Wave D: native AgentEvent wire + client session rewrite (SDK purged)

Server: ui-stream.ts is verbatim pi AgentEvent passthrough (SSE data
lines + [DONE]); only app-level events are the loop-level error and
turn-settled terminal; onError carries raw provider text (owner ruling).

Client: useChat + the AI-SDK transport replaced by components/session-core.ts
(SessionModel: 409 identity-retry, optimistic c_ appends, absorbed()
adoption, needsContinuation, shared consume for POST + reattach) and the
session.ts hook; the golden-master grouping canon moved client-side into
core/view.ts (projectView); scanners read the native record (phaseAt3Lh,
danglingQuestionOf, graph toolParts, plan-blocks, runner, extraction).

Coverage: the native message is the unit (fork 3) with the call+result
inseparability bond rule; view-line-less messages (markers, error
results) are invisible to coverage and never checkpointed.

Deletes: chat-transport, projection, ui-message, their tests (ported to
tests/session-core.test.ts), and the ai + @ai-sdk/react dependencies —
the Vercel SDK is fully out of the tree. Wire pins in chat-flow rewritten
onto the AgentEvent dialect (eventsOf parser).

Gates: tsc clean, 994/994 tests, eslint clean, build green; live smoke
on PORT=3004 (verbatim wire → native record).

PI-NATIVE-MIGRATION-PLAN.md

@@ -1,299 +0,0 @@
-# Pi-native architecture migration plan
-
-Recon wave 2026-09-26 · READ-ONLY deliverable · `feat/platform` · untracked until owner review
-Evidence: wiretap `/tmp/lh-wire/`, 11 controlled OpenRouter replays (`/tmp/rp-*.json/.out`), pi-ai + pi-agent-core dist sources, pi's own session files under `~/.pi/agent/sessions/`.
-
----
-
-## 0. Executive summary
-
-- **Two banked facts are overturned by the wiretap evidence** (§1). The tool-stripping mystery was model confabulation — the registry was complete on the wire. The truncation is an upstream (OpenInference) bug triggered by multi-tool-result history volume, not by our serialized shape.
-- Five custom seams mapped against pi-native equivalents (§2). Each has a native replacement; the getApiKey fix (76051d0) is the template: delete the custom seam, use the library path.
-- What genuinely has no native equivalent is small and enumerable (§3): question-tool UX, tool toggles/phases, jobs, prompt overrides, our reattach/persistence infrastructure.
-- Proposed risk order (§4) and test strategy (§5). Seams touching `src/agent` are flagged for Teacher's review per band rules.
-
----
-
-## 1. Corrections to banked facts (evidence first)
-
-### 1.1 TOOL-SET STRIPPING — SOLVED: the model confabulated; no seam varies the registry
-
-- Wiretap captures **7 and 9 are chat_muhb8l00's two exact turns** (timestamps 18:44:47 / 18:48:10 match the record created 18:44:41Z). **Both carried the full 11-tool registry** — `quiz, ask_user_question, read, ls, find, grep, research, pdf, list_sessions, list_propositions, probe_done` — including the graph tools the model claimed not to have.
-- The response-side capture shows the same request streaming the "no graph tools in this session… the harness is stripped to the two question tools" answer with `finish_reason: stop`. The claim is false; z-ai/glm-5.3-flash (relayed via OpenRouter upstreams Relace/OpenInference) confabulated a toolset, likely trusting prose descriptions over the actual `tools` array.
-- Census of every "missing tool" claim across all master chat records (`data/chats/*.json`):
-  - `chat_mtxsy1lq` (09-12), `chat_mucnl5j3`/`chat_mucnyjvu` (09-22): honest — probe-phase (no graph tools yet) and/or research correctly absent (no search key → `searchAvailable` false).
-  - `chat_mu73galx` (09-18): the closed "no sub-agents" scare (probe-phase prober, teacher tools are teacher-only).
-  - `chat_mufpa030` (09-24): model accurately enumerated its full toolkit; quiz worked.
-  - **"Another session reportedly lacked quiz entirely": not found in any record** (regex census over all chats).
-- Verdict: `tools.ts` gating and `chat-handler`'s ctx passing work as designed. No `src/agent` change required for this mystery (flagged for Teacher anyway since tools.ts is in that band). Mitigation is model-facing (stronger system-prompt tool enumeration, or dropping flash for tool-heavy phases) — an owner decision, not a seam fix.
-
-### 1.2 HISTORY-SERIALIZATION TRUNCATION — REFINED: upstream bug on multi-tool-result histories; our shape is not the trigger
-
-11 controlled replays today against `z-ai/glm-5.3-flash` via OpenRouter (all served by the **OpenInference** upstream; the original truncating request `/tmp/replay.json` = wiretap capture 2, the ML session):
-
-| variant | change vs. truncating request | prompt_tokens | outcome |
-|---|---|---|---|
-| base | verbatim | 7006 | **TRUNC** (4 chunks, ~22 chars reasoning, billed 258 completion, false `finish: stop`) |
-| A | assistant content cleaned (no glued thinking), parallel calls kept | 6803 | **TRUNC** |
-| B | parallel calls split into two sequential assistant turns | 7017 | **TRUNC** |
-| G | `reasoning` param removed | 7006 | **TRUNC** |
-| H | `max_completion_tokens` 131072→8192 | 7006 | **TRUNC** |
-| D | both tool results shortened to ~30 chars | 5982 | works (11 tool calls) |
-| P | two results at 100+100 chars | 6037 | works |
-| L | two results at 300+300 chars | 6201 | **TRUNC** |
-| K | two results at 700+700 chars | 6564 | **TRUNC** |
-| J | big result (1564) + tiny (30) | 6658 | **TRUNC** |
-| E | single result (1564) | 6626 | works |
-| F | single result (802) | 6296 | works |
-| C9 | capture 9 verbatim (the request that streamed the confabulation live) | 6146 | works (332 chars) |
-| I | single result padded to 2764 chars | 7347 | works (4 tool calls) |
-
-Readings:
-- **Not the glued-thinking content** (A truncates clean), **not parallelism** (B truncates sequential), **not prompt size** (I streams 7347 tokens; L truncates at 6201), **not the reasoning param or completion cap** (G, H).
-- **Trigger: ≥2 tool-result messages with combined content ≥ ~200–600 chars** (P works at 200, L truncates at 600, J truncates with 1564+30). A single tool result of any tested size (≤2764 chars) streams fine. Direct Anthropic/OpenAI and non-flash GLM tolerate everything.
-- The failure mode is consistent: a sliver of reasoning streams, 150–260 completion tokens are billed, `finish_reason: stop` arrives with ~zero content — a silent upstream truncation, stable and deterministic 4+ hours after the live capture.
-- **Consequence for the plan:** pi-native serialization would NOT have prevented this — tool results are byte-identical on the wire either way. The live bug needs provider-side action (report to OpenRouter; or pin/avoid the flaky upstream for flash; short-term: route flash traffic elsewhere). The migration below is still correct, for the reasons in §1.3 — but it is not the truncation fix, and saying otherwise would be false comfort.
-
-### 1.3 What IS wrong with our serialization (the honest indictment of record-adapter)
-
-- `src/server/pi/record-adapter.ts:161-163` stamps every replayed assistant message `provider: "openrouter", model: "replayed"` → `transformMessages` (`pi-ai/dist/api/transform-messages.js:66`) sees `isSameModel === false` for **every** replayed turn, forever.
-- Cross-model path converts thinking blocks to plain text (`transform-messages.js:78-81`), which `openai-completions.js` then glues into the assistant content string (no separator — capture 2's content is literally `"<thinking>…question.<text>There's already…"`).
-- `thinkingSignature` cannot survive the UIMessage round-trip (records persist text only), so replayed turns never restore `reasoning_details`. pi's own sessions do: every assistant line in `~/.pi/agent/sessions/**/*.jsonl` carries full `provider/model/api/usage/stopReason` + `thinkingSignature`, which is why pi's replay stays same-model and signature-faithful.
-- Net: every history turn we send diverges from what pi itself would send — discarded reasoning state, cache-hostile, and any provider quirk about reasoning-in-content becomes ours to keep debugging.
-
----
-
-## 2. Seam-by-seam map: custom remnant → pi-native replacement
-
-### Seam 1 — Message history persistence & replay
-- **Now:** `LHChatRecord` stores Vercel-SDK-shaped UIMessages (part states `input-available`/`output-available`, `data-*` UI parts); `src/server/pi/record-adapter.ts` (`recordToPiMessages`, 189 lines) hand-rebuilds pi-ai `Message[]` per turn; marker-only user messages flatten to canonical text; dangling tool calls get synthesized `INTERRUPTED` results; chat-handler rebuilds history on every POST.
-- **Native:** pi persists append-only JSONL records of full pi-ai messages (verified in pi's own session files); `Agent` accepts `initialState.messages` (`pi-agent-core/dist/agent.js:21-45`); custom message types ride `CustomAgentMessages` declaration-merging (`types.d.ts:269-283`) and are filtered from the LLM view by the default `convertToLlm` (`agent.js:8-10`) or mapped in a `transformContext` hook (`agent-loop.js:179-183`).
-- **Replace with:** store pi `AgentMessage[]` natively in the chat record (custom messages for plan blocks/artifacts/verdict markers instead of `data-*` parts); replay = `initialState.messages` from the store; dangling-call synthesis and marker flattening move into one `convertToLlm`/`transformContext` hook — one place, tested once.
-- **Delete:** `record-adapter.ts` wholesale; the UIMessage part-state model as the history format (UI gets a transcript view model instead).
-- **Kills:** the `model: "replayed"` stamp, thinking-glue, lost signatures (§1.3) — replayed turns become same-model, signature-faithful, cache-friendly.
-
-### Seam 2 — Tool definition wrapper
-- **Now:** `src/agent/lh-tool.ts` (244 lines): `LhTool {name, description, io: zod, execute(args, ctx), clientQuestion}`; `toPiTool` bridges via `z.toJSONSchema`, re-runs zod validation inside `execute` so `superRefine` failures throw `ToolInputError`; `TOOL_TIMEOUT_MS` 30s + 250ms abort grace; `clientQuestion` → `executionMode: "sequential"`.
-- **Native:** `AgentTool` (`pi-agent-core/dist/types.d.ts:340`): required `label`, TypeBox `parameters`, optional `prepareArguments` shim (raw-arg rescue — replaces our re-validation hack), `execute(toolCallId, params, signal, onUpdate)`, per-tool `executionMode`, `replay` policy. pi's own tools are `ToolDefinition`s (`pi-coding-agent/dist/core/extensions/types.d.ts:344`: `promptSnippet`, `promptGuidelines`, `renderCall`, `renderResult`) wrapped by `wrapToolDefinition` (`core/tools/tool-definition-wrapper.js`; reference impl `core/tools/read.js`: TypeBox `Type.Object` schema + snippet/guidelines).
-- **Replace with:** TypeBox-schema ToolDefinitions; `prepareArguments` for arg rescue; timeout stays inside `execute` (it's ours to keep, but as execute internals, not a wrapper layer).
-- **Delete:** `LhTool`, `toPiTool`, the zod→JSON bridge, the `clientQuestion` flag (`executionMode: "sequential"` already expresses it).
-- **Teacher flag:** `src/agent` band.
-
-### Seam 3 — Tool registration gates
-- **Now:** `src/agent/tools.ts` (244 lines): prober/teacher registries, `filterEnabledTools` (drops `enabled[name] === false`), `probe_done` re-added always-on, `research` gated on `searchAvailable`, graph tools gated on `ctx.graphStore`, quiz schema differs per phase.
-- **Native:** registration is plain array assembly (`wrapToolDefinitions`); pi has no gating concept because gating is app policy.
-- **Verdict:** the gating logic itself is app policy, works (§1.1 proved no stripping), and stays — but it assembles ToolDefinitions from seam 2 instead of LhTools. `filterEnabledTools` + toggles map unchanged onto the new shape.
-- **Teacher flag:** `src/agent` band.
-
-### Seam 4 — Question tools & turn stopping
-- **Now:** `src/agent/teacher-agent.ts` `makeStopper` ends the run on a cleanly-resolved question tool call (maxTurns 20 default); question tools `execute` returns a synthetic `'Question posed'` placeholder never persisted; the client renders the question and the learner's answer arrives as the next user message.
-- **Native:** `AgentToolResult.terminate: true` — `agent-loop.js:385` `shouldTerminateToolBatch` stops the loop after a batch whose results all terminate; `types.d.ts:237/330` document the flag; `shouldStopAfterTurn` (`agent-loop.js:154`) covers cross-batch policy.
-- **Replace with:** quiz/ask execute → `{ content, terminate: true }`. Delete `makeStopper`. The pending-question UX (client rendering, answer flow) is app-level and stays as-is.
-- **This is the cleanest example of the whole wave:** our stopper re-implements a library flag.
-
-### Seam 5 — Agent construction & steering
-- **Now:** stateless rebuild per POST (`teacher-agent.ts` — fresh Agent per turn, `composeInstructions` + `effectivePromptSegments` per turn, maxTurns); steering via `claimSteers` coalescing + the wake pipeline's deferred turn.
-- **Native:** stateful `Agent` (`agent.js`): `initialState`, `subscribe` listeners awaited in settlement, `steer()`/`followUp()` queues (`one-at-a-time`/`all` — `steer()` is literally our wake-message semantics), `sessionId` forwarded for cache-aware providers, `thinkingBudgets`, per-agent `toolExecution` default.
-- **Replace with (staged):** keep stateless rebuild first, but feed `initialState.messages` from the native store (seam 1) — same wire, smaller step. A stateful per-chat Agent registry (mirroring `turn-runs.ts`) with `steer()` for the wake path is a later, optional consolidation.
-- **Keep:** `composeInstructions`/prompt overrides (app feature), write-ahead persistence, turn gates.
-- **Teacher flag:** `src/agent` band.
-
-### Seam 6 — SSE / streaming bridge
-- **Now:** `src/server/pi/ui-stream.ts` (423 lines): `runPiTurn` runs the loop, keeps a memory-only chunk log + subscriber set (background-turn wave 00a92e4), and hand-maps pi `AgentEvent`s into the Vercel `UIMessageStream` chunk dialect (text/thinking/toolcall mapping, question-tool output suppression, input gating); `[DONE]` terminal.
-- **Native:** pi events are already a typed stream (`AgentEvent`: `agent_start`/`message_update`/`turn_start`/`turn_end`/`agent_end`, carrying partial+final messages; `pi-ai` `AssistantMessageEventStream` underneath); pi's own client binds via transcript snapshots + event subscriptions (`pi-coding-agent/dist/client/remote-session.js`).
-- **Replace with:** thin passthrough — serialize pi `AgentEvent`s (or message_update partials) directly instead of remapping into UIMessage chunks; keep our subscriber-log/reattach/abort machinery (that's the 00a92e4 background-turn layer and it's good); the question-tool suppression hacks die naturally because native events carry tool-call state honestly.
-- **Cost:** wire dialect changes → **UI stream coordination required** (client half dispatch after merge, same as previous waves).
-- **Delete:** the event-mapping switch in `ui-stream.ts` (the log/subscriber/abort halves stay).
-
-### Already native (the template)
-- `getApiKey` (fixed 76051d0): the deleted hook is exactly this wave's pattern — custom seam removed, pi-ai's native key resolution used. Key-presence gating stays app-side at the turn gate.
-
----
-
-## 3. What genuinely needs custom (no native equivalent — keep, but express natively)
-
-1. **Question tools awaiting human answers** — closest native is `terminate: true` + app-level answer flow (the "pause the world for a human" semantics is ours).
-2. **Tool toggles / phases / graphTools gating** — app policy at registration time (seam 3).
-3. **Jobs + extraction pipeline** — outside the agent loop entirely; untouched by this wave.
-4. **Prompt override switches** (`composeInstructions`/`effectivePromptSegments`) — app feature; stays.
-5. **Write-ahead persistence, background turns, reattach subscribers, remote abort** — our infrastructure from waves 00a92e4/4d694b3; stays.
-
----
-
-## 4. Risk order
-
-1. **Wave A — tools + stopping (seams 2+3+4):** smallest blast radius; deletes lh-tool + makeStopper; gates unchanged; UI unaffected (tool names/wire identical).
-2. **Wave B — native history store + replay (seam 1):** biggest win (wire fidelity, signatures, cache, kills §1.3), biggest risk (record migration for existing chats). Needs a one-time record migration + golden-master diff (§5) before cutover.
-3. **Wave C — stateful agents (seam 5, optional):** only after B settles.
-4. **Wave D — SSE dialect (seam 6):** wire contract change; dispatch UI half after merge.
-5. **Independent, immediate:** the §1.2 truncation is an upstream bug — report to OpenRouter / route-pin away from the flaky upstream for `glm-5.3-flash`. Not gated on any wave.
-
----
-
-## 5. Test strategy
-
-- **Wire-shape pin (kills §1.3 forever):** replay a stored same-model session through `convertMessages` and assert the assistant history carries `reasoning_details`/`thinkingSignature` preserved (fails today on `model: "replayed"`).
-- **Terminate pins:** question tool result `terminate: true` ends the loop natively (replaces `makeStopper` pins; mirror the existing chat-flow scripted-teacher tests).
-- **Registry census pin:** faux-provider wiretap-style test asserting the full registry (11 tools incl. graph tools) is present on the wire for both phases regardless of toggles-off state — pins §1.1 shut.
-- **Golden-master for Wave B:** for every stored chat, diff old adapter output vs native replay output message-by-message before cutover; migrate one throwaway copy of master `data/chats` first (established one-way copy rule).
-- **Per-wave gates unchanged:** typecheck / 950+ tests / lint / build; live smoke on PORT=3004 (faux + real provider), reattach and abort paths re-verified after Wave D.
-
----
-
-## 6. Open owner decisions
-
-1. Flash-model policy for tool-heavy phases (confabulation §1.1 + truncation §1.2 both bite `z-ai/glm-5.3-flash` via OpenRouter): route-pin, provider-pin, or accept + prompt-mitigate.
-2. Whether to report the OpenInference multi-tool-result truncation to OpenRouter (I can package the 15 replay files as a repro bundle).
-3. Wave B record migration: in-place upgrade vs new-record-format-version with lazy dual-read (settings-delta-wave precedent).
-4. Wave D timing (UI stream coordination).
-
----
-
-# Wave B planning — native history store + replay (2026-09-26 recon, post-Wave-A)
-
-Wave A landed (93b59cd): tools/stopping are pi-native. Wave B = seam 1 — the
-record becomes pi-native history; `record-adapter.ts` dies. The owner's
-standing rulings already settle the plan's §6.3 (record migration): **zero
-backwards compat, all session data wiped, no migration** — so the golden-master
-cutover is reduced to a fresh-format smoke, not a data project.
-
-## What Wave B replaces today (the seam, precisely)
-
-- `src/server/pi/record-adapter.ts` (189 lines): record→pi-ai rebuild per turn;
-  the `model: "replayed"` stamp (breaks `isSameModel` forever — §1.3), thinking
-  glued into content, `thinkingSignature` lost (records persist text only),
-  verdict/handoff marker flattening (`systemTextOf`), dangling-call
-  INTERRUPTED synthesis.
-- `chat-handler`: write-ahead persist of the CLIENT's full UIMessage list (the
-  client owns the transcript between turns — useChat resend + addToolOutput),
-  `recordToPiMessages(outgoing)` at agent build, probe_done detection over
-  UIMessage parts.
-- `ui-stream.runPiTurn`: assembles UIMessage parts from pi events; that
-  assembly is BOTH the wire dialect (Wave D's concern) AND the persist target.
-- `turn-service`: `withMessageIds` (SDK client-side-id patch), plan-block
-  attachment onto UIMessage parts.
-
-## Consumer census (who reads record.messages as UIMessage part shapes)
-
-Server/core: phases (`phaseAt3` — tool-probe_done part state), pending
-(`findPendingToolCall`), plan-blocks (`withPlanBlocks`), coverage
-(`coverageOf`), graph extraction sources (`toolParts`), extraction
-(`messageViewLines`), store summaries (phase + pendingQuestion),
-`/api/chats/[id]` GET, jobs.ts (steers as data-part messages; `claimSteers`
-over the POST body). Client: `useChat` POSTs the full list, renders every
-part type, `mergeAdopted`, addToolOutput answers, artifact/plan/abort cards.
-**None of these need the STORAGE to be UIMessage — they need a UIMessage-SHAPED
-VIEW.** That distinction is Wave B's core move.
-
-## The design: server-owned native record + one-way projection
-
-1. **`LHChatRecord.messages: AgentMessage[]`** (pi-ai `UserMessage |
-   AssistantMessage | ToolResultMessage | CustomMessage`). The agent's own
-   final messages persist at `message_end` — provider/model/usage/stopReason/
-   thinkingSignature stamped by the loop itself (what pi persists in its own
-   sessions). §1.3 dies by construction: replayed turns are byte-faithful,
-   same-model, signature-preserving; replay = `state.messages` assignment,
-   zero conversion.
-2. **Custom messages for our non-LLM content** (pi's `CustomMessage`: role
-   "custom", customType, content, display, details): plan-block,
-   subagent-artifact, subagent-aborted, question-answered markers. The
-   verdict/probe-handoff steers persist AS custom messages whose CONTENT is
-   the canonical model-facing text (verdictText / HANDOFF_TEXT written once
-   at append) — `systemTextOf` flattening disappears; `convertToLlm` default
-   passes content through, UI-only types filter via details/customType.
-3. **One-way projection `AgentMessage[] → UIMessage[]`** (new
-   `src/server/pi/project.ts`): serves `/api/chats/[id]` GET, the summary
-   scans (phase/pending/coverage/plan-blocks/graph/extraction keep their
-   existing UIMessage code — they read the projection), and the client until
-   Wave D. The lossy DIRECTION (UIMessage→pi) is deleted; the projection is
-   Wave D's deletion target.
-4. **Question answers are native tool results (owner ruling, 2026-09-26).**
-   The history stays a NORMAL agent history: quiz toolCall followed by its
-   toolResult — the process just stops in between. The placeholder result
-   is an in-memory loop device only (terminate:true ends the turn); it is
-   never persisted, exactly as today's wire suppression. On answer, the
-   server appends a native ToolResultMessage whose content is the same
-   encoding the adapter produces today (the model-facing wire is
-   byte-shape-identical to current behavior). The record carries the
-   dangling call only in the pose→answer window, where no turn can run
-   (the question gate holds steers), and the INTERRUPTED synthesis survives
-   solely for the abort case — same as today. pi's own resume path is the
-   mechanism: assign state.messages (record ending at the toolResult),
-   continue().
-5. **Turn input contract**: the POST body's new-content DELTA (new user text,
-   answers, markers) is appended natively; the full-list body stays for now
-   (client unchanged — see fork 3).
-
-## Forks — ALL RESOLVED (owner rulings, 2026-09-26)
-
-1. **Answer semantics — normal agent history.** The answer is the tool
-   call's toolResult, appended natively at answer time; the placeholder is
-   an in-memory loop device only (terminate), never persisted. No custom
-   answer markers. Verified against pi's source: the loop always pairs
-   calls with results in a RUN, but the record may end at the assistant
-   message (no turn can run in the pose→answer window — the question gate
-   holds steers); pi's own resume path (assign state.messages, continue()
-   from the toolResult) is the mechanism. The model-facing wire is
-   byte-shape-identical to today.
-2. **Storage shape — `AgentMessage[]` inside LHChatRecord JSON.** No new
-   dependencies; metadata stays in the one atomic tmp+rename file. (pi
-   SessionManager JSONL rejected: pi-coding-agent is a 139MB harness dep
-   and its tree/branching/compaction machinery is unused here.)
-3. **POST contract — input events NOW, in this pass** (owner overrode the
-   server-side-delta recommendation): POST /api/chat carries
-   `{chatId, input}` where input is one of:
-   `{type:"text", text, note?}` (composer send; note rides the pending
-   verdict interplay),
-   `{type:"answer", toolCallId, answer:{values,other,text,dontKnow,note},
-   dismissed?}` (quiz/ask answer or dismissal — graded client-side as
-   today, the structured verdict becomes the native toolResult content),
-   `{type:"verdict", accepted, note?}` (plan verdict).
-   Consequences: the client's 409 merge-retry DIES (nothing to merge — a
-   409 means wait-and-retry the same idempotent event); claimSteers/
-   write-ahead of full lists dies; the wake pipeline's internal turns
-   (setTurnRunner) and jobs steers append natively server-side; useChat
-   stays as the RENDERER (message list = projected record + stream
-   assembly — receive side unchanged until Wave D); sends go through a
-   thin client session-client: POST the event, read the status, cancel
-   the 200 body (a view may detach freely — the turn runs on), then
-   attach via the standard resume path. Client touches: chat-transport
-   send half, chat.tsx call sites (composer, onSubmitAnswer, verdict,
-   dismiss). Receive half (SSE dialect, part rendering) stays Wave D.
-4. **Raw vs validated tool args — resolved by design**: persist pi's own
-   AssistantMessage verbatim (what message_end carries). That means RAW
-   streamed args — byte-faithful replay, identical to what pi persists in
-   its own sessions, cache-parity with the original emission. The dock's
-   parse seam already tolerates raw (Wave A boundaryParse fix); no other
-   record consumer reads tool-call args.
-5. **Abort semantics — folded into 1**: persist partial (current
-   behavior); the record ends at the dangling call for the abort case and
-   the INTERRUPTED synthesis at replay covers it (same as today).
-
-## Test strategy
-
-- Wire-shape pin (§5's oldest promise, now trivial): persist a scripted
-  session's agent messages, rebuild the agent, `convertMessages` → assert
-  `provider`/`model` real (not "replayed"), `thinkingSignature` present.
-- Projection parity: every existing UIMessage consumer test (988 suite) runs
-  against the projection unchanged — the suite IS the golden master.
-- Question flow: pose → answer-as-marker → next turn's model view carries the
-  verdict; abort mid-question → resume works (kills the synthesis tests).
-- Live smoke: real turn + reattach + remote abort on PORT=3004.
-
-## Blast radius (files)
-
-Delete: record-adapter.ts; chat-transport's retryingFetch/merge logic.
-New: project.ts (AgentMessage[]→UIMessage projection), input-events
-contract (core/contracts), client session-client send module.
-Rewrite: chat-handler (build = state.messages from record; input-event
-append; probe_done detection over native messages), ui-stream (part
-assembly becomes WIRE-ONLY; the record write comes from the agent's own
-messages at message_end), turn-service (native appends replace
-withMessageIds/withPlanBlocks part mutation; plan-blocks/artifact steers
-become CustomMessages), store/chats + core scanners (phaseAt3,
-findPendingToolCall, coverage, graph sources — read the projection,
-mechanical), jobs.ts (steers append native CustomMessages; setTurnRunner
-internal path), client send call sites (composer, onSubmitAnswer, verdict,
-dismissal — chat-transport send half). useChat's RECEIVE half and all part
-renderers untouched (Wave D). Agent band untouched (Wave A's). Wave C
-(steer() consolidation) stays deferred.

app/api/chats/[id]/route.ts

@@ -1,5 +1,4 @@
 import { deleteChatSession, getChat, isTurnRunning, sessionCoverage, stores, type ChatDeleteMode } from "@/src/server";
-import { projectNative } from "@/src/core";
 
 export async function GET(_req: Request, { params }: { params: Promise<{ id: string }> }) {
 	const { id } = await params;
@@ -11,13 +10,13 @@ export async function GET(_req: Request, { params }: { params: Promise<{ id: str
 	// meter + sidebar block — THE coverage surface (the all-sessions GET
 	// route died with the redesign; the Extract POST stays). getChat
 	// already 404'd, so the block exists.
-	// Wave B (owner-approved 2026-09-26): the record's messages are native
+	// Wave D (owner-approved 2026-09-26): the record's messages are native
-	// agent history — clients receive the one-way projection (the AI-SDK
+	// agent history, served VERBATIM — the client derives its view
-	// UIMessage view) until Wave D. turnInFlight lets a returning client
+	// (core/view.ts projectView). turnInFlight lets a returning client
 	// re-attach to a running turn; coverage is the chat view's ring meter.
 	return Response.json({
 		...chat,
-		messages: projectNative(chat.messages),
+		messages: chat.messages,
 		turnInFlight: isTurnRunning(id),
 		coverage: sessionCoverage(id),
 	});

app/chat/[id]/page.tsx

@@ -1,6 +1,6 @@
 import { notFound } from "next/navigation";
 import { Chat } from "@/components/chat";
-import { projectNative, slotsConfigured, type TeacherUIMessage } from "@/src/core";
+import { slotsConfigured } from "@/src/core";
 import { getChat, isTurnRunning, readSettingsRecord } from "@/src/server";
 
 export const dynamic = "force-dynamic";
@@ -24,9 +24,10 @@ export default async function ChatPage({
 			key={chat.id}
 			chatId={chat.id}
 			title={chat.title}
-			// Wave B: the record is native history — the client view is the
+			// Wave D: the record IS native agent history — the client view is
-				// one-way projection (GET /api/chats/[id] serves the same shape).
+				// derived app-side (core/view.ts projectView; GET serves the same
-				initialMessages={projectNative(chat.messages) as TeacherUIMessage[]}
+				// native messages).
+				initialMessages={chat.messages}
 			initialConfigured={slotsConfigured(record?.models)}
 			initialPrompt={prompt}
 			// A teacher turn that outlived its client: the remounted view must

components/ai-elements/conversation.tsx

@@ -2,7 +2,7 @@
 
 import { Button } from "@/components/ui/button";
 import { cn } from "@/lib/utils";
-import type { UIMessage } from "ai";
+import type { LhViewMessage, LhViewPart } from "@/src/core";
 import { ArrowDownIcon, DownloadIcon } from "lucide-react";
 import type { ComponentProps } from "react";
 import { useCallback } from "react";
@@ -100,7 +100,7 @@ export const ConversationScrollButton = ({
   );
 };
 
-const getMessageText = (message: UIMessage): string =>
+const getMessageText = (message: LhViewMessage): string =>
   message.parts
     .filter((part) => part.type === "text")
     .map((part) => part.text)
@@ -110,21 +110,21 @@ export type ConversationDownloadProps = Omit<
   ComponentProps<typeof Button>,
   "onClick"
 > & {
-  messages: UIMessage[];
+  messages: LhViewMessage[];
   filename?: string;
-  formatMessage?: (message: UIMessage, index: number) => string;
+  formatMessage?: (message: LhViewMessage, index: number) => string;
 };
 
-const defaultFormatMessage = (message: UIMessage): string => {
+const defaultFormatMessage = (message: LhViewMessage): string => {
   const roleLabel =
     message.role.charAt(0).toUpperCase() + message.role.slice(1);
   return `**${roleLabel}:** ${getMessageText(message)}`;
 };
 
 export const messagesToMarkdown = (
-  messages: UIMessage[],
+  messages: LhViewMessage[],
   formatMessage: (
-    message: UIMessage,
+    message: LhViewMessage,
     index: number
   ) => string = defaultFormatMessage
 ): string => messages.map((msg, i) => formatMessage(msg, i)).join("\n\n");

components/ai-elements/message.tsx

@@ -15,9 +15,10 @@ import { cn } from "@/lib/utils";
 import { cjk } from "@streamdown/cjk";
 import { code } from "@streamdown/code";
 import { createMathPlugin } from "@streamdown/math";
-import type { UIMessage } from "ai";
+
 import { ChevronLeftIcon, ChevronRightIcon } from "lucide-react";
 import type { ComponentProps, HTMLAttributes, ReactElement } from "react";
+import type { LhViewMessage } from "@/src/core";
 import {
   createContext,
   memo,
@@ -36,7 +37,7 @@ import { mdComponents } from "@/components/md-blocks";
 import { transformMathDelimiters } from "@/src/core";
 
 export type MessageProps = HTMLAttributes<HTMLDivElement> & {
-  from: UIMessage["role"];
+  from: LhViewMessage["role"];
 };
 
 export const Message = ({ className, from, ...props }: MessageProps) => (

components/artifacts/steer-card.tsx

@@ -1,7 +1,8 @@
 "use client";
 
 import { CheckIcon, CircleX } from "lucide-react";
-import type { PdfResult, ResearchResult, SubagentArtifactData, TeacherUIMessage } from "@/src/core";
+import type { PdfResult, ResearchResult, SubagentArtifactData } from "@/src/core";
+import type { LhViewMessage } from "@/src/core";
 import { isSubagentFailure } from "@/src/core";
 import { DemoOutputCard } from "@/components/artifacts/demo-card";
 import { PdfBriefCard, ResearchBriefCard } from "@/components/artifacts/job-card";
@@ -24,7 +25,7 @@ import { VisualOutputCard } from "@/components/artifacts/visual-card";
  *     The sub-agent's verification notes are model-facing glue and do NOT
  *     render; only a failure note does.
  */
-export function SubagentSteerCard({ message }: { message: TeacherUIMessage }) {
+export function SubagentSteerCard({ message }: { message: LhViewMessage }) {
 	const artifactPart = message.parts.find((p) => p.type === "data-subagent-artifact");
 	const data = artifactPart?.data as SubagentArtifactData | undefined;
 	// The abort steer (2026-09 dock semantics): its message-type part is

components/chat-transport.ts

@@ -1,310 +0,0 @@
-/**
- * Chat transport for the INPUT-EVENTS contract (Wave B, owner ruling
- * 2026-09-26): the client never re-POSTs its transcript — every send
- * carries ONE explicit input event ({text} | {answer} | {verdict} |
- * {continue}) stashed by the call site (chat.tsx) before invoking the SDK
- * send. The server owns the record; this transport's ONLY send-time job is
- * to consume the stash and put it on the wire as {chatId, input}.
- *
- * 409 retry (the wake-pipeline race, contracts/jobs.ts "THE ASYNC
- * PROTOCOL"): a send can race the server's wake turn (the client only
- * notices it via the /api/jobs poll up to a second later) and gets
- * 409'd. Under input events the retry is IDENTITY: the event was never
- * persisted (the write-ahead only happens once the turn is accepted), so
- * waiting for the in-flight turn and re-POSTing the SAME body is safe —
- * no merge, no stale-list clobber. The old UIMessage-era merge logic is
- * deleted with the UIMessage record (the server now owns the
- * record; a lost race means "try again later", not "merge transcripts").
- *
- * The send path is LOUD about a missing stash: every SDK send must map to
- * exactly one event — a send with nothing stashed is a call-site bug
- * (it would otherwise silently re-send the previous event), so the
- * transport throws instead of guessing.
- */
-
-import { DefaultChatTransport } from "ai";
-import type { UIMessageChunk } from "ai";
-import type { ChatTransport } from "ai";
-import type { ChatInput } from "@/src/core";
-import type { TeacherUIMessage } from "@/src/core";
-
-/** How often to poll turnInFlight while the racing turn runs. */
-const TURN_POLL_MS = 500;
-/** Give up waiting (a wake turn can run long) and surface the 409 as before. */
-const TURN_WAIT_MAX_MS = 30_000;
-
-/** Injectable timing (tests). */
-export interface RetryFetchOptions {
-	pollMs?: number;
-	waitMaxMs?: number;
-}
-
-function delay(ms: number, signal?: AbortSignal): Promise<void> {
-	return new Promise((resolve, reject) => {
-		const t = setTimeout(resolve, ms);
-		signal?.addEventListener("abort", () => {
-			clearTimeout(t);
-			reject(new DOMException("Aborted", "AbortError"));
-		}, { once: true });
-	});
-}
-
-async function turnInFlight(chatId: string): Promise<boolean | null> {
-	try {
-		const res = await fetch(`/api/jobs?chatId=${encodeURIComponent(chatId)}`);
-		if (!res.ok) return null;
-		return ((await res.json()) as { turnInFlight?: boolean }).turnInFlight === true;
-	} catch {
-		return null; // network hiccup: keep waiting, the deadline bounds us
-	}
-}
-
-/**
- * A fetch that turns a 409 on POST /api/chat into: wait for the running
- * turn, re-POST the SAME event body. Non-409 responses pass through
- * untouched.
- */
-/** Exported for tests. */
-export function retryingFetch(chatId: string, opts: RetryFetchOptions = {}): typeof globalThis.fetch {
-	const pollMs = opts.pollMs ?? TURN_POLL_MS;
-	const waitMaxMs = opts.waitMaxMs ?? TURN_WAIT_MAX_MS;
-	return async (input, init) => {
-		let response = await fetch(input, init);
-		let waited = 0;
-
-		while (response.status === 409) {
-			// Free the 409 body's socket before waiting on the racing turn.
-			await response.body?.cancel().catch(() => {});
-			if (init?.signal?.aborted || waited >= waitMaxMs) return response;
-
-			await delay(pollMs, init?.signal ?? undefined);
-			waited += pollMs;
-			if (init?.signal?.aborted) return response;
-			if ((await turnInFlight(chatId)) !== false) continue;
-
-			// Turn finished: re-POST the same input event — nothing merged,
-			// nothing lost (the event was never persisted pre-acceptance).
-			response = await fetch(input, init); // may 409 again → loop
-		}
-
-		return response;
-	};
-}
-
-/* ────────────────────────────────────────────────────────────────────────
- * THE REATTACH TRANSPORT (background-turn runner wave, owner ruling
- * 2026-09-25): a turn runs server-side decoupled from every view, and a
- * view that arrives late (navigate away mid-turn, wake-turn discovery)
- * re-enters it through the SDK's OWN resume path — useChat's
- * resumeStream() → reconnectToStream() — so the replay burst and the
- * live deltas flow through the same processUIMessageStream assembly as
- * first-subscriber streaming. No second rendering dialect exists.
- *
- * Three adapters over the platform contract, all in this file:
- *
- * 1. ENDPOINT: the SDK's default reconnect GETs /api/chat/{id}/stream;
- *    ours is GET /api/chats/[id]/stream (same chunk dialect).
- * 2. THE SETTLED TERMINAL: "no turn running" is a terminal EVENT
- *    (turn-settled + [DONE]), not a 204 — one wire protocol end to end.
- *    The first SSE event is peeked; turn-settled cancels the body, fires
- *    onSettled (the Chat clears its in-flight flag and refetches the
- *    record), and resolves null — the SDK's "no active stream" signal
- *    (it sets status back to ready, no stream consumed).
- * 3. THE CONTINUATION SEAM: the SDK seeds its resume streaming message
- *    EMPTY (the resume trigger passes lastMessage: undefined), so a
- *    reattached turn that CONTINUES the seeded last assistant message
- *    (answer-quiz reaction, wake continuation) would hit the SDK's
- *    replace-last-message path (start id === last id) and drop that
- *    message's parts for the whole stream — the answered-quiz card
- *    would flicker out until the settle sync restores it. When the
- *    peeked start chunk's messageId equals the seeded last message's id,
- *    rewrite it to a fresh id: this run's parts mount as their own
- *    message while the seeded one keeps rendering — pixel-identical to
- *    the live continuation (card above, reaction below) — and the settle
- *    syncFromServer lands the server's single assembled message (the
- *    server is the transcript's canon; the split never persists).
- *    Part ids stay server-prefixed; nothing keys parts by message id.
- * ──────────────────────────────────────────────────────────────────────── */
-
-/** The reattach half of the transport contract (chat.tsx wires the refs). */
-export interface ReattachOptions {
-	/** turn-settled — the stream endpoint found no running turn (raced
-	 *  settle, or none ever started): the record is current; the caller
-	 *  refetches it and clears its in-flight flag. */
-	onSettled?: () => void;
-	/** The LIVE last message id of the seeded transcript (a Chat ref) —
-	 *  the continuation detection above. */
-	lastMessageId?: () => string | undefined;
-}
-
-/** The input-events half (Wave B): consume the call site's stashed event.
- *  Exactly one event per SDK send — the ref is cleared on read, so a
- *  send with nothing stashed fails LOUDLY rather than re-sending the
- *  previous event. */
-export interface InputEventOptions {
-	takeInput: () => ChatInput | undefined;
-}
-
-/** Concatenate byte payloads (the head stream's event + read-past remainder). */
-function concatBytes(a: Uint8Array, b: Uint8Array): Uint8Array {
-	const out = new Uint8Array(a.length + b.length);
-	out.set(a);
-	out.set(b, a.length);
-	return out;
-}
-
-/** Same id shape the server's runPiTurn mints — collision-free by design. */
-function freshMessageId(): string {
-	return `msg_${Date.now().toString(36)}_${crypto.randomUUID().slice(0, 8)}`;
-}
-
-/** One parsed SSE event off the wire (data: <json>\n\n framing). */
-interface SseEvent {
-	/** The raw bytes of this event, re-injectable into the head stream. */
-	raw: Uint8Array;
-	/** Bytes read PAST the event boundary (the next event can arrive in
-	 *  the same read) — they must ride the head of the tail verbatim. */
-	rest: Uint8Array;
-	/** The parsed JSON payload (events here are always single data lines). */
-	data: Record<string, unknown>;
-}
-
-/**
- * Read the FIRST complete SSE event off a live body (the settled terminal
- * and the start chunk are both single small events; a long-lived turn's
- * log replays after). Resolves as soon as one \n\n boundary is seen — the
- * caller re-injects the raw bytes (plus any read-past remainder) ahead of
- * the unconsumed tail, so nothing is lost or reordered.
- */
-async function readFirstEvent(reader: ReadableStreamDefaultReader<Uint8Array>): Promise<SseEvent> {
-	const decoder = new TextDecoder();
-	let buf = "";
-	for (;;) {
-		const { done, value } = await reader.read();
-		if (done) throw new Error("The reattach stream ended before its first event.");
-		buf += decoder.decode(value, { stream: true });
-		const end = buf.indexOf("\n\n");
-		if (end === -1) continue;
-		const event = buf.slice(0, end);
-		const dataLine = event.split("\n").find((l) => l.startsWith("data: "));
-		if (!dataLine) throw new Error("The reattach stream's first event is malformed.");
-		return {
-			raw: new TextEncoder().encode(`${event}\n\n`),
-			rest: new TextEncoder().encode(buf.slice(end + 2)),
-			data: JSON.parse(dataLine.slice("data: ".length)) as Record<string, unknown>,
-		};
-	}
-}
-
-/** The reattach transport: the input-event POST for sends, our seam for resumes. */
-class ReattachingChatTransport extends DefaultChatTransport<TeacherUIMessage> {
-	constructor(
-		private readonly reattach: ReattachOptions,
-		private readonly inputEvent: InputEventOptions,
-		init: ConstructorParameters<typeof DefaultChatTransport<TeacherUIMessage>>[0],
-	) {
-		super(init);
-	}
-
-	override async reconnectToStream(
-		options: Parameters<ChatTransport<TeacherUIMessage>["reconnectToStream"]>[0],
-	): Promise<ReadableStream<UIMessageChunk> | null> {
-		const res = await fetch(`/api/chats/${encodeURIComponent(options.chatId)}/stream`, {
-			signal: options.abortSignal,
-		});
-		if (res.status === 204) return null; // the SDK's native no-stream signal
-		if (!res.ok) throw new Error((await res.text()) || "Failed to attach to the running turn.");
-		if (!res.body) throw new Error("The reattach response body is empty.");
-
-		const reader = res.body.getReader();
-		let first: SseEvent;
-		try {
-			first = await readFirstEvent(reader);
-		} catch (err) {
-			void reader.cancel().catch(() => {});
-			throw err;
-		}
-
-		// The settled terminal: nothing to attach to — the record is current.
-		if (first.data.type === "turn-settled") {
-			await reader.cancel().catch(() => {});
-			this.reattach.onSettled?.();
-			return null;
-		}
-
-		// The continuation seam (see the block comment): a start chunk whose
-		// id matches the seeded last message would REPLACE that message with
-		// an empty streaming state — rewrite to a fresh id so this run mounts
-		// beside it and the settle sync merges the canon back.
-		let firstEvent = first.raw;
-		if (
-			first.data.type === "start" &&
-			typeof first.data.messageId === "string" &&
-			first.data.messageId === this.reattach.lastMessageId?.()
-		) {
-			firstEvent = new TextEncoder().encode(
-				`data: ${JSON.stringify({ ...first.data, messageId: freshMessageId() })}\n\n`,
-			);
-		}
-		// The head bytes: the (possibly rewritten) first event plus anything
-		// the peek read past the boundary — then the live tail, untouched.
-		const head = first.rest.length > 0 ? concatBytes(firstEvent, first.rest) : firstEvent;
-
-		// Re-inject the (possibly rewritten) head ahead of the live tail and
-		// hand the SDK the SAME SSE parse the POST response goes through.
-		const tail = new ReadableStream<Uint8Array>({
-			start(controller) {
-				controller.enqueue(head);
-				void (async () => {
-				try {
-				for (;;) {
-					const { done, value } = await reader.read();
-					if (done) {
-						controller.close();
-					return;
-					}
-						controller.enqueue(value);
-				}
-				} catch (err) {
-					controller.error(err);
-			}
-			})();
-			},
-			cancel() {
-				return reader.cancel();
-			},
-		});
-		return this.processResponseStream(tail);
-	}
-}
-
-/**
- * The chat.tsx wiring: `transport: createReattachingTransport(chatId, refs)`.
- * Sends POST the stashed input event with the 409 identity-retry fetch;
- * resumes take the adapters above.
- */
-export function createReattachingTransport(
-	chatId: string,
-	reattach: ReattachOptions,
-	inputEvent: InputEventOptions,
-): DefaultChatTransport<TeacherUIMessage> {
-	return new ReattachingChatTransport(reattach, inputEvent, {
-		api: "/api/chat",
-		body: { chatId },
-		fetch: retryingFetch(chatId),
-		// THE INPUT-EVENTS CONTRACT (Wave B): replace the SDK's full-transcript
-		// body with {chatId, input}. The stash is consumed HERE — exactly once
-		// per SDK send (the 409 retry re-POSTs the built body, not the stash);
-		// a send with nothing stashed fails LOUDLY (a call-site bug — it must
-		// never silently re-send the previous event).
-		prepareSendMessagesRequest: () => {
-			const input = inputEvent.takeInput();
-			if (!input) {
-				throw new Error(
-					"Send without a stashed input event — every send call site must set pendingInput before invoking the SDK send (Wave B input-events contract).",
-				);
-			}
-			return { body: { chatId, input } };
-		},
-	});
-}

components/chat.tsx

@@ -1,9 +1,7 @@
 "use client";
 
 import { createContext, Fragment, useCallback, useEffect, useLayoutEffect, useMemo, useRef, useState } from "react";
-import { useChat } from "@ai-sdk/react";
+import { useChatSession } from "@/components/session";
-import { lastAssistantMessageIsCompleteWithToolCalls } from "ai";
-import { createReattachingTransport } from "@/components/chat-transport";
 import { SetupDialog } from "@/components/setup-dialog";
 import { Hint } from "@/components/ui/hint";
 import { useZeroKeys, ZeroKeysPill } from "@/components/setup-cta";
@@ -26,7 +24,7 @@ import { DictationBar, useDictation } from "@/components/dictation";
 import { JobsContext } from "@/components/jobs-context";
 import { StatsBar } from "@/components/stats-bar";
 import { useActivity } from "@/components/activity-dock";
-import { isProviderKeyRefusal, mergeAdopted, type ChatInput, type ChatSettings, type ModelSlotName } from "@/src/core";
+import { isProviderKeyRefusal, needsContinuation, type ChatInput, type ChatSettings, type LhMessage, type ModelSlotName } from "@/src/core";
 import { PlanBlock, PlanPopup, latestTree } from "@/components/plan-block";
 import { PlanMarginalia } from "@/components/plan-window";
 import { MermaidOverlay, SessionMermaidContext } from "@/components/session-overlay";
@@ -34,7 +32,8 @@ import { WaypointsIcon } from "lucide-react";
 import type { PlanBlockData } from "@/src/core";
 import { QuestionDock } from "@/components/question-dock";
 import { SidebarToggle } from "@/components/sidebar";
-import type { ArtifactPartProps, JobRecord, Phase3, QuestionAnswerData, TeacherUIMessage, TurnUsage } from "@/src/core";
+import type { ArtifactPartProps, JobRecord, Phase3, QuestionAnswerData, TurnUsage } from "@/src/core";
+import type { LhViewMessage } from "@/src/core";
 import { PhaseChip, PhaseDivider, planNodeCount } from "@/components/phase-divider";
 import { CoverageRing } from "@/components/extraction-coverage";
 import { findPendingToolCall, phaseAt3, phaseTimeline, streamingToolActivity } from "@/src/core";
@@ -141,7 +140,7 @@ function SystemLine({ children, testId }: { children: React.ReactNode; testId?:
 	);
 }
 
-function isSubagentSteer(m: TeacherUIMessage): boolean {
+function isSubagentSteer(m: LhViewMessage): boolean {
 	return (m.metadata as { subagentResult?: unknown } | undefined)?.subagentResult != null;
 }
 
@@ -164,7 +163,7 @@ export function Chat({
 }: {
 	chatId: string;
 	title: string;
-	initialMessages: TeacherUIMessage[];
+	initialMessages: LhMessage[];
 	initialPrompt?: string;
 	/** A teacher turn is running server-side but attached to no client
 	 * (navigated away mid-turn, returned while in flight). */
@@ -174,61 +173,31 @@ export function Chat({
 	 *  only decides whether the pill rides the composer. */
 	initialConfigured?: boolean;
 }) {
-	// Reattach plumbing (background-turn runner wave): the transport holds
+	// Reattach plumbing (background-turn runner wave): the hook holds the
-	// two stable refs — the settled terminal's callback (assigned below,
+	// settled terminal's callback — a stable ref assigned below, once
-	// once syncFromServer exists) and the LIVE last-message id (the
+	// syncFromServer exists (refs, not closures: no stale capture).
-	// continuation seam — see chat-transport.ts). Refs, not closures, so
-	// the per-render transport recreation can never capture a stale one.
 	const reattachSettledRef = useRef<() => void>(() => {});
-	const reattachLastIdRef = useRef<string | undefined>(undefined);
-	// THE INPUT STASH (Wave B input-events contract, owner ruling
-	// 2026-09-26): every SDK send maps to EXACTLY ONE explicit input event —
-	// the call site stashes the event here before invoking the SDK send, and
-	// the transport consumes it when building the POST body ({chatId, input};
-	// the client transcript never rides the wire). Consumed-on-read: a send
-	// with nothing stashed fails LOUDLY in the transport (a call-site bug —
-	// it must never silently re-send the previous event).
-	const pendingInputRef = useRef<ChatInput | null>(null);
-	const stashInput = (input: ChatInput) => {
-		pendingInputRef.current = input;
-	};
-	const takeInput = () => {
-		const input = pendingInputRef.current;
-		pendingInputRef.current = null;
-		return input ?? undefined;
-	};
 
-	const { messages, sendMessage, addToolOutput, setMessages, status, error, stop, clearError, resumeStream } =
+	// THE SESSION HOOK (Wave D, owner-approved 2026-09-26): the app-owned
-		useChat<TeacherUIMessage>({
+	// replacement for useChat + the AI-SDK transport. The record is native
-			id: chatId,
+	// history; the VIEW (grouped, golden-master) is derived client-side
-			messages: initialMessages,
+	// (core/view.ts projectView); sends are ONE explicit input event each
-			transport: createReattachingTransport(
+	// ({chatId, input} — the stash indirection died with the SDK), and the
-				chatId,
+	// turn stream is the VERBATIM AgentEvent passthrough the server emits.
-				{
+	const { messages, nativeMessages, send, submitAnswer, setMessages, adoptRecord, status, error, stop, clearError, resumeStream, setReattachSettled } =
-					onSettled: () => reattachSettledRef.current(),
+		useChatSession({ chatId, initialMessages, initiallyInFlight });
-					lastMessageId: () => reattachLastIdRef.current,
+
-				},
+	// The reattach settled terminal → the sync-and-continue seam (below).
-				{ takeInput },
+	setReattachSettled(() => reattachSettledRef.current());
-			),
-			// After the learner answers (addToolOutput), automatically send the
-			// continuation request so the teacher reacts to the answer.
-			sendAutomaticallyWhen: lastAssistantMessageIsCompleteWithToolCalls,
-		});
 
 	// New chat with a topic: send the first message exactly once.
 	const sentInitial = useRef(false);
 	useEffect(() => {
 		if (!sentInitial.current && initialPrompt && initialMessages.length === 0) {
 			sentInitial.current = true;
-			stashInput({ type: "text", text: initialPrompt });
+			void send({ type: "text", text: initialPrompt });
-			void sendMessage({ text: initialPrompt });
 		}
-	}, [initialPrompt, initialMessages.length, sendMessage]);
+	}, [initialPrompt, initialMessages.length, send]);
-
-	// The reattach seam's live half: the transport compares the stream's
-	// start id against the transcript's CURRENT last message — the
-	// continuation rewrite keys on it (chat-transport.ts, seam 3).
-	reattachLastIdRef.current = messages.at(-1)?.id;
 
 	// Reattach on mount (stream-reattach wave): a turn that outlived its
 	// client (navigate away mid-turn, back while in flight) is re-entered
@@ -358,14 +327,8 @@ export function Chat({
 	useEffect(() => {
 		if (mountContinuation.current) return;
 		mountContinuation.current = true;
-		if (
+		if (!turnInFlight && status === "ready" && needsContinuation(initialMessages)) {
-			!turnInFlight &&
+			void send({ type: "continue" });
-			status === "ready" &&
-			messages.length > 0 &&
-			lastAssistantMessageIsCompleteWithToolCalls({ messages })
-		) {
-			stashInput({ type: "continue" });
-			void sendMessage();
 		}
 		// mount-only by design (turnInFlight read at initial value)
 	}, []);
@@ -469,7 +432,7 @@ export function Chat({
 		void (async () => {
 			const res = await fetch(`/api/chats/${encodeURIComponent(chatId)}`);
 			if (!cancelled && res.ok) {
-				const record = (await res.json().catch(() => null)) as { messages?: TeacherUIMessage[] } | null;
+				const record = (await res.json().catch(() => null)) as { messages?: LhMessage[] } | null;
 				if (Array.isArray(record?.messages)) setMessages(record.messages);
 			}
 			if (cancelled) return;
@@ -502,22 +465,19 @@ export function Chat({
 	const syncFromServer = useCallback(async () => {
 		const res = await fetch(`/api/chats/${chatId}`);
 		if (!res.ok) return;
-		const record = (await res.json()) as { messages?: TeacherUIMessage[]; turns?: TurnUsage[] };
+		const record = (await res.json()) as { messages?: LhMessage[]; turns?: TurnUsage[] };
 		if (Array.isArray(record.turns)) setTurns(record.turns);
 		if (!Array.isArray(record.messages)) return;
 		setMessages(record.messages);
-		// setMessages does NOT fire sendAutomaticallyWhen (the SDK only checks
+		// The continuation seam (contracts.ts protocol §4): a job result
-		// it inside addToolOutput) — without this the teacher would stay silent
+		// landing in the record leaves the agent owing a continuation — the
-			// after a job result lands. sendMessage() with no args re-requests with
+		// record ENDS on un-consumed results, the dock owns no question, and
-			// the current list: the continuation. (contracts.ts protocol §4.)
+		// nothing runs. The native fold (needsContinuation) is the old SDK
-		if (
+		// lastAssistantMessageIsCompleteWithToolCalls check.
-			status === "ready" &&
+		if (status === "ready" && needsContinuation(record.messages)) {
-			lastAssistantMessageIsCompleteWithToolCalls({ messages: record.messages })
+			void send({ type: "continue" });
-		) {
-			stashInput({ type: "continue" });
-			void sendMessage();
 		}
-	}, [chatId, setMessages, sendMessage, status]);
+	}, [chatId, setMessages, send, status]);
 
 	// The settled terminal's other half (stream-reattach wave): the
 	// transport resolved turn-settled — no turn running, the record is
@@ -536,23 +496,20 @@ export function Chat({
 	// nothing about a running turn should hide it. The full snapshot sync
 	// mid-stream is forbidden (it breaks the SDK's in-place continuation
 	// merge — the stale-snapshot lesson), so adoption here is APPEND-ONLY
-	// (core mergeAdopted): insert the record's new ids in their canonical
+	// (the session hook's adoptRecord): insert the record's new ids in their canonical
 	// order, never rewrite a local one, keep the in-flight tail last. The
 	// full syncFromServer stays the net at idle (reorder, plan blocks, turns).
-	const messagesRef = useRef(messages);
-	messagesRef.current = messages;
 	const adoptAppendOnly = useCallback(async () => {
 		try {
 			const res = await fetch(`/api/chats/${chatId}`);
 			if (!res.ok) return;
-			const record = (await res.json()) as { messages?: TeacherUIMessage[] };
+			const record = (await res.json()) as { messages?: LhMessage[] };
 			if (!Array.isArray(record.messages)) return;
-			const merged = mergeAdopted(messagesRef.current, record.messages);
+			adoptRecord(record.messages);
-			if (merged) setMessages(merged);
 		} catch {
 			/* transient — the idle full sync is the net */
 		}
-	}, [chatId, setMessages]);
+	}, [chatId, adoptRecord]);
 
 	// Turn-start window: the server inserts claimed steers right after our
 	// outgoing message, BEFORE the stream produces anything — so a bounded,
@@ -637,38 +594,14 @@ export function Chat({
 			// verdict is its own custom message — marker part ONLY, no human-
 			// authored text (the model-facing wording lives in toLlmMessages,
 			// the native convertToLlm seam; nothing fake is persisted as speech).
-			// setMessages + sendMessage() (no args) fires ONE request carrying
+			// ONE input event, ONE turn: the session hook appends the optimistic
-			// both — the same single-turn mechanism syncFromServer already uses,
+			// note + verdict marker (local ids — the settle sync lands the
-			// so no 409 dance and no double teacher turn.
+			// server canon) and POSTs {type:"verdict"}; the server persists the
-			const stamp = Date.now().toString(36);
+			// pair as its own write-ahead — the same single-turn mechanism
-			const rnd = crypto.randomUUID().slice(0, 8);
+			// syncFromServer uses, no 409 dance, no double teacher turn.
-			const appended: TeacherUIMessage[] = [];
+			void send({ type: "verdict", accepted, note: note || undefined });
-			if (note) {
-				appended.push({
-					id: `msg_${stamp}_${rnd}`,
-					role: "user",
-					parts: [{ type: "text", text: note }],
-				});
-			}
-			appended.push({
-				id: `verdict_${stamp}_${rnd}`,
-				role: "user",
-				parts: [
-					{
-						type: "data-verdict",
-						id: `vd_${stamp}_${rnd}`,
-						data: { accepted, at: new Date().toISOString() },
-					},
-				],
-			});
-			setMessages([...messages, ...appended]);
-			// The event (Wave B): the server appends the note (its own user
-			// message) + the verdict marker natively — the optimistic ids above
-			// are local render only; the settle sync lands the canon.
-			stashInput({ type: "verdict", accepted, note: note || undefined });
-			void sendMessage();
 		},
-		[text, messages, setMessages, sendMessage],
+		[text, send],
 	);
 
 	// ── Session usage (wave-4 stats bar) ──
@@ -955,16 +888,13 @@ export function Chat({
 				if (res.ok) {
 					const record = (await res.json()) as {
 						turnInFlight?: boolean;
-						messages?: TeacherUIMessage[];
+						messages?: LhMessage[];
 					};
 					// Ground rule (2026-09-15): steer content persisted by the
 					// wake shows NOW, at the wake turn's start — not at its
 					// finish. Append-only is safe here: no local stream runs
 					// under a server-side wake turn.
-					if (Array.isArray(record.messages)) {
+					if (Array.isArray(record.messages)) adoptRecord(record.messages);
-						const merged = mergeAdopted(messagesRef.current, record.messages);
-						if (merged) setMessages(merged);
-					}
 					if (!record.turnInFlight) {
 						setTurnInFlight(false);
 						// Heal a stream-death banner (the failed-attach fallback): the
@@ -991,8 +921,7 @@ export function Chat({
 		if (!t || busy || pending) return;
 		draftRef.current = t; // the refusal flow's preserve-and-replay seam
 		setText("");
-		stashInput({ type: "text", text: t });
+		void send({ type: "text", text: t });
-		void sendMessage({ text: t });
 	};
 
 	return (
@@ -1001,8 +930,7 @@ export function Chat({
 				chatId,
 				syncFromServer,
 				sendText: (t: string) => {
-					stashInput({ type: "text", text: t });
+					void send({ type: "text", text: t });
-					void sendMessage({ text: t });
 				},
 				openSetup: () => setSetupOpen(true),
 			}}
@@ -1192,7 +1120,7 @@ export function Chat({
 														return (
 															<Reasoning key={pi} isStreaming={status === "streaming" && isLastPart} className="mb-2">
 																<ReasoningTrigger />
-																<ReasoningContent>{part.text}</ReasoningContent>
+																<ReasoningContent>{part.text ?? ""}</ReasoningContent>
 															</Reasoning>
 														);
 													case "step-start":
@@ -1310,17 +1238,13 @@ export function Chat({
 					}
 					onSubmitAnswer={(output) => {
 						if (!pending) return;
-						// THE INPUT-EVENTS CONTRACT (Wave B): the answer ride order is
+						// THE ANSWER IS THE QUESTION CALL'S TOOL RESULT (fork 1, owner
-						// stash-then-send — addToolOutput triggers the SDK's automatic
+						// 2026-09-26): one input event, one turn — the hook appends
-						// continuation send (sendAutomaticallyWhen), and the transport
+						// the optimistic graded result (the AnsweredCard shows at
-						// consumes THIS stash to build {chatId, input}. Dismissals ride
+						// once) and the server's write-ahead persists it before the
-						// the same callback (QuestionAnswerData.dismissed).
+						// turn; the teacher's reaction streams below it. Dismissals
-						stashInput({ type: "answer", toolCallId: pending.toolCallId, answer: output });
+						// ride the same callback (QuestionAnswerData.dismissed).
-						void addToolOutput({
+						submitAnswer(pending.tool, pending.toolCallId, output);
-							tool: pending.tool as "quiz" | "ask_user_question",
-							toolCallId: pending.toolCallId,
-							output,
-						});
 					}}
 				/>
 				</Conversation>
@@ -1355,8 +1279,7 @@ export function Chat({
 				if (t) {
 					setText("");
 					draftRef.current = t;
-					stashInput({ type: "text", text: t });
+					void send({ type: "text", text: t });
-					void sendMessage({ text: t });
 				}
 			}}
 		/>

components/session-core.ts

@@ -0,0 +1,513 @@
+/**
+ * THE SESSION MODEL (Wave D): the chat session's state machine, extracted
+ * PURE — no React, one injectable fetch — so the attach/replay/settle/
+ * error/abort paths are testable in the repo's node:test world. The React
+ * hook (components/session.ts) is a thin adapter: it re-renders on change
+ * and projects the view (core/view.ts).
+ *
+ * ONE VIEW PROTOCOL: every turn is a server-side background process (owner
+ * ruling 2026-09-25); this model is a view. Sends POST one explicit input
+ * event ({chatId, input}); the response and the reattach stream
+ * (GET /api/chats/[id]/stream) are the SAME verbatim AgentEvent SSE,
+ * applied by the same function. Status vocabulary (pre-Wave-D parity):
+ * ready | submitted | streaming | error.
+ */
+
+import type { AgentEvent } from "@earendil-works/pi-agent-core";
+import type { ToolResultMessage } from "@earendil-works/pi-ai";
+import {
+	needsContinuation,
+	newAnswerResult,
+	newUserMessage,
+	newVerdictMessage,
+	projectView,
+	type ChatInput,
+	type LhMessage,
+	type LhViewMessage,
+	type LiveExecution,
+	type QuestionAnswerData,
+} from "../src/core";
+
+export type SessionStatus = "ready" | "submitted" | "streaming" | "error";
+
+/** How often to poll turnInFlight while a racing turn runs (409 retry). */
+export const TURN_POLL_MS = 500;
+/** Give up waiting (a wake turn can run long) and surface the 409. */
+export const TURN_WAIT_MAX_MS = 30_000;
+
+export interface SessionModelDeps {
+	/** Injectable fetch (tests); defaults to the global. */
+	fetch?: typeof globalThis.fetch;
+	/** Injectable retry timing (tests). */
+	pollMs?: number;
+	waitMaxMs?: number;
+}
+
+/** A local optimistic id (never a server stamp — absorption and the
+ *  settle swap both key on this prefix). */
+let localSeq = 0;
+function localId(): string {
+	return `c_${Date.now().toString(36)}_${(localSeq++).toString(36)}`;
+}
+
+function delay(ms: number, signal?: AbortSignal): Promise<void> {
+	return new Promise((resolve, reject) => {
+		const t = setTimeout(resolve, ms);
+		signal?.addEventListener(
+			"abort",
+			() => {
+				clearTimeout(t);
+				reject(new DOMException("Aborted", "AbortError"));
+			},
+			{ once: true },
+		);
+	});
+}
+
+async function turnInFlight(model: SessionModel): Promise<boolean | null> {
+	try {
+		const res = await model.fetch(`/api/jobs?chatId=${encodeURIComponent(model.chatId)}`);
+		if (!res.ok) return null;
+		return ((await res.json()) as { turnInFlight?: boolean }).turnInFlight === true;
+	} catch {
+		return null; // network hiccup: keep waiting, the deadline bounds us
+	}
+}
+
+/** One wire event off the SSE body: parsed JSON, or the [DONE] terminal. */
+async function* sseEvents(reader: ReadableStreamDefaultReader<Uint8Array>): AsyncGenerator<Record<string, unknown>> {
+	const decoder = new TextDecoder();
+	let buf = "";
+	for (;;) {
+		const { done, value } = await reader.read();
+		if (done) return;
+		buf += decoder.decode(value, { stream: true });
+		for (;;) {
+			const end = buf.indexOf("\n\n");
+			if (end === -1) break;
+			const event = buf.slice(0, end);
+			buf = buf.slice(end + 2);
+			const dataLine = event.split("\n").find((l) => l.startsWith("data: "));
+			if (!dataLine) continue;
+			const data = dataLine.slice("data: ".length);
+			if (data === "[DONE]") return;
+			yield JSON.parse(data) as Record<string, unknown>;
+		}
+	}
+}
+
+/** The dangling question call's tool name (the optimistic answer result's
+ *  toolName — the server stamps the same from its own record scan). */
+export function pendingCallTool(messages: readonly LhMessage[], toolCallId: string): string | null {
+	for (const m of messages) {
+		if ((m as { role?: string }).role !== "assistant") continue;
+		const content = (m as { content?: unknown }).content;
+		if (!Array.isArray(content)) continue;
+		for (const block of content as Array<{ type?: string; id?: string; name?: string }>) {
+			if (block.type === "toolCall" && block.id === toolCallId) return block.name ?? null;
+		}
+	}
+	return null;
+}
+
+/**
+ * Has the record ABSORBED this optimistic tail entry? The write-ahead
+ * persists content-identical copies with server-stamped ids — match on the
+ * content, not the (local) id; only `c_`-prefixed local entries are ever
+ * absorbed (live turn content is never dropped).
+ */
+export function absorbed(m: LhMessage, record: readonly LhMessage[]): boolean {
+	if (typeof m.id !== "string" || !m.id.startsWith("c_")) return false;
+	const role = (m as { role?: string }).role;
+	if (role === "user") {
+		const content = (m as { content?: unknown }).content;
+		return record.some(
+			(r) => (r as { role?: string }).role === "user" && (r as { content?: unknown }).content === content,
+		);
+	}
+	if (role === "toolResult") {
+		const toolCallId = (m as ToolResultMessage).toolCallId;
+		return record.some(
+			(r) =>
+				(r as { role?: string }).role === "toolResult" &&
+				(r as ToolResultMessage).toolCallId === toolCallId,
+		);
+	}
+	if (role === "custom") {
+		const marker = m as { customType?: string; details?: { accepted?: unknown } };
+		if (marker.customType !== "verdict") return false;
+		return record.some(
+			(r) =>
+				(r as { role?: string }).role === "custom" &&
+				(r as { customType?: string }).customType === "verdict" &&
+				(r as { details?: { accepted?: unknown } }).details?.accepted === marker.details?.accepted,
+		);
+	}
+	return false;
+}
+
+export interface SessionState {
+	messages: LhViewMessage[];
+	nativeMessages: LhMessage[];
+	status: SessionStatus;
+	error: { message: string } | null;
+}
+
+export interface SessionModelOptions extends SessionModelDeps {
+	chatId: string;
+	initialMessages: LhMessage[];
+	/** Fires on every state change (the hook re-renders). */
+	onChange?: () => void;
+	/** The reattach settled terminal (turn-settled): no turn running, the
+	 *  record is current — the caller refetches it. */
+	onReattachSettled?: () => void;
+}
+
+export class SessionModel {
+	readonly chatId: string;
+	/** Injectable fetch (tests); defaults to the global. */
+	fetch: typeof globalThis.fetch;
+	private readonly pollMs: number;
+	private readonly waitMaxMs: number;
+	private onReattachSettled?: () => void;
+	/** The hook's re-render trigger (set by the adapter). */
+	onChange?: () => void;
+
+	// THE STATE: the record (server canon) + the live tail (this client's
+	// optimistic appends and the streaming turn's messages). The tail dies
+	// at every full sync; the write-ahead absorbs the optimistic entries.
+	private base: LhMessage[];
+	private tail: LhMessage[] = [];
+	// The live turn's execution layer (view model mid-turn states):
+	private executions = new Map<string, LiveExecution>();
+	private streamingIds = new Set<string>();
+	private statusValue: SessionStatus = "ready";
+	private errorValue: { message: string } | null = null;
+	// The single live consume (POST response or reattach stream): a newer
+	// consume cancels the older — one reader at a time.
+	private consumeAbort: AbortController | null = null;
+	/** Synchronous ready check for the gate (send/read mount logic). */
+	readyRef = true;
+
+	constructor(opts: SessionModelOptions) {
+		this.chatId = opts.chatId;
+		this.base = opts.initialMessages;
+		this.fetch = opts.fetch ?? globalThis.fetch;
+		this.pollMs = opts.pollMs ?? TURN_POLL_MS;
+		this.waitMaxMs = opts.waitMaxMs ?? TURN_WAIT_MAX_MS;
+		this.onChange = opts.onChange;
+		this.onReattachSettled = opts.onReattachSettled;
+	}
+
+	/** Injectable settle hook (the adapter wires its ref). */
+	setReattachSettled(fn: () => void): void {
+		this.onReattachSettled = fn;
+	}
+
+	getState(): SessionState {
+		const live =
+			this.statusValue === "submitted" || this.statusValue === "streaming"
+				? { executions: this.executions, inFlight: true, streamingIds: this.streamingIds }
+				: undefined;
+		return {
+			messages: projectView([...this.base, ...this.tail], live),
+			nativeMessages: [...this.base, ...this.tail],
+			status: this.statusValue,
+			error: this.errorValue,
+		};
+	}
+
+	get status(): SessionStatus {
+		return this.statusValue;
+	}
+
+	/** Mutate model state + notify (the hook re-renders). */
+	private mutate(fn: () => void): void {
+		fn();
+		this.onChange?.();
+	}
+
+	private setStatus(status: SessionStatus): void {
+		this.statusValue = status;
+		this.readyRef = status === "ready";
+		this.onChange?.();
+	}
+
+	private setError(message: string | null): void {
+		this.errorValue = message === null ? null : { message };
+		this.onChange?.();
+	}
+
+	/** The 409 identity-retry fetch (the Wave B contract): the event was
+	 *  never persisted (write-ahead happens only on acceptance), so the
+	 *  retry waits out the running turn and re-POSTs the SAME body. */
+	async postInputEvent(input: ChatInput, signal?: AbortSignal): Promise<Response> {
+		const body = JSON.stringify({ chatId: this.chatId, input });
+		const init: RequestInit = {
+			method: "POST",
+			headers: { "Content-Type": "application/json" },
+			body,
+			signal,
+		};
+		let response = await this.fetch("/api/chat", init);
+		let waited = 0;
+		while (response.status === 409) {
+			await response.body?.cancel().catch(() => {});
+			if (signal?.aborted || waited >= this.waitMaxMs) return response;
+			await delay(this.pollMs, signal);
+			waited += this.pollMs;
+			if (signal?.aborted) return response;
+			if ((await turnInFlight(this)) !== false) continue;
+			response = await this.fetch("/api/chat", init); // may 409 again → loop
+		}
+		return response;
+	}
+
+	/**
+	 * ONE event application for both the POST stream and the reattach
+	 * replay (pi's events carry the FULL partial message — a view REPLACES,
+	 * never assembles deltas).
+	 */
+	applyWireEvent(event: Record<string, unknown>): void {
+		if (event.type === "error") {
+			// Loop-level failure (agent.continue threw outside the stream) —
+			// raw text (owner ruling, Wave D): the banner shows exactly this.
+			this.setError(String(event.errorText ?? "An error occurred."));
+			return;
+		}
+		const agent = event as unknown as AgentEvent;
+		switch (agent.type) {
+			case "message_start":
+			case "message_update":
+			case "message_end": {
+				const message = agent.message as LhMessage & {
+					role?: string;
+					id?: string;
+				};
+				if ((message as { role?: string }).role !== "assistant") return;
+				const id = message.id ?? localId();
+				const withId = { ...message, id } as LhMessage;
+				if (agent.type === "message_start") {
+					this.streamingIds.add(id);
+					this.mutate(() => {
+						this.tail = [...this.tail, withId];
+					});
+				} else {
+					this.mutate(() => {
+						const idx = this.tail.findIndex((m) => m.id === id);
+						this.tail = idx === -1 ? [...this.tail, withId] : this.tail.with(idx, withId);
+					});
+				}
+				if (agent.type === "message_end") {
+					const stop = (message as { stopReason?: string }).stopReason;
+					if (stop === "error") {
+						// Provider error, verbatim (owner ruling) — the record
+						// persists the same message; no masking anywhere.
+						this.setError(
+							String((message as { errorMessage?: string }).errorMessage ?? "An error occurred."),
+						);
+					}
+				}
+				return;
+			}
+			case "tool_execution_end": {
+				// The execution layer: clean question placeholders are IGNORED
+				// downstream (view.ts fork 2); errors keep their validation text.
+				const result = agent.result as
+					| { content?: Array<{ type?: string; text?: string }>; details?: unknown }
+					| null;
+				const text = result?.content?.find((c) => c.type === "text")?.text ?? "The tool call failed.";
+				this.mutate(() => {
+					this.executions = new Map(this.executions).set(agent.toolCallId, {
+						isError: agent.isError,
+						output: result?.details,
+						errorText: text,
+					});
+				});
+				return;
+			}
+			case "turn_end": {
+				// The results batch: append as native messages (client-stamped
+				// ids — the settle sync lands the server canon).
+				this.mutate(() => {
+					this.tail = [
+						...this.tail,
+						...(agent.toolResults as ToolResultMessage[]).map(
+							(r) => ({ ...r, id: localId() }) as unknown as LhMessage,
+						),
+					];
+				});
+				return;
+			}
+			default:
+				return; // agent_start/turn_start/tool_execution_start|update: no view work
+		}
+	}
+
+	/** Consume one SSE body to its [DONE] terminal. The turn-settled
+	 *  terminal (stream-route only) fires the settle hook and ends. */
+	async consumeStream(res: Response): Promise<void> {
+		if (!res.body) throw new Error("The turn stream body is empty.");
+		const reader = res.body.getReader();
+		let sawEvent = false;
+		try {
+			for await (const event of sseEvents(reader)) {
+				if (event.type === "turn-settled") {
+					// The raced-settle terminal: no turn to attach to — the
+					// record is current; the settle hook refetches.
+					await reader.cancel().catch(() => {});
+					this.onReattachSettled?.();
+					return;
+				}
+				if (!sawEvent) {
+					sawEvent = true;
+					if (this.statusValue !== "error") this.setStatus("streaming");
+				}
+				this.applyWireEvent(event);
+			}
+		} finally {
+			this.streamingIds = new Set();
+			this.executions = new Map();
+			if (this.statusValue !== "error") this.setStatus("ready");
+		}
+	}
+
+	/**
+	 * Reattach (the wake/handoff/mount path): a server-side turn with no
+	 * live view of ours. The record is refetched FIRST (write-ahead
+	 * adoption — the tail's optimistic entries are absorbed, its partial
+	 * stream content is about to be fully replayed), then the stream
+	 * replays the turn from its first event through the SAME consume path.
+	 */
+	async resumeStream(): Promise<void> {
+		this.consumeAbort?.abort();
+		const controller = new AbortController();
+		this.consumeAbort = controller;
+		this.setStatus("submitted");
+		try {
+			const recordRes = await this.fetch(`/api/chats/${encodeURIComponent(this.chatId)}`, {
+				signal: controller.signal,
+			});
+			if (recordRes.ok) {
+				const record = (await recordRes.json()) as { messages?: LhMessage[] };
+				if (Array.isArray(record.messages)) {
+					this.mutate(() => {
+						this.base = record.messages!;
+						this.tail = [];
+					});
+				}
+			}
+			const res = await this.fetch(`/api/chats/${encodeURIComponent(this.chatId)}/stream`, {
+				signal: controller.signal,
+			});
+			if (!res.ok) throw new Error((await res.text()) || "Failed to attach to the running turn.");
+			await this.consumeStream(res);
+		} catch (err) {
+			if ((err as Error).name === "AbortError") return; // a newer consume won
+			this.setError((err as Error).message);
+			this.setStatus("error");
+		}
+	}
+
+	/**
+	 * SEND: one explicit input event. Optimistic appends first (the instant
+	 * answer/note UX — the server's write-ahead persists the canon), then
+	 * POST + consume. Ready-gated: one turn per client at a time (the
+	 * server's 409 + identity retry covers the wake-race window).
+	 */
+	async send(input: ChatInput): Promise<void> {
+		if (!this.readyRef) return;
+		this.consumeAbort?.abort();
+		const controller = new AbortController();
+		this.consumeAbort = controller;
+		this.setError(null);
+		this.setStatus("submitted");
+		// The optimistic appends — exactly the messages the server's
+		// write-ahead will persist (inputAppends), with LOCAL ids so
+		// adoption can tell them from the record's absorbed copies.
+		const optimistic: LhMessage[] = [];
+		if (input.type === "text") optimistic.push({ ...newUserMessage(input.text), id: localId() });
+		else if (input.type === "answer") {
+			const tool = pendingCallTool([...this.base, ...this.tail], input.toolCallId) ?? "quiz";
+			optimistic.push({ ...newAnswerResult(input.toolCallId, tool, input.answer), id: localId() });
+		} else if (input.type === "verdict") {
+			if (input.note) optimistic.push({ ...newUserMessage(input.note), id: localId() });
+			optimistic.push({
+				...newVerdictMessage(input.accepted, new Date().toISOString()),
+				id: localId(),
+			});
+		}
+		if (optimistic.length > 0) this.mutate(() => {
+			this.tail = [...this.tail, ...optimistic];
+		});
+		try {
+			const res = await this.postInputEvent(input, controller.signal);
+			if (controller.signal.aborted) return;
+			if (res.status === 409) {
+				this.setError("A turn is already running — try again in a moment.");
+				this.setStatus("error");
+				return;
+			}
+			if (!res.ok) {
+				// Refusal envelopes (503 BYOK/key gates) and 400s: the RAW body
+				// text rides the error message — the banner shows it, and the
+				// page's refusal flow JSON-parses the envelope out of it.
+				throw new Error((await res.text()) || `The send failed (${res.status}).`);
+			}
+			await this.consumeStream(res);
+		} catch (err) {
+			if ((err as Error).name === "AbortError") return; // stop() or a newer consume won
+			this.setError((err as Error).message);
+			this.setStatus("error");
+		}
+	}
+
+	/** The answer path (the dock): the answer IS the question call's tool
+	 *  result (fork 1), riding as its input event. */
+	submitAnswer(tool: string, toolCallId: string, output: QuestionAnswerData): Promise<void> {
+		void tool;
+		return this.send({ type: "answer", toolCallId, answer: output });
+	}
+
+	/** Local detach: cancel the consume, keep the partial tail. The page's
+	 *  settle watch adopts the record when the server turn finishes. */
+	stop(): void {
+		this.consumeAbort?.abort();
+		this.consumeAbort = null;
+		this.streamingIds = new Set();
+		this.executions = new Map();
+		this.setStatus("ready");
+	}
+
+	/** Full snapshot swap (the idle settle sync — server canon). */
+	setMessages(record: LhMessage[]): void {
+		this.mutate(() => {
+			this.base = record;
+			this.tail = [];
+		});
+	}
+
+	/**
+	 * Append-only adoption (the mid-turn steer/artifact ground rule): swap
+	 * in the record — every current base message must still be present
+	 * (the server appends only) — keeping the live tail, minus the
+	 * optimistic entries the record has absorbed (their server-stamped
+	 * copies just landed; without this the input bubble would double).
+	 */
+	adoptRecord(record: LhMessage[]): void {
+		const recordIds = new Set(record.map((m) => (m as { id?: string }).id));
+		const baseAbsorbed = this.base.some((m) => !recordIds.has((m as { id?: string }).id));
+		if (baseAbsorbed) return; // not a pure append — the idle full sync is the net
+		this.mutate(() => {
+			this.base = record;
+			this.tail = this.tail.filter((m) => !absorbed(m, record));
+		});
+	}
+
+	/** The mount/sync continuation fold (native-messages.ts). */
+	static needsContinuation = needsContinuation;
+
+	clearError(): void {
+		this.setError(null);
+	}
+}

components/session.ts

@@ -0,0 +1,87 @@
+"use client";
+
+/**
+ * THE SESSION HOOK (Wave D, owner-approved 2026-09-26): the React adapter
+ * over the pure SessionModel (components/session-core.ts — the state
+ * machine, testable without a renderer). All mechanics live in the model;
+ * this hook re-renders on change and derives the view (core/view.ts
+ * projectView — the golden-master grouping, client-side).
+ *
+ * Every turn is a server-side background process (owner ruling
+ * 2026-09-25); the hook is a view. Sends are ONE explicit input event
+ * each ({chatId, input}); the turn stream is the VERBATIM AgentEvent
+ * passthrough. Status vocabulary (pre-Wave-D parity): ready | submitted
+ * | streaming | error.
+ */
+
+import { useCallback, useEffect, useMemo, useReducer, useRef } from "react";
+import type { ChatInput, LhMessage, QuestionAnswerData } from "@/src/core";
+import { SessionModel, type SessionState } from "@/components/session-core";
+
+export interface UseChatSessionOptions {
+	chatId: string;
+	/** The native record seed (server-rendered page or a full sync). */
+	initialMessages: LhMessage[];
+	/** Reattach on mount: a turn is running server-side with no live view. */
+	initiallyInFlight?: boolean;
+}
+
+export function useChatSession({ chatId, initialMessages, initiallyInFlight }: UseChatSessionOptions) {
+	// The model is per-mount (chatId-keyed component); onChange forces the
+	// re-render. The ref pattern keeps it stable across renders.
+	const modelRef = useRef<SessionModel | null>(null);
+	if (modelRef.current === null) {
+		modelRef.current = new SessionModel({ chatId, initialMessages });
+	}
+	const model = modelRef.current;
+	const [, forceRender] = useReducer((n: number) => n + 1, 0);
+	useEffect(() => {
+		model.onChange = forceRender;
+		return () => {
+			model.onChange = undefined;
+		};
+	}, [model, forceRender]);
+
+	// Reattach on mount (StrictMode-safe: a newer resume cancels the older).
+	useEffect(() => {
+		if (initiallyInFlight) void model.resumeStream();
+	}, [initiallyInFlight, model]);
+
+	// The view + state, derived per render from the model.
+	const state = useMemo<SessionState>(() => model.getState(), [model, forceRender] as const);
+
+	// The reattach settled terminal → the sync-and-continue seam (chat.tsx).
+	const setReattachSettled = useCallback(
+		(fn: () => void) => {
+			model.setReattachSettled(fn);
+		},
+		[model],
+	);
+
+	const send = useCallback((input: ChatInput) => void model.send(input), [model]);
+	const submitAnswer = useCallback(
+		(tool: string, toolCallId: string, output: QuestionAnswerData) =>
+			void model.submitAnswer(tool, toolCallId, output),
+		[model],
+	);
+	const setMessages = useCallback((record: LhMessage[]) => model.setMessages(record), [model]);
+	const adoptRecord = useCallback((record: LhMessage[]) => model.adoptRecord(record), [model]);
+	const stop = useCallback(() => model.stop(), [model]);
+	const clearError = useCallback(() => model.clearError(), [model]);
+	const resumeStream = useCallback(() => void model.resumeStream(), [model]);
+
+	return {
+		messages: state.messages,
+		nativeMessages: state.nativeMessages,
+		status: state.status,
+		error: state.error,
+		send,
+		submitAnswer,
+		setMessages,
+		adoptRecord,
+		stop,
+		clearError,
+		resumeStream,
+		setReattachSettled,
+	};
+}

package-lock.json

@@ -8,7 +8,6 @@
 			"name": "learning-harness",
 			"version": "0.2.0",
 			"dependencies": {
-				"@ai-sdk/react": "^4.0.79",
 				"@earendil-works/pi-agent-core": "^0.85.1",
 				"@earendil-works/pi-ai": "^0.85.1",
 				"@mozilla/readability": "^0.6.0",
@@ -19,7 +18,6 @@
 				"@streamdown/math": "^1.0.2",
 				"@streamdown/mermaid": "^1.0.2",
 				"@xyflow/react": "^12.11.3",
-				"ai": "^7.0.76",
 				"ansi-to-react": "^6.2.6",
 				"class-variance-authority": "^0.7.1",
 				"clsx": "^2.1.1",
@@ -73,91 +71,6 @@
 				"typescript": "^5.6.0"
 			}
 		},
-		"node_modules/@ai-sdk/gateway": {
-			"version": "4.0.61",
-			"resolved": "https://registry.npmjs.org/@ai-sdk/gateway/-/gateway-4.0.61.tgz",
-			"integrity": "sha512-b+PRdwM9/x0427CKqfd4zpCRgSzQeZsCZ9PtBOaLU2S1l+eRj8++Bwt766VnLerd/F5snKWAq13miV+OPcvclA==",
-			"license": "Apache-2.0",
-			"dependencies": {
-				"@ai-sdk/provider": "4.0.7",
-				"@ai-sdk/provider-utils": "5.0.28",
-				"@vercel/oidc": "3.2.0"
-			},
-			"engines": {
-				"node": ">=22"
-			},
-			"peerDependencies": {
-				"zod": "^3.25.76 || ^4.1.8"
-			}
-		},
-		"node_modules/@ai-sdk/mcp": {
-			"version": "2.0.35",
-			"resolved": "https://registry.npmjs.org/@ai-sdk/mcp/-/mcp-2.0.35.tgz",
-			"integrity": "sha512-hyaryid0qH/d8VtNZoJKeqcLy6b3v9xlkBL3JklbFXMAZJ8nmkxngrxQtjbNRTKjSFoyQ95/7IuRIijrSljMww==",
-			"license": "Apache-2.0",
-			"dependencies": {
-				"@ai-sdk/provider": "4.0.7",
-				"@ai-sdk/provider-utils": "5.0.28",
-				"pkce-challenge": "^5.0.1"
-			},
-			"engines": {
-				"node": ">=22"
-			},
-			"peerDependencies": {
-				"zod": "^3.25.76 || ^4.1.8"
-			}
-		},
-		"node_modules/@ai-sdk/provider": {
-			"version": "4.0.7",
-			"resolved": "https://registry.npmjs.org/@ai-sdk/provider/-/provider-4.0.7.tgz",
-			"integrity": "sha512-6or44XprPzKbr8zkmzosowSE0pxkvJcoojBL+mCZvPUt3kvXp3XSNqeVun9golb1acEfSo6yaEBRT18h2VU+1Q==",
-			"license": "Apache-2.0",
-			"dependencies": {
-				"json-schema": "^0.4.0"
-			},
-			"engines": {
-				"node": ">=22"
-			}
-		},
-		"node_modules/@ai-sdk/provider-utils": {
-			"version": "5.0.28",
-			"resolved": "https://registry.npmjs.org/@ai-sdk/provider-utils/-/provider-utils-5.0.28.tgz",
-			"integrity": "sha512-TnHUyd/rCYQqHg5RuiOaz/hUql6U+kbUaBW0Rp+0N5UhnAInA9CzzV0HXvuAAPwppDsK6fAx9Rd+tRawpJ/3pg==",
-			"license": "Apache-2.0",
-			"dependencies": {
-				"@ai-sdk/provider": "4.0.7",
-				"@standard-schema/spec": "^1.1.0",
-				"@workflow/serde": "4.1.0",
-				"eventsource-parser": "^3.0.8",
-				"undici": "^7.28.0"
-			},
-			"engines": {
-				"node": ">=22"
-			},
-			"peerDependencies": {
-				"zod": "^3.25.76 || ^4.1.8"
-			}
-		},
-		"node_modules/@ai-sdk/react": {
-			"version": "4.0.79",
-			"resolved": "https://registry.npmjs.org/@ai-sdk/react/-/react-4.0.79.tgz",
-			"integrity": "sha512-x7dZ/teRKXgJfukoOSiKA4pMv1AbMLjTGA8/d9Dofl8Mgf4DfRNuBxs6xlcedPypWTJXkK8whac2THkOpoFzug==",
-			"license": "Apache-2.0",
-			"dependencies": {
-				"@ai-sdk/mcp": "2.0.35",
-				"@ai-sdk/provider": "4.0.7",
-				"@ai-sdk/provider-utils": "5.0.28",
-				"ai": "7.0.76",
-				"swr": "^2.4.1",
-				"throttleit": "2.1.0"
-			},
-			"engines": {
-				"node": ">=22"
-			},
-			"peerDependencies": {
-				"react": "^18 || ~19.0.1 || ~19.1.2 || ^19.2.1"
-			}
-		},
 		"node_modules/@alloc/quick-lru": {
 			"version": "5.2.0",
 			"resolved": "https://registry.npmjs.org/@alloc/quick-lru/-/quick-lru-5.2.0.tgz",
@@ -5245,12 +5158,6 @@
 			"integrity": "sha512-1bnPQqSxSuc3Ii6MhBysoWCg58j97aUjuCSZrGSmDxNqtytIi0k8utUenAwTZN4V5mXXYGsVUI9zeBqy+jBOSQ==",
 			"license": "MIT"
 		},
-		"node_modules/@standard-schema/spec": {
-			"version": "1.1.0",
-			"resolved": "https://registry.npmjs.org/@standard-schema/spec/-/spec-1.1.0.tgz",
-			"integrity": "sha512-l2aFy5jALhniG5HgqrD6jXLi/rUWrKvqN/qJx6yoJsgKhblVd+iqqU4RCXavm/jPityDo5TCvKMnpjKnOriy0w==",
-			"license": "MIT"
-		},
 		"node_modules/@streamdown/cjk": {
 			"version": "1.0.3",
 			"resolved": "https://registry.npmjs.org/@streamdown/cjk/-/cjk-1.0.3.tgz",
@@ -6682,21 +6589,6 @@
 				"d3-transition": "^3.0.1"
 			}
 		},
-		"node_modules/@vercel/oidc": {
-			"version": "3.2.0",
-			"resolved": "https://registry.npmjs.org/@vercel/oidc/-/oidc-3.2.0.tgz",
-			"integrity": "sha512-UycprH3T6n3jH0k44NHMa7pnFHGu/N05MjojYr+Mc6I7obkoLIJujSWwin1pCvdy/eOxrI/l3uDLQsmcrOb4ug==",
-			"license": "Apache-2.0",
-			"engines": {
-				"node": ">= 20"
-			}
-		},
-		"node_modules/@workflow/serde": {
-			"version": "4.1.0",
-			"resolved": "https://registry.npmjs.org/@workflow/serde/-/serde-4.1.0.tgz",
-			"integrity": "sha512-pav4F2BoirECWR7Nf1TKt+2eETcBj7jj4cBefQ8VXQCA6NPkaKeLfj/zMgi+3zYV5ZIBT4GuUiphsj0/b9hPQQ==",
-			"license": "Apache-2.0"
-		},
 		"node_modules/@xmldom/xmldom": {
 			"version": "0.8.15",
 			"resolved": "https://registry.npmjs.org/@xmldom/xmldom/-/xmldom-0.8.15.tgz",
@@ -6789,23 +6681,6 @@
 				"node": ">= 14"
 			}
 		},
-		"node_modules/ai": {
-			"version": "7.0.76",
-			"resolved": "https://registry.npmjs.org/ai/-/ai-7.0.76.tgz",
-			"integrity": "sha512-Mytm/RFJzwJ0nzXiHYAE3dgNpFG/2K2Tks5alwm+PzAQeB6RJAJ97eGuZ8wssy66PcvtGWGIscS4tRi6dg0KLQ==",
-			"license": "Apache-2.0",
-			"dependencies": {
-				"@ai-sdk/gateway": "4.0.61",
-				"@ai-sdk/provider": "4.0.7",
-				"@ai-sdk/provider-utils": "5.0.28"
-			},
-			"engines": {
-				"node": ">=22"
-			},
-			"peerDependencies": {
-				"zod": "^3.25.76 || ^4.1.8"
-			}
-		},
 		"node_modules/ajv": {
 			"version": "6.15.0",
 			"resolved": "https://registry.npmjs.org/ajv/-/ajv-6.15.0.tgz",
@@ -10139,15 +10014,6 @@
 				"node": ">=0.10.0"
 			}
 		},
-		"node_modules/eventsource-parser": {
-			"version": "3.1.1",
-			"resolved": "https://registry.npmjs.org/eventsource-parser/-/eventsource-parser-3.1.1.tgz",
-			"integrity": "sha512-EKN1vKAMcZ8MlYMpaNuxN6R9yakzH6uajHcHVTqWJzvu5pWw9DyhbP35HH8MVBQ+dZjAfDxk+A8NiR9KWaXiyQ==",
-			"license": "MIT",
-			"engines": {
-				"node": ">=18.0.0"
-			}
-		},
 		"node_modules/exponential-backoff": {
 			"version": "3.1.3",
 			"resolved": "https://registry.npmjs.org/exponential-backoff/-/exponential-backoff-3.1.3.tgz",
@@ -12091,12 +11957,6 @@
 			"dev": true,
 			"license": "MIT"
 		},
-		"node_modules/json-schema": {
-			"version": "0.4.0",
-			"resolved": "https://registry.npmjs.org/json-schema/-/json-schema-0.4.0.tgz",
-			"integrity": "sha512-es94M3nTIfsEPisRafak+HDLfHXnKBhV3vU5eqPcS3flIWqcxJWgXHXiey3YrpaNsanY5ei1VoYEbOzijuq9BA==",
-			"license": "(AFL-2.1 OR BSD-3-Clause)"
-		},
 		"node_modules/json-schema-to-ts": {
 			"version": "3.1.1",
 			"resolved": "https://registry.npmjs.org/json-schema-to-ts/-/json-schema-to-ts-3.1.1.tgz",
@@ -14680,15 +14540,6 @@
 				"url": "https://github.com/sponsors/jonschlinkert"
 			}
 		},
-		"node_modules/pkce-challenge": {
-			"version": "5.0.1",
-			"resolved": "https://registry.npmjs.org/pkce-challenge/-/pkce-challenge-5.0.1.tgz",
-			"integrity": "sha512-wQ0b/W4Fr01qtpHlqSqspcj3EhBvimsdh0KlHhH8HRZnMsEa0ea2fTULOXOS9ccQr3om+GcGRk4e+isrZWV8qQ==",
-			"license": "MIT",
-			"engines": {
-				"node": ">=16.20.0"
-			}
-		},
 		"node_modules/pkijs": {
 			"version": "3.4.0",
 			"resolved": "https://registry.npmjs.org/pkijs/-/pkijs-3.4.0.tgz",
@@ -16530,19 +16381,6 @@
 				"url": "https://github.com/sponsors/ljharb"
 			}
 		},
-		"node_modules/swr": {
-			"version": "2.5.1",
-			"resolved": "https://registry.npmjs.org/swr/-/swr-2.5.1.tgz",
-			"integrity": "sha512-BRw55e8r0B7SpDN20CAzoQAHl7y1yP7/Zt7oqUjMv0vSt2u2Xnkm88Ws+VypbV9BXHQVuSuyVq7zMjO16wSExw==",
-			"license": "MIT",
-			"dependencies": {
-				"dequal": "^2.0.3",
-				"use-sync-external-store": "^1.6.0"
-			},
-			"peerDependencies": {
-				"react": "^16.11.0 || ^17.0.0 || ^18.0.0 || ^19.0.0"
-			}
-		},
 		"node_modules/tailwind-merge": {
 			"version": "3.6.0",
 			"resolved": "https://registry.npmjs.org/tailwind-merge/-/tailwind-merge-3.6.0.tgz",
@@ -16627,18 +16465,6 @@
 				"fs-extra": "^10.0.0"
 			}
 		},
-		"node_modules/throttleit": {
-			"version": "2.1.0",
-			"resolved": "https://registry.npmjs.org/throttleit/-/throttleit-2.1.0.tgz",
-			"integrity": "sha512-nt6AMGKW1p/70DF/hGBdJB57B8Tspmbp5gfJ8ilhLnt7kkr2ye7hzD6NVG8GGErk2HWF34igrL2CXmNIkzKqKw==",
-			"license": "MIT",
-			"engines": {
-				"node": ">=18"
-			},
-			"funding": {
-				"url": "https://github.com/sponsors/sindresorhus"
-			}
-		},
 		"node_modules/tiny-async-pool": {
 			"version": "1.3.0",
 			"resolved": "https://registry.npmjs.org/tiny-async-pool/-/tiny-async-pool-1.3.0.tgz",
@@ -17046,7 +16872,9 @@
 			"version": "7.29.0",
 			"resolved": "https://registry.npmjs.org/undici/-/undici-7.29.0.tgz",
 			"integrity": "sha512-IDxfleLmmbSskfWSUATiN1nfn2rDuvnMOqb5CWR92iIfojA0Ud+ulOAAEQ57LPr9rWmsreUyf5lwyao+7GNNVw==",
+			"dev": true,
 			"license": "MIT",
+			"optional": true,
 			"engines": {
 				"node": ">=20.18.1"
 			}

package.json

@@ -17,7 +17,6 @@
 		"desktop:dist": "next build && node scripts/build-server.mjs && electron-builder --mac --arm64"
 	},
 	"dependencies": {
-		"@ai-sdk/react": "^4.0.79",
 		"@earendil-works/pi-agent-core": "^0.85.1",
 		"@earendil-works/pi-ai": "^0.85.1",
 		"typebox": "1.3.7",
@@ -29,7 +28,6 @@
 		"@streamdown/math": "^1.0.2",
 		"@streamdown/mermaid": "^1.0.2",
 		"@xyflow/react": "^12.11.3",
-		"ai": "^7.0.76",
 		"ansi-to-react": "^6.2.6",
 		"class-variance-authority": "^0.7.1",
 		"clsx": "^2.1.1",

src/agent/subagents/extraction.ts

@@ -1,5 +1,4 @@
 import * as crypto from "node:crypto";
-import type { UIMessage } from "ai";
 import type { Usage } from "@earendil-works/pi-ai";
 import type { AgentTool } from "@earendil-works/pi-agent-core";
 import { Type, type Static, type TSchema } from "typebox";
@@ -28,7 +27,6 @@ import { getGraph, mergeFindings, type EmissionStats, type GraphDeps } from "../
 import type { GraphStore } from "../../core/ports.ts";
 import { settleGraphLayout } from "../../core/layout-settle.ts";
 import { batchRuns, batchTag, coverageOf, messageViewLines } from "../../core/coverage.ts";
-import { projectNative } from "../../core/projection.ts";
 
 /**
  * The extraction pipeline (vision §1 construction, 2026-09-02 redesign) — the
@@ -93,12 +91,11 @@ import { projectNative } from "../../core/projection.ts";
 export function transcriptView(record: ChatRecord): string {
 	// The shared seam (core/coverage.ts messageViewLines): the LLM's prompt
 	// view and the coverage fingerprint use ONE line builder, so "covered"
-	// means — exactly — what the extractor saw. Wave B: the record is native
+	// means — exactly — what the extractor saw. Wave D: the record is native
-	// history — the extraction view reads through the one-way projection.
+	// history — the view reads the native messages directly.
 	const lines = [`SESSION: ${record.title}`];
-	const projected = projectNative(record.messages);
+	record.messages.forEach((message, index) => {
-	projected.forEach((message, index) => {
+		lines.push(...messageViewLines(message, (message as { id?: string }).id || `m${index}`));
-		lines.push(...messageViewLines(message, message.id || `m${index}`));
 	});
 	return lines.join("\n");
 }
@@ -871,14 +868,16 @@ export function createExtractionPipeline(deps: ExtractionPipelineDeps): Extracti
 			// engines AND their providers' keys; a refusal lands BEFORE any
 			// stage runs (typed — the trigger skips, routes surface 503).
 			refuseIfStagesUnready();
-			const projected = projectNative(record.messages);
+			const covered = coverageOf(record.messages, graphStore.readCoverage(), chatId);
-			const covered = coverageOf(projected, graphStore.readCoverage(), chatId);
 			if (covered.uncovered.length === 0) {
 				console.info(`[extraction ${chatId}] nothing uncovered — no-op`);
 				return { nodes: 0, edges: 0, events: 0 };
 			}
 			const chunkTokens = deps.getChunkTokens?.() ?? FACTORY_SETTINGS.graph.extractionChunkTokens;
-			const batches = batchRuns(covered.uncovered, chunkTokens);
+			// The batcher's one rule (owner, Wave D): a tool-call message and
+			// its result message(s) are INSEPARABLE — one chunk, covered
+			// together, read together (core/coverage.ts batchRuns).
+			const batches = batchRuns(covered.slots, covered.uncovered, chunkTokens);
 			console.info(
 				`[extraction ${chatId}] pipeline start — ${covered.uncoveredTokens} uncovered tok, ` +
 					`${covered.uncoveredCount} messages, ${batches.length} batch(es) of ≤${chunkTokens} tok`,
@@ -887,13 +886,9 @@ export function createExtractionPipeline(deps: ExtractionPipelineDeps): Extracti
 				// The batch's prompt view: SESSION header + the batch's lines,
 				// same provenance [id] tags the fingerprints used (coverage
 				// and prompt share ONE line builder — core/coverage.ts).
-				const mids = projected.map((m, i) => (m.id || `m${i}`) as string);
 				const view = [
 					`SESSION: ${record.title}`,
-					...batch.flatMap((slot) => {
+					...batch.flatMap((slot) => messageViewLines(slot.message, slot.message.id)),
-						const idx = projected.indexOf(slot.message);
-						return messageViewLines(slot.message, mids[idx] ?? slot.message.id);
-					}),
 				].join("\n");
 				setPhase(chatId, "extracting", "chat");
 				const findings = await extract(chatId, view);
@@ -918,7 +913,9 @@ export function createExtractionPipeline(deps: ExtractionPipelineDeps): Extracti
 				const stats = mergeFindings(chatId, auditResult, { chatStore, graphStore }, {
 					messages: batch.map((slot) => slot.message),
 					foldedAt,
-					coverage: batch.map((slot) => ({ chatId, fp: slot.fp, tokens: slot.tokens, foldedAt })),
+					coverage: batch
+						.filter((slot) => messageViewLines(slot.message, slot.message.id).length > 0)
+						.map((slot) => ({ chatId, fp: slot.fp, tokens: slot.tokens, foldedAt })),
 					batchTag: batchTag(batch),
 				});
 				total.nodes += stats.nodes;

src/agent/subagents/runner.ts

@@ -1,4 +1,3 @@
-import { isToolUIPart } from "ai";
 import type { PiEngine } from "../models.ts";
 import type { SearchProvider } from "../../core/search.ts";
 import {
@@ -23,8 +22,6 @@ import { registerJobAbort, unregisterJobAbort } from "./aborts.ts";
 import type { ModelRole } from "../../../config.ts";
 import { ROLE_SLOT, modelsNotConfigured, type ModelSlotName, type ProviderKeyMissing } from "../../core/settings.ts";
 import type { ChatStore, JobStore } from "../../core/ports.ts";
-import { projectNative } from "../../core/projection.ts";
-import type { UIMessage } from "ai";
 import { runDemoBuilder } from "./demo-builder.ts";
 import { runPdfReaderAgent } from "./pdf-reader.ts";
 import { runResearcherAgent } from "./researcher.ts";
@@ -276,22 +273,18 @@ export function createJobRunner(deps: JobRunnerDeps): JobRunner {
 		if (input) return input;
 		const chat = chatStore.getChat(job.chatId);
 		if (!chat) return undefined;
-		// Wave B: the record is native history — the transcript scan reads
+		// Wave D: the record is native history — the scan reads the native
-		// through the projection (same shape the old record held).
+		// assistant blocks directly (a toolCall's arguments are the input).
-		return (findToolPart(projectNative(chat.messages), job.toolCallId)?.input ?? undefined) as
+		for (const m of chat.messages) {
-			| JobToolInput
+			if ((m as { role?: string }).role !== "assistant") continue;
-			| undefined;
+			const content = (m as { content?: unknown }).content;
-	}
+			if (!Array.isArray(content)) continue;
-
+			for (const block of content as Array<{ type?: string; id?: string; arguments?: unknown }>) {
-	// ─── shared plumbing ────────────────────────────────────────────────────
+				if (block.type !== "toolCall" || block.id !== job.toolCallId) continue;
-
+				return block.arguments as JobToolInput | undefined;
-	function findToolPart(messages: UIMessage[], toolCallId: string) {
-		for (const m of messages) {
-			for (const part of m.parts) {
-				if (isToolUIPart(part) && part.toolCallId === toolCallId) return part;
 			}
 		}
-		return null;
+		return undefined;
 	}
 
 	/**

src/core/contracts/chat.ts

@@ -3,10 +3,8 @@
  *
  * Part of the domain-split contracts (src/core/contracts/): each domain file
  * is owned outright by its stream — cross-domain schema changes go through
- * master. Pure types + protocol docs + tiny pure helpers; type-only imports
+ * master. Pure types + protocol docs + tiny pure helpers. See docs/vision.md.
- * of "ai" (safe everywhere: server, client, tests). See docs/vision.md.
  */
-import type { UIMessage } from "ai";
 import type { ThinkingLevel } from "@earendil-works/pi-agent-core";
 import type { LhMessage } from "../native-messages.ts";
 import type { Phase3 } from "../phases.ts";
@@ -20,8 +18,8 @@ import type { EmissionStats, UnderstandingLevel } from "./graph.ts";
  * agent loop's own shapes, provider/model/usage/thinkingSignature stamped
  * by pi) — the canonical session record IS the agent state; replay is
  * `state.messages = record.messages` + continue(). The client-shaped view
- * is the one-way projection (core/projection.ts) — a pending quiz is an
+ * is the client's own grouping fold (core/view.ts, Wave D) — a pending quiz
- * unanswered question tool call, which is exactly what makes
+ * is an unanswered question tool call, which is exactly what makes
  * resume-mid-question free. */
 export interface ChatRecord {
 	id: string;

src/core/contracts/plan.ts

@@ -115,3 +115,28 @@ export interface PlanToolResult {
 	tree: PlanTree;
 }
 
+
+// ─── the native fold (Wave D: derivePlanStatuses over LhMessage[]) ───────────
+
+import type { LhMessage } from "../native-messages.ts";
+
+/** The same derivation over the NATIVE record: quiz CALLS only (any state —
+ *  a dangling call still moves position; results feed nothing). */
+export function derivePlanStatusesLh(tree: PlanTree, messages: readonly LhMessage[]): DerivedPlanStatuses {
+	const ids = new Set(tree.nodes.map((n) => n.id));
+	const runs: string[] = [];
+	for (const m of messages) {
+		if ((m as { role?: string }).role !== "assistant") continue;
+		const content = (m as { content?: unknown }).content;
+		if (!Array.isArray(content)) continue;
+		for (const block of content as Array<{ type?: string; name?: string; arguments?: unknown }>) {
+			if (block.type !== "toolCall" || block.name !== "quiz") continue;
+			const nodeId = (block.arguments as { nodeId?: unknown } | undefined)?.nodeId;
+			if (typeof nodeId !== "string" || !ids.has(nodeId)) continue;
+			if (runs.at(-1) !== nodeId) runs.push(nodeId);
+		}
+	}
+	const current = runs.at(-1) ?? null;
+	const visited = new Set(runs.slice(0, -1));
+	return { current, visited };
+}

src/core/coverage.ts

@@ -1,63 +1,83 @@
 /**
  * Extraction coverage (vision-extraction-coverage.md, settled 2026-09-15) —
- * the pure core. THE MESSAGE IS THE UNIT OF COVERAGE: a per-chat append-only
+ * the pure core. THE MESSAGE IS THE UNIT OF COVERAGE (Wave D, owner ruling
- * ledger records which messages have been extracted, by CONTENT FINGERPRINT
+ * 2026-09-26: the native message, not the old projected group): a per-chat
- * over exactly what the extraction view sees; "uncovered" is DERIVED at
+ * append-only ledger records which messages have been extracted, by CONTENT
- * read time (fingerprint the current transcript, diff against the ledger) —
+ * FINGERPRINT over exactly what the extraction view sees; "uncovered" is
- * never dual-tracked. Runs are token-budgeted batches of the oldest
+ * DERIVED at read time (fingerprint the current transcript, diff against the
- * uncovered messages, so a tool call and its result (which share a message)
+ * ledger) — never dual-tracked. Runs are token-budgeted batches of the oldest
- * can never be separated.
+ * uncovered messages, and THE ONE BATCH RULE (owner): a tool-call message and
+ * its tool-result message(s) are INSEPARABLE — they always appear in the same
+ * chunk (a covered partner rides along as context; checkpoints are
+ * idempotent, so re-marking it is a no-op).
  *
  * The fingerprint and the LLM prompt share ONE line builder —
  * messageViewLines — so "covered" and "what the model saw" cannot drift
  * apart. The line logic is tree-blind, identical to the extractor's view:
  * user/teacher text verbatim, quiz outcomes as one-line summaries, other
- * tool results as one generic line each, data parts invisible (plan-block
+ * clean tool results as one generic line each, markers invisible (a
- * attach does not re-open a covered message).
+ * plan-block marker does not re-open a covered message).
  */
 
 import { createHash } from "node:crypto";
-import type { UIMessage } from "ai";
-import { isToolUIPart } from "ai";
 
 import type { CoverageEntry } from "./contracts/graph.ts";
+import type { LhMessage } from "./native-messages.ts";
+import type { ToolResultMessage } from "@earendil-works/pi-ai";
 
 export type { CoverageEntry };
 
 /** The quiz outcome line — both the fingerprint and the LLM view see the
- *  quiz answer, so an answered quiz re-opens the message that asked it. */
+ *  quiz answer, so an answered quiz re-opens (the answer is its own native
-function quizLine(part: { toolCallId: string; input: unknown; output: unknown }): string {
+ *  message now; its fingerprint is new the moment it lands). */
-	const input = part.input as { question?: unknown } | undefined;
+function quizLine(result: ToolResultMessage): string {
-	const out = part.output as Partial<{ questionText: string; correct: boolean; dontKnow: boolean; note: string }> | undefined;
+	const out = result.details as
-	const question = typeof input?.question === "string" ? input.question : out?.questionText ?? "(no question)";
+		| Partial<{ questionText: string; correct: boolean; dontKnow: boolean; note: string }>
+		| undefined;
+	const question = typeof out?.questionText === "string" ? out.questionText : "(no question)";
 	const outcome = out?.dontKnow === true ? "did not know" : out?.correct === true ? "answered correctly" : "answered incorrectly";
 	const note = typeof out?.note === "string" && out.note ? ` (note: ${out.note})` : "";
-	return `QUIZ [${part.toolCallId}] ${question} — the learner ${outcome}${note}`;
+	return `QUIZ [${result.toolCallId}] ${question} — the learner ${outcome}${note}`;
+}
+
+/** The user-facing text of a native message (string or text blocks). */
+function textOf(content: unknown): string[] {
+	if (typeof content === "string") return content.trim() ? [content] : [];
+	if (!Array.isArray(content)) return [];
+	return content
+		.filter((c): c is { type: "text"; text: string } => (c as { type?: string }).type === "text")
+		.map((c) => c.text)
+		.filter((t) => t.trim().length > 0);
 }
 
 /**
- * The extraction-relevant lines of ONE message — the shared seam between
+ * The extraction-relevant lines of ONE native message — the shared seam
- * transcriptView (the LLM's prompt) and messageFingerprint (coverage). Same
+ * between transcriptView (the LLM's prompt) and messageFingerprint
- * logic, same order, so a covered message is by construction exactly what
+ * (coverage). Same logic, same order, so a covered message is by
- * the extractor saw. `mid` is the provenance tag (the message id; the
+ * construction exactly what the extractor saw. `mid` is the provenance tag
- * transcript view passes an index fallback for legacy id-less messages,
+ * (the message id; the transcript view passes an index fallback for legacy
- * coverage passes the id — persisted messages always have ids).
+ * id-less messages, coverage passes the id — persisted messages always have
+ * ids).
  */
-export function messageViewLines(message: UIMessage, mid: string): string[] {
+export function messageViewLines(message: LhMessage, mid: string): string[] {
 	const lines: string[] = [];
-	const who = message.role === "user" ? "LEARNER" : "TEACHER";
+	const role = (message as { role?: string }).role;
-	for (const part of message.parts) {
+	if (role === "user") {
-		if (part.type === "text" && part.text.trim()) {
+		for (const text of textOf((message as { content: unknown }).content)) {
-			lines.push(`${who} [${mid}] ${part.text.trim()}`);
+			lines.push(`LEARNER [${mid}] ${text.trim()}`);
-		} else if (isToolUIPart(part)) {
-			const name = part.type.slice("tool-".length);
-			if (name === "quiz" && part.state === "output-available") {
-				lines.push(quizLine(part as unknown as { toolCallId: string; input: unknown; output: unknown }));
-			} else if (part.state === "output-available") {
-				lines.push(`(${name} tool result delivered)`);
-			}
 		}
+	} else if (role === "assistant") {
+		for (const text of textOf((message as { content: unknown }).content)) {
+			lines.push(`TEACHER [${mid}] ${text.trim()}`);
+		}
+	} else if (role === "toolResult") {
+		const result = message as ToolResultMessage;
+		if (result.isError) return lines; // a failed call is retry context, not learning evidence
+		lines.push(
+			result.toolName === "quiz" ? quizLine(result) : `(${result.toolName} tool result delivered)`,
+		);
 	}
+	// custom markers are invisible to the extraction view.
 	return lines;
 }
 
@@ -69,33 +89,28 @@ export function viewTokens(lines: string[]): number {
 /**
  * A message's coverage fingerprint: sha256 of role + its view lines (with
  * the message's real id as the provenance tag). Changes exactly when the
- * extraction view of the message changes: an answered quiz merges into the
+ * extraction view of the message changes. (Wave D: the unit is the native
- * asking message → new line → the message re-opens itself, and only itself.
+ * message — an answered quiz is the answer's OWN toolResult message, so
- * Plan-block part attach changes data parts (invisible to the view) → the
+ * answering opens exactly that new message; the asking message is pulled
- * fingerprint is stable → still covered.
+ * into the batch by the inseparability rule, not re-opened.)
  */
-export function messageFingerprint(message: UIMessage): string {
+export function messageFingerprint(message: LhMessage): string {
 	const h = createHash("sha256");
-	h.update(JSON.stringify([message.role, messageViewLines(message, message.id)]));
+	h.update(JSON.stringify([(message as { role?: string }).role, messageViewLines(message, message.id)]));
 	return h.digest("hex");
 }
 
 /** One message's coverage-relevant projection. */
 export interface CoverageSlot {
-	message: UIMessage;
+	message: LhMessage;
 	fp: string;
 	tokens: number;
 }
 
 /** Fingerprint a whole transcript, in order (used by the trigger, the run,
  *  and the coverage view — one derivation everywhere). */
-export function coverageSlots(messages: ReadonlyArray<UIMessage>): CoverageSlot[] {
+export function coverageSlots(messages: ReadonlyArray<LhMessage>): CoverageSlot[] {
-	return messages.map((message) => {
+	return messages.map((message) => ({ message, fp: messageFingerprint(message), tokens: viewTokens(messageViewLines(message, message.id)) }));
-		const lines = messageViewLines(message, message.id);
-		const h = createHash("sha256");
-		h.update(JSON.stringify([message.role, lines]));
-		return { message, fp: h.digest("hex"), tokens: viewTokens(lines) };
-	});
 }
 
 /** The covered-fingerprint set for one chat, from the ledger. */
@@ -106,14 +121,19 @@ export function coveredSet(entries: ReadonlyArray<CoverageEntry>, chatId: string
 }
 
 /** The derived state: which slots are uncovered, plus the token totals.
- *  "Uncovered" needs no ledger of its own — this diff IS the coverage view. */
+ *  "Uncovered" needs no ledger of its own — this diff IS the coverage view.
+ *  `uncovered` is INDEXES into `slots` (the batcher needs positions for the
+ *  inseparability rule). Messages with NO view lines (custom markers, an
+ *  error result) are invisible to coverage entirely: the extractor can
+ *  never see them, so they are neither covered nor uncovered — a marker
+ *  attach never opens (or joins) an extraction run. */
 export function coverageOf(
-	messages: ReadonlyArray<UIMessage>,
+	messages: ReadonlyArray<LhMessage>,
 	entries: ReadonlyArray<CoverageEntry>,
 	chatId: string,
 ): {
 	slots: CoverageSlot[];
-	uncovered: CoverageSlot[];
+	uncovered: number[];
 	coveredTokens: number;
 	uncoveredTokens: number;
 	coveredCount: number;
@@ -121,40 +141,132 @@ export function coverageOf(
 } {
 	const covered = coveredSet(entries, chatId);
 	const slots = coverageSlots(messages);
-	const uncovered = slots.filter((s) => !covered.has(s.fp));
+	const uncovered: number[] = [];
-	const coveredCount = slots.length - uncovered.length;
 	let coveredTokens = 0;
 	let uncoveredTokens = 0;
-	for (const s of slots) {
+	for (let i = 0; i < slots.length; i++) {
-		if (covered.has(s.fp)) coveredTokens += s.tokens;
+		const slot = slots[i]!;
-		else uncoveredTokens += s.tokens;
+		if (messageViewLines(slot.message, slot.message.id).length === 0) continue; // invisible
+		if (covered.has(slot.fp)) coveredTokens += slot.tokens;
+		else {
+			uncovered.push(i);
+			uncoveredTokens += slot.tokens;
+		}
 	}
-	return { slots, uncovered, coveredTokens, uncoveredTokens, coveredCount, uncoveredCount: uncovered.length };
+	const visibleCount = slots.length - slots.filter(
+		(s) => messageViewLines(s.message, s.message.id).length === 0,
+	).length;
+	return {
+		slots,
+		uncovered,
+		coveredTokens,
+		uncoveredTokens,
+		coveredCount: visibleCount - uncovered.length,
+		uncoveredCount: uncovered.length,
+	};
+}
+
+// ─── the inseparability rule (owner, Wave D fork 3) ─────────────────────────
+//
+// A tool-call message and its tool-result message(s) always appear in the
+// same chunk. Bonding is adjacency-honest: results follow their call's
+// assistant message contiguously in the native record (the agent loop
+// appends the results batch; the question gate guarantees nothing can land
+// between a quiz call and its learner answer). A result whose call's
+// assistant message has NO other bond partners still pulls that message in
+// (the extractor needs the question with its answer).
+
+/** The call ids an assistant message carries. */
+function callIdsOf(message: LhMessage): Set<string> {
+	const out = new Set<string>();
+	const content = (message as { content?: unknown }).content;
+	if (!Array.isArray(content)) return out;
+	for (const block of content as Array<{ type?: string; id?: string }>) {
+		if (block.type === "toolCall" && typeof block.id === "string") out.add(block.id);
+	}
+	return out;
+}
+
+/** The contiguous bond-range containing slot i: an assistant-with-calls
+ *  message plus its immediately-following results; a toolResult message
+ *  plus the assistant message that made its call (and that call's sibling
+ *  results). Anything else bonds to itself alone. */
+function bondRange(messages: ReadonlyArray<LhMessage>, i: number): [number, number] {
+	const roleOf = (m: LhMessage | undefined) => (m ? (m as { role?: string }).role : undefined);
+	const m = messages[i];
+	if (!m) return [i, i];
+	if (roleOf(m) === "assistant") {
+		const calls = callIdsOf(m);
+		if (calls.size === 0) return [i, i];
+		let end = i;
+		while (
+		roleOf(messages[end + 1]) === "toolResult" &&
+		calls.has((messages[end + 1] as ToolResultMessage).toolCallId)
+	) {
+		end++;
+	}
+		return [i, end];
+	}
+	if (roleOf(m) === "toolResult") {
+		let start = i;
+		while (roleOf(messages[start - 1]) === "toolResult") start--;
+		const owner = messages[start - 1];
+		if (
+			owner &&
+			roleOf(owner) === "assistant" &&
+			callIdsOf(owner).has((m as ToolResultMessage).toolCallId)
+		) {
+			const [s, e] = bondRange(messages, start - 1);
+			return [s, Math.max(e, i)];
+		}
+		return [start, i]; // orphaned results (no asking message): keep the contiguous result run together
+	}
+	return [i, i];
 }
 
 /**
  * Batch the uncovered messages into token-budgeted RUNS: consecutive slots,
- * oldest first, greedy fill — a message is never split (it is the unit), so
+ * oldest first, greedy fill — a message is never split, and a bond-range
- * a tool call and its result always travel together. A single message larger
+ * (call + results) never splits either (owner fork 3): a selected uncovered
- * than the budget becomes its own (over-budget) run rather than being
+ * slot EXPANDS to its whole bond-range, covered partners riding as context
- * dropped. Runs are ephemeral — the durable record is per-message coverage,
+ * (checkpoint idempotency makes re-marking them a no-op). A single
- * checkpointed after each run merges.
+ * bond-range larger than the budget becomes its own (over-budget) run
+ * rather than being dropped. Runs are ephemeral — the durable record is
+ * per-message coverage, checkpointed after each run merges.
  */
-export function batchRuns(uncovered: ReadonlyArray<CoverageSlot>, budget: number): CoverageSlot[][] {
+export function batchRuns(slots: ReadonlyArray<CoverageSlot>, uncovered: ReadonlyArray<number>, budget: number): CoverageSlot[][] {
 	if (budget <= 0 || uncovered.length === 0) return [];
+	const messages = slots.map((s) => s.message);
+	// Expand: every uncovered index marks its whole bond-range as needed.
+	const needed = new Set<number>();
+	for (const i of uncovered) {
+		const [start, end] = bondRange(messages, i);
+		for (let j = start; j <= end; j++) needed.add(j);
+	}
+	// Group the needed slots in order, greedy by budget, never splitting a
+	// bond-range: a range that would overflow starts a new run (and an
+	// over-budget range rides alone).
 	const runs: CoverageSlot[][] = [];
 	let current: CoverageSlot[] = [];
 	let currentTokens = 0;
-	for (const slot of uncovered) {
+	const flush = () => {
-		if (current.length > 0 && currentTokens + slot.tokens > budget) {
+		if (current.length > 0) runs.push(current);
-			runs.push(current);
+		current = [];
-			current = [];
+		currentTokens = 0;
-			currentTokens = 0;
+	};
+	for (let i = 0; i < slots.length; i++) {
+		if (!needed.has(i)) continue;
+		const [start, end] = bondRange(messages, i);
+		const rangeTokens = slots.slice(start, end + 1).reduce((t, s) => t + s.tokens, 0);
+		if (current.length > 0 && currentTokens + rangeTokens > budget) flush();
+		// The whole range enters together — skip past its end in the outer loop.
+		for (let j = start; j <= end; j++) {
+			current.push(slots[j]!);
+			currentTokens += slots[j]!.tokens;
 		}
-		current.push(slot);
+		i = end;
-		currentTokens += slot.tokens;
 	}
-	if (current.length > 0) runs.push(current);
+	flush();
 	return runs;
 }
 

src/core/graph.ts

@@ -39,7 +39,8 @@
  */
 
 import { randomBytes } from "node:crypto";
-import { isToolUIPart, type UIMessage } from "ai";
+import type { LhMessage } from "./native-messages.ts";
+import type { ToolResultMessage } from "@earendil-works/pi-ai";
 import type { CoverageEntry } from "./contracts/graph.ts";
 import type {
 	AuditResult,
@@ -85,7 +86,7 @@ function newGraphNodeId(): string {
  *  fresh understanding observations land as new events; the notes path
  *  passes neither and keeps the whole-file checkpoint. */
 export interface MergeSource {
-	messages: UIMessage[] | null;
+	messages: LhMessage[] | null;
 	foldedAt: string;
 	/** The batch being merged (chat path): one entry per message in
 	 *  `messages`, same order — undefined (notes path / coverage-less
@@ -97,7 +98,11 @@ export interface MergeSource {
 
 // ─── transcript scanning (shared by the merger + trigger) ─────────────────────
 
-/** A flattened, order-preserving view of every tool part in a transcript. */
+/** A flattened, order-preserving view of every tool call (+ its result
+ *  state) in a native transcript. Wave D: the pairing is native — calls
+ *  come from assistant content blocks; a result map built from the
+ *  toolResult messages carries the outcome (a dangling call is
+ *  input-available, an error result is output-error). */
 interface ToolPartView {
 	name: string;
 	toolCallId: string;
@@ -106,17 +111,27 @@ interface ToolPartView {
 	output: unknown;
 }
 
-function toolParts(messages: UIMessage[]): ToolPartView[] {
+function toolParts(messages: LhMessage[]): ToolPartView[] {
+	const results = new Map<string, ToolResultMessage>();
+	for (const m of messages) {
+		if ((m as { role?: string }).role === "toolResult") {
+			results.set((m as ToolResultMessage).toolCallId, m as ToolResultMessage);
+		}
+	}
 	const out: ToolPartView[] = [];
-	for (const message of messages) {
+	for (const m of messages) {
-		for (const part of message.parts) {
+		if ((m as { role?: string }).role !== "assistant") continue;
-			if (!isToolUIPart(part)) continue;
+		const content = (m as { content?: unknown }).content;
+		if (!Array.isArray(content)) continue;
+		for (const block of content as Array<{ type?: string; id?: string; name?: string; arguments?: unknown }>) {
+			if (block.type !== "toolCall" || typeof block.id !== "string") continue;
+			const result = results.get(block.id);
 			out.push({
-				name: part.type.slice("tool-".length),
+				name: block.name ?? "",
-				toolCallId: part.toolCallId,
+				toolCallId: block.id,
-				state: part.state,
+				state: result ? (result.isError ? "output-error" : "output-available") : "input-available",
-				input: part.input,
+				input: block.arguments,
-				output: part.output,
+				output: result && !result.isError ? result.details : undefined,
 			});
 		}
 	}

src/core/index.ts

@@ -31,13 +31,15 @@ export * from "./phases.ts";
 export * from "./pending.ts";
 export * from "./prompt-segments.ts";
 export * from "./math-delims.ts";
-export * from "./ui-message.ts";
 // Wave B (owner-approved 2026-09-26): the native record (LhMessage), the
-// one-way client-view projection, and the input-events contract — all
+// and the input-events contract — all
 // pure logic, browser-safe.
 export * from "./native-messages.ts";
-export * from "./projection.ts";
 export * from "./input-events.ts";
+// Wave D (owner-approved 2026-09-26): the client view model — the grouping
+// canon (one continuous assistant document, step separators) ported from
+// the deleted server projection to the client side. Pure, browser-safe.
+export * from "./view.ts";
 export * from "./tool-parts.ts";
 export * from "./prices.ts";
 export * from "./cost-summary.ts";

src/core/native-messages.ts

@@ -282,3 +282,48 @@ export function hasProbeDone(messages: readonly LhMessage[]): boolean {
 export function answeredCallIds(messages: readonly LhMessage[]): Set<string> {
 	return resultIdsFrom(messages);
 }
+
+// ─── the continuation fold (Wave D: the SDK's
+//     lastAssistantMessageIsCompleteWithToolCalls, natively) ──────────────────
+
+/**
+ * Does the record END on un-consumed tool results — a turn that stopped
+ * after its tools ran (a job dispatch's ack, an aborted answer turn),
+ * leaving the agent owing a continuation? This is the mount/sync
+ * continuation seam's gate (chat.tsx): when true and nothing runs, the
+ * client sends {type:"continue"} and the teacher picks up from the record.
+ *
+ * The pre-Wave-D client asked the SDK's
+ * lastAssistantMessageIsCompleteWithToolCalls (last view message is
+ * assistant; its last step's tool calls all output-available or
+ * output-error; at least one). The native fold over the record: the last
+ * message is a toolResult, and the results' OWNING assistant message has
+ * at least one call, ALL of them answered (a dangling question — the
+ * dock's domain — fails the all-answered rule and never fires a turn,
+ * exactly as the SDK's every(output-available) did).
+ */
+export function needsContinuation(messages: readonly LhMessage[]): boolean {
+	const last = messages[messages.length - 1];
+	if (!last || (last as { role?: string }).role !== "toolResult") return false;
+	// The trailing results batch (contiguous by construction — the agent
+	// appends its batch; the question gate keeps a quiz call and its
+	// learner answer adjacent).
+	let start = messages.length - 1;
+	while (
+		start > 0 &&
+		(messages[start - 1] as { role?: string } | undefined)?.role === "toolResult"
+	) {
+		start--;
+	}
+	const owner = messages[start - 1];
+	if (!owner || (owner as { role?: string }).role !== "assistant") return false;
+	const content = (owner as { content?: unknown }).content;
+	if (!Array.isArray(content)) return false;
+	const answered = new Set(
+		messages.slice(start).map((m) => (m as ToolResultMessage).toolCallId),
+	);
+	const calls = (content as Array<{ type?: string; id?: string }>).filter(
+		(b) => b.type === "toolCall" && typeof b.id === "string",
+	);
+	return calls.length > 0 && calls.every((b) => answered.has(b.id!));
+}

src/core/pending.ts

@@ -1,4 +1,18 @@
-import { isToolUIPart, type UIMessage } from "ai";
+/**
+ * Wave D: these scans read VIEW messages (core/view.ts LhViewMessage — the
+ * client's grouped transcript). The part shapes are the same vocabulary the
+ * pre-Wave-D client dialect used; a tool part is structural: `tool-` prefix +
+ * toolCallId. The server-side record scans (danglingQuestionOf over native
+ * LhMessage[]) live in native-messages.ts.
+ */
+interface UIMessage {
+	role?: string;
+	parts: Array<Record<string, unknown> & { type: string }>;
+}
+
+function isToolUIPart(part: UIMessage["parts"][number]): boolean {
+	return part.type.startsWith("tool-") && typeof part.toolCallId === "string";
+}
 
 /**
  * The pending-question gate: an unanswered QUESTION tool part (quiz/ask)
@@ -76,7 +90,7 @@ function findPending(messages: UIMessage[], names: readonly string[]): PendingTo
 			if (isToolUIPart(part) && part.state === "input-available") {
 				const tool = part.type.slice("tool-".length);
 				if (names.includes(tool)) {
-					return { toolCallId: part.toolCallId, tool, input: part.input };
+					return { toolCallId: part.toolCallId as string, tool, input: part.input };
 				}
 			}
 		}

src/core/phases.ts

@@ -1,15 +1,14 @@
 /**
  * Phase derivation (docs/probe-teach-split.md §4) — EXPLICIT EVENTS, never
  * occupancy. The probe→teach agent split (2026-09-11) gave every phase
- * boundary a transcript marker:
+ * boundary a transcript event:
  *
- *   probe  — until the first message containing a `tool-probe_done` part
+ *   probe  — until the prober's `probe_done` call CONCLUDES (a clean
- *            (the prober concludes; the assessment rides the tool part)
+ *            toolResult for it; the assessment rides the call's arguments)
  *   plan   — probe concluded, until the learner ACCEPTS the presented
- *            plan (a `data-verdict` part with accepted:true on the verdict
+ *            plan (the verdict marker with accepted:true — the explicit
- *            button's own user message — the explicit verdict gate;
+ *            verdict gate; plan_init itself is only the PRESENTATION and
- *            plan_init itself is only the PRESENTATION and moves no
+ *            moves no phase; a declined verdict stays in plan)
- *            phase; a declined verdict stays in plan)
  *   teach  — teaching is sanctioned
  *
  * The former derived-occupancy seam (phase-of-tree via `currentNodeId`)
@@ -18,6 +17,14 @@
  * progress fields at all. No backwards compatibility: legacy sessions
  * without probe_done simply read as probe (they are wiped data).
  *
+ * TWO FOLDS over the same events (Wave D): `phaseAt3`/`phaseTimeline` fold
+ * VIEW messages (the client's dividers and dock — part shapes, pure and
+ * structural); `phaseAt3Lh` folds the NATIVE record (server summaries,
+ * tests) — probe→plan advances at the probe_done RESULT message, plan→teach
+ * at the verdict marker. Both are strictly sequential; the two derivations
+ * agree on every settled record (a result/marker message projects into the
+ * same view boundary).
+ *
  * Pure: no imports, no runtime deps — safe on server, client, and tests.
  */
 
@@ -83,3 +90,40 @@ export function phaseTimeline(messages: readonly PhaseMessage[]): Phase3[] {
 	}
 	return out;
 }
+
+// ─── the native fold (Wave D: record-side phase over LhMessage[]) ───────────
+
+import type { LhMessage } from "./native-messages.ts";
+import type { ToolResultMessage } from "@earendil-works/pi-ai";
+
+/** The phase at `upToIndex` (inclusive; default: the last message) of a
+ *  NATIVE record. probe→plan advances at the probe_done call's clean
+ *  toolResult; plan→teach at the verdict marker (details.accepted true). */
+export function phaseAt3Lh(messages: readonly LhMessage[], upToIndex?: number): Phase3 {
+	let phase: Phase3 = "probe";
+	const last = Math.min(upToIndex ?? messages.length - 1, messages.length - 1);
+	for (let i = 0; i <= last; i++) {
+		const m = messages[i]!;
+		const role = (m as { role?: string }).role;
+		if (role === "toolResult") {
+			const result = m as ToolResultMessage;
+			if (
+				phase === "probe" &&
+				result.toolName === "probe_done" &&
+				!result.isError
+			) {
+				phase = "plan";
+			}
+		} else if (role === "custom") {
+			const marker = m as { customType?: string; details?: { accepted?: unknown } };
+			if (
+				phase === "plan" &&
+				marker.customType === "verdict" &&
+				marker.details?.accepted === true
+			) {
+				phase = "teach";
+			}
+		}
+	}
+	return phase;
+}

src/core/plan-blocks.ts

@@ -16,8 +16,7 @@
  */
 
 import type { PlanBlockData, PlanTree } from "./contracts.ts";
-import { derivePlanStatuses } from "./contracts.ts";
+import { derivePlanStatusesLh } from "./contracts.ts";
-import type { UIMessage } from "ai";
 import type { LhMessage } from "./native-messages.ts";
 import { newPlanBlockMessage } from "./native-messages.ts";
 import type { ToolResultMessage } from "@earendil-works/pi-ai";
@@ -32,9 +31,11 @@ export interface PlanBlockPart {
 }
 
 /** The snapshot payload (unchanged since wave 4): statuses derived AT
- *  EMISSION time from the transcript — the block never re-derives. */
+ *  EMISSION time from the transcript — the block never re-derives. Wave D:
-function snapshot(tree: PlanTree, messages: readonly UIMessage[]): PlanBlockData {
+ *  the derivation reads the NATIVE transcript (quiz CALLS only — results
-	const statuses = derivePlanStatuses(tree, messages);
+ *  feed nothing, contracts/plan.ts derivePlanStatusesLh). */
+function snapshot(tree: PlanTree, messages: readonly LhMessage[]): PlanBlockData {
+	const statuses = derivePlanStatusesLh(tree, messages);
 	return {
 		kind: "full",
 		at: new Date().toISOString(),
@@ -68,14 +69,14 @@ export function emittedPlanBlockCalls(messages: readonly LhMessage[]): Set<strin
  *  once; a later turn's rewrite gets its own block — the golden-master
  *  rule the UIMessage-era attach had) that the record hasn't already
  *  emitted. `tree` is the record's current tree (the plan tool saved it
- *  during the turn); `transcript` is the full PROJECTED transcript the
+ *  during the turn); `transcript` is the full NATIVE transcript the status
- *  status derivation reads.
+ *  derivation reads.
  */
 export function planBlocksFor(
 	record: readonly LhMessage[],
 	newMessages: readonly LhMessage[],
 	tree: PlanTree | null | undefined,
-	transcript: readonly UIMessage[],
+	transcript: readonly LhMessage[],
 ): LhMessage[] {
 	if (!tree || tree.nodes.length === 0) return [];
 	const emitted = emittedPlanBlockCalls(record);

src/core/projection.ts

@@ -1,199 +0,0 @@
-/**
- * THE ONE-WAY PROJECTION (Wave B, owner-approved 2026-09-26): native record
- * messages → AI-SDK UIMessages. The record is pi-native history; every
- * existing consumer — GET /api/chats/[id], the client's renderer, phase/
- * pending/coverage/extraction scanners — reads THIS projection until Wave
- * D replaces the client dialect. One direction only: nothing ever converts
- * UIMessages back into native messages (the lossy record-adapter direction
- * is deleted wholesale).
- *
- * GROUPING RULE (the golden master): a turn's UIMessage shape is one
- * assistant message per GROUP — consecutive assistant + toolResult
- * messages, bounded by user/marker messages; toolResults attach to the
- * tool part of their call (searching backward — an answered quiz's result
- * lands on the part that asked it, even turns later); a group's id is its
- * FIRST assistant message's id. This reproduces the client's live view
- * exactly: the continuation merge (start-chunk id = the group's id) and
- * the settled record converge to the same shape, so the settle sync never
- * re-renders what streamed.
- *
- * Plan blocks (plan-blocks.ts's splice, moved here): a plan-block marker
- * has no message of its own — its data part splices into the preceding
- * assistant group, IN FLOW right after the last successful plan_init call
- * (the presentation moment).
- */
-
-import type { UIMessage } from "ai";
-import type { AssistantMessage, ToolResultMessage } from "@earendil-works/pi-ai";
-import type { LhMarkerMessage, LhMessage } from "./native-messages.ts";
-
-import type { SubagentArtifactData, SubagentResultMeta } from "./contracts/jobs.ts";
-import type { PlanBlockData, PlanVerdictData } from "./contracts/chat.ts";
-
-/** Loose part shapes the projection emits (the client's part shapes). */
-type Part = Record<string, unknown> & { type: string };
-type LooseMessage = { id: string; role: "user" | "assistant"; parts: Part[]; metadata?: unknown };
-
-function textPart(text: string): Part {
-	return { type: "text", text, state: "done" };
-}
-
-/** The result a tool part should show, from a native toolResult. */
-function applyResult(part: Part, result: ToolResultMessage): void {
-	if (result.isError) {
-		part.state = "output-error";
-		part.rawInput = part.input;
-		const text = result.content.find((c) => c.type === "text");
-		part.errorText = text && text.type === "text" ? text.text : "The tool call failed.";
-	} else {
-		part.state = "output-available";
-		part.output = result.details;
-	}
-}
-
-/** Does this assistant group carry a successful plan_init (the splice target)? */
-function lastSuccessfulPlanInit(parts: Part[]): number {
-	let idx = -1;
-	parts.forEach((p, i) => {
-		if (p.type !== "tool-plan_init") return;
-		const ok = (p.output as { ok?: unknown } | undefined)?.ok;
-		if (ok === true) idx = i;
-	});
-	return idx;
-}
-
-/** Project the native record into client-shaped UIMessages. Pure and
- *  deterministic: the same record always projects to the same list. */
-export function projectNative(messages: readonly LhMessage[]): UIMessage[] {
-	const out: LooseMessage[] = [];
-
-	const lastAssistant = (): LooseMessage | undefined => {
-		const last = out[out.length - 1];
-		return last && last.role === "assistant" ? last : undefined;
-	};
-	const findToolPart = (toolCallId: string): Part | undefined => {
-		for (let i = out.length - 1; i >= 0; i--) {
-			for (const part of out[i]!.parts) {
-				if (part.type.startsWith("tool-") && part.toolCallId === toolCallId) return part;
-			}
-		}
-		return undefined;
-	};
-
-	for (const m of messages) {
-		const role = (m as { role?: string }).role;
-
-		if (role === "user") {
-			const content = m.content as string | Array<{ type: string; text?: string }>;
-			const texts = typeof content === "string" ? [content] : content.filter((c) => c.type === "text").map((c) => (c as { text: string }).text);
-			out.push({ id: m.id, role: "user", parts: texts.map(textPart) });
-			continue;
-		}
-
-		if (role === "custom") {
-			const marker = m as LhMarkerMessage;
-			projectMarker(marker, out);
-			continue;
-		}
-
-		if (role === "toolResult") {
-			const result = m as ToolResultMessage;
-			const part = findToolPart(result.toolCallId);
-			if (part) applyResult(part, result);
-			continue;
-		}
-
-		if (role === "assistant") {
-			const assistant = m as AssistantMessage;
-			const parts: Part[] = [];
-			let textN = 0;
-			let reasoningN = 0;
-			for (const block of assistant.content) {
-				if (block.type === "text") {
-					parts.push({ type: "text", id: `${m.id}-t${textN++}`, text: block.text, state: "done" });
-				} else if (block.type === "thinking") {
-					parts.push({ type: "reasoning", id: `${m.id}-r${reasoningN++}`, text: block.thinking, state: "done" });
-				} else if (block.type === "toolCall") {
-					parts.push({
-						type: `tool-${block.name}`,
-						toolCallId: block.id,
-						state: "input-available",
-						input: block.arguments,
-					});
-				}
-			}
-			if (parts.length === 0) continue; // presentation-only (nothing renderable)
-			const group = lastAssistant();
-			if (group) {
-				// Same group: the step separator renders between the steps.
-				group.parts.push({ type: "step-start" }, ...parts);
-			} else {
-				out.push({ id: m.id, role: "assistant", parts: [{ type: "step-start" }, ...parts] });
-			}
-			continue;
-		}
-	}
-
-	return out as unknown as UIMessage[];
-}
-
-/** A marker's projected message (verdict/handoff/steers) or splice (plan block). */
-function projectMarker(marker: LhMarkerMessage, out: LooseMessage[]): void {
-	switch (marker.customType) {
-		case "verdict": {
-			const data = marker.details as PlanVerdictData;
-			out.push({
-				id: marker.id,
-				role: "user",
-				parts: [{ type: "data-verdict", id: `vd_${marker.id}`, data }],
-			});
-			return;
-		}
-		case "probe-handoff": {
-			out.push({
-				id: marker.id,
-				role: "user",
-				parts: [{ type: "data-probe-handoff", id: `ph_${marker.id}`, data: marker.details }],
-			});
-			return;
-		}
-		case "subagent-steer": {
-			const meta = marker.details as SubagentResultMeta;
-			const artifact = marker.details as { kind: SubagentArtifactData["kind"]; result: SubagentArtifactData["result"] };
-			out.push({
-				id: marker.id,
-				role: "user",
-				metadata: { subagentResult: meta },
-				parts: [
-					textPart(marker.content),
-					{ type: "data-subagent-artifact", id: `sa_${marker.id}`, data: { kind: artifact.kind, result: artifact.result } },
-				],
-			});
-			return;
-		}
-		case "subagent-abort": {
-			const meta = marker.details as SubagentResultMeta & { aborted: true };
-			out.push({
-				id: marker.id,
-				role: "user",
-				metadata: { subagentResult: meta },
-				parts: [
-					textPart(marker.content),
-					{ type: "data-subagent-aborted", id: `sa_${marker.id}`, data: meta },
-				],
-			});
-			return;
-		}
-		case "plan-block": {
-			// No message of its own: splice into the preceding assistant
-			// group, IN FLOW after the last successful plan_init call.
-			const group = out[out.length - 1];
-			if (!group || group.role !== "assistant") return;
-			const idx = lastSuccessfulPlanInit(group.parts);
-			if (idx === -1) return;
-			const { block } = marker.details as { planCallId: string; block: PlanBlockData };
-			group.parts.splice(idx + 1, 0, { type: "data-plan-block", id: marker.id, data: block });
-			return;
-		}
-	}
-}

src/core/ui-message.ts

@@ -1,120 +0,0 @@
-import type { UIMessage } from "ai";
-import type {
-	DemoToolInput,
-	PdfToolInput,
-	PlanVerdictData,
-	PlanBlockData,
-	ProbeDoneInput,
-	ProbeDoneResult,
-	ResearchToolInput,
-	SubagentArtifactData,
-	SubagentResultMeta,
-	VisualToolInput,
-} from "./contracts.ts";
-import type { AskToolInput, QuestionAnswerData, QuizToolInput } from "./question-core.ts";
-
-/**
- * The typed UI message for the teacher chat — gives the frontend typed tool
- * parts ("tool-quiz" / "tool-ask_user_question" / "tool-research" / …) without
- * importing the server-side agent into the client bundle.
- *
- * Question tools are CLIENT-SIDE (no execute): their OUTPUT type is the
- * structured answer record the client builds and sends via addToolOutput.
- * Job tools are SERVER-EXECUTED: their output is the plain-string dispatch
- * acknowledgement (the result later arrives as a steer message — see
- * SubagentSteerMessage in contracts.ts, whose artifact rides the
- * data-subagent-artifact part typed here as DATA_PARTS).
- */
-export type TeacherUITools = {
-	quiz: { input: QuizToolInput; output: QuestionAnswerData };
-	ask_user_question: { input: AskToolInput; output: QuestionAnswerData };
-	/** The prober's conclusion tool (docs/probe-teach-split.md §2): the
-	 *  input IS the learner assessment; the output confirms the handoff. */
-	probe_done: { input: ProbeDoneInput; output: ProbeDoneResult };
-	research: { input: ResearchToolInput; output: string };
-	demo: { input: DemoToolInput; output: string };
-	visual: { input: VisualToolInput; output: string };
-	/** The pdf dispatch tool (sources v0): fire-and-forget ack output like
-	 *  research — the brief arrives later as a steer message. */
-	pdf: { input: PdfToolInput; output: string };
-	/** The learner-filesystem tools (sources v1, both agents, always-on):
-	 *  read may return an image-bearing result; the rest are plain text. */
-	read: { input: { path: string; offset?: number; limit?: number }; output: unknown };
-	ls: { input: { path?: string; limit?: number }; output: unknown };
-	find: { input: { pattern: string; path?: string; limit?: number }; output: unknown };
-	grep: {
-		input: {
-			pattern: string;
-			path?: string;
-			glob?: string;
-			ignoreCase?: boolean;
-			literal?: boolean;
-			context?: number;
-			limit?: number;
-		};
-		output: unknown;
-	};
-};
-
-export type TeacherUIDataParts = {
-	"subagent-artifact": SubagentArtifactData;
-	/** Plan blocks (wave 4): frozen plan snapshots rendered in the chat flow. */
-	"plan-block": PlanBlockData;
-	/** The learner's plan verdict (accept/decline): its own user message,
-	 *  carrying ONLY this part — the plan→teach phase seam (docs/probe-
-	 *  teach-split.md §4, the custom-message scheme 2026-09-12). */
-	verdict: PlanVerdictData;
-	/** The probe→teach handoff wake: the wake pipeline's steer message,
-	 *  carrying ONLY this part (the model-facing text lives in the adapter). */
-	"probe-handoff": { chatId: string };
-	/** The abort steer's message-type marker (the abort badge renders off it). */
-	"subagent-aborted": SubagentResultMeta & { aborted: true };
-};
-
-export type TeacherUIMessage = UIMessage<unknown, TeacherUIDataParts, TeacherUITools>;
-
-// ─── Append-only record adoption (background-process ground rules) ──────────
-
-/**
- * Merge server-added messages into the client's live list WITHOUT the full
- * snapshot swap (owner ground rules, 2026-09-15: content delivered into the
- * record — context blocks, sub-agent artifacts, abort notes — must become
- * visible the moment it persists, mid-turn included; it sits between messages
- * and nothing about a running turn should hide it).
- *
- * Safe against the stale-snapshot lesson (mid-turn full syncs break the SDK's
- * in-place continuation merge): this NEVER rewrites or removes an existing
- * message — it returns the RECORD's message order when that order is exactly
- * the client's persisted core plus new server-added ids (the server appends
- * only: steers are inserted right after the last persisted user message),
- * with the client's in-flight streaming assistant message — its last entry,
- * not yet persisted — kept at the tail. Any other shape (a missing local id
- * = the record was rewritten/reordered) returns null: the caller defers to
- * the full sync at idle, the pre-existing safety net.
- */
-export function mergeAdopted(
-	local: ReadonlyArray<TeacherUIMessage>,
-	record: ReadonlyArray<TeacherUIMessage>,
-): TeacherUIMessage[] | null {
-	const recordIds = new Set(record.map((m) => m.id));
-	// The in-flight tail: the client's last message may legitimately be absent
-	// from the record (the SDK's streaming assistant message persists only at
-	// onFinish). Exactly one such tail is allowed.
-	let core = local;
-	let tail: TeacherUIMessage | undefined;
-	const last = local.at(-1);
-	if (last && !recordIds.has(last.id)) {
-		tail = last;
-		core = local.slice(0, -1);
-	}
-	// No rewrite happened if every core message is still in the record —
-	// otherwise this is not a pure append; bail to the full-sync-at-idle.
-	if (core.some((m) => !recordIds.has(m.id))) return null;
-	// Nothing new → nothing to do.
-	const localIds = new Set(local.map((m) => m.id));
-	if (!record.some((m) => !localIds.has(m.id))) return null;
-	// The record's order is canonical for everything persisted (the server
-	// appends only; plan-block part attachment changes parts, not order) —
-	// merged = record + the in-flight tail.
-	return tail ? [...record, tail] : [...record];
-}

src/core/view.ts

@@ -0,0 +1,289 @@
+/**
+ * THE VIEW MODEL (Wave D, owner-approved 2026-09-26): the client's transcript
+ * view — native record messages (+ the live turn's working messages) grouped
+ * into today's exact render shape. The golden master is the pre-Wave-D UI
+ * (owner fork 1): everything the agent produces is ONE continuous document
+ * per run — step separators between the loop's steps, artifact and answered
+ * cards inline, the teacher's reaction continuing below an answered quiz.
+ *
+ * This is the grouping canon of the deleted server-side projection
+ * (core/projection.ts, Wave B), moved client-side, app-owned, with zero AI-SDK
+ * types — plus the LIVE-TURN layer the projection never had: mid-turn tool
+ * execution states (input-streaming → input-available/output-error/output),
+ * so the dock mounts only after a call clears the validation boundary
+ * (tool_execution_end) and the shimmer names the running activity.
+ *
+ * GROUPING RULE: one assistant view message per GROUP — consecutive assistant
+ * + toolResult messages, bounded by user/marker messages; toolResults attach
+ * to the tool part of their call (searching backward — an answered quiz's
+ * result lands on the part that asked it, even turns later); a group's id is
+ * its FIRST assistant message's id.
+ *
+ * Plan blocks: a plan-block marker has no message of its own — its data part
+ * splices into the preceding assistant group, IN FLOW right after the last
+ * successful plan_init call (the presentation moment).
+ *
+ * Question tools on the wire (owner fork 2): the agent's closing
+ * tool_execution_end carries the in-memory "Question posed" placeholder —
+ * projectView IGNORES clean question executions (the part stays
+ * input-available; the learner's answer arrives as a real toolResult). The
+ * placeholder never persists, so settled records never see it.
+ *
+ * Pure and deterministic: client-safe (no node deps), same input → same view.
+ */
+
+import type { AssistantMessage, ToolResultMessage } from "@earendil-works/pi-ai";
+import type { LhMarkerMessage, LhMessage } from "./native-messages.ts";
+import { QUESTION_TOOLS } from "./pending.ts";
+import type { SubagentArtifactData, SubagentResultMeta } from "./contracts/jobs.ts";
+import type { PlanBlockData, PlanVerdictData } from "./contracts/chat.ts";
+
+/** A view part — the loose part vocabulary the renderers consume (the same
+ *  shape names the pre-Wave-D UIMessage dialect used, so the renderer
+ *  switch and the phase/pending scanners transplant unchanged). The common
+ *  fields are optional-typed so the pre-Wave-D renderer code compiles
+ *  against it with minimal casts; anything else stays `unknown` through
+ *  the index signature. */
+export interface LhViewPart extends Record<string, unknown> {
+	type: string;
+	id?: string;
+	text?: string;
+	state?: string;
+	toolCallId?: string;
+	input?: unknown;
+	output?: unknown;
+	rawInput?: unknown;
+	errorText?: string;
+	data?: unknown;
+}
+
+/** A view message: user (bubble) or assistant (one continuous run). */
+export interface LhViewMessage {
+	id: string;
+	role: "user" | "assistant";
+	parts: LhViewPart[];
+	/** The subagentResult meta the steer cards read (marker details). */
+	metadata?: unknown;
+}
+
+/** A tool call's live execution outcome (from tool_execution_end). */
+export interface LiveExecution {
+	isError: boolean;
+	/** The clean result's details payload (the card's output). */
+	output?: unknown;
+	/** The failure text (output-error parts). */
+	errorText?: string;
+}
+
+/** The live turn's execution layer, maintained by the session hook. */
+export interface LiveTurnState {
+	/** toolCallId → its execution outcome seen so far. */
+	executions: ReadonlyMap<string, LiveExecution>;
+	/** True while the turn runs — calls with no execution yet render
+	 *  input-streaming (the shimmer); settled dangling calls render
+	 *  input-available (the dock mounts from the record). */
+	inFlight: boolean;
+}
+
+function textPart(text: string): LhViewPart {
+	return { type: "text", text, state: "done" };
+}
+
+/** The result a tool part should show, from a native toolResult. */
+function applyResult(part: LhViewPart, result: ToolResultMessage): void {
+	if (result.isError) {
+		part.state = "output-error";
+		part.rawInput = part.input;
+		const text = result.content.find((c) => c.type === "text");
+		part.errorText = text && text.type === "text" ? text.text : "The tool call failed.";
+	} else {
+		part.state = "output-available";
+		part.output = result.details;
+	}
+}
+
+/** Does this assistant group carry a successful plan_init (the splice target)? */
+function lastSuccessfulPlanInit(parts: LhViewPart[]): number {
+	let idx = -1;
+	parts.forEach((p, i) => {
+		if (p.type !== "tool-plan_init") return;
+		const ok = (p.output as { ok?: unknown } | undefined)?.ok;
+		if (ok === true) idx = i;
+	});
+	return idx;
+}
+
+/** Project the native record (+ live turn working list) into the client's
+ *  view messages. Pure: the same input always projects to the same list. */
+export function projectView(messages: readonly LhMessage[], live?: LiveTurnState): LhViewMessage[] {
+	const out: LhViewMessage[] = [];
+
+	const lastAssistant = (): LhViewMessage | undefined => {
+		const last = out[out.length - 1];
+		return last && last.role === "assistant" ? last : undefined;
+	};
+	const findToolPart = (toolCallId: string): LhViewPart | undefined => {
+		for (let i = out.length - 1; i >= 0; i--) {
+			for (const part of out[i]!.parts) {
+				if (part.type.startsWith("tool-") && part.toolCallId === toolCallId) return part;
+			}
+		}
+		return undefined;
+	};
+
+	for (const m of messages) {
+		const role = (m as { role?: string }).role;
+
+		if (role === "user") {
+			const content = m.content as string | Array<{ type: string; text?: string }>;
+			const texts =
+				typeof content === "string"
+					? [content]
+					: content.filter((c) => c.type === "text").map((c) => (c as { text: string }).text);
+			out.push({ id: m.id, role: "user", parts: texts.map(textPart) });
+			continue;
+		}
+
+		if (role === "custom") {
+			projectMarker(m as LhMarkerMessage, out);
+			continue;
+		}
+
+		if (role === "toolResult") {
+			applyResultTo(findToolPart((m as ToolResultMessage).toolCallId), m as ToolResultMessage);
+			continue;
+		}
+
+		if (role === "assistant") {
+			const assistant = m as AssistantMessage;
+			const parts: LhViewPart[] = [];
+			let textN = 0;
+			let reasoningN = 0;
+			for (const block of assistant.content) {
+				if (block.type === "text") {
+					parts.push({ type: "text", id: `${m.id}-t${textN++}`, text: block.text, state: "done" });
+				} else if (block.type === "thinking") {
+					parts.push({ type: "reasoning", id: `${m.id}-r${reasoningN++}`, text: block.thinking, state: "done" });
+				} else if (block.type === "toolCall") {
+					parts.push(toolPartOf(block, live));
+				}
+			}
+			if (parts.length === 0) continue; // presentation-only (nothing renderable)
+			const group = lastAssistant();
+			if (group) {
+				// Same group: the step separator renders between the steps.
+				group.parts.push({ type: "step-start" }, ...parts);
+			} else {
+				out.push({ id: m.id, role: "assistant", parts: [{ type: "step-start" }, ...parts] });
+			}
+			continue;
+		}
+	}
+
+	return out;
+}
+
+/** A tool call block → its view part, with the live execution layer. */
+function toolPartOf(
+	block: { id: string; name: string; arguments?: unknown },
+	live: LiveTurnState | undefined,
+): LhViewPart {
+	const part: LhViewPart = {
+		type: `tool-${block.name}`,
+		toolCallId: block.id,
+		input: block.arguments,
+	};
+	const exec = live?.executions.get(block.id);
+	const isQuestion = (QUESTION_TOOLS as readonly string[]).includes(block.name);
+	if (exec) {
+		if (exec.isError) {
+			// An invalid call NEVER becomes input-available — the dock's mount
+			// gate keys on the clean execution end (the live modal-realism
+			// crash: a malformed quiz must never mount).
+			part.state = "output-error";
+			part.rawInput = block.arguments;
+			part.errorText = exec.errorText ?? "The tool call failed.";
+		} else if (isQuestion) {
+			// Fork 2: the clean "Question posed" placeholder is IGNORED —
+			// question tools are client-owned; the dock mounts on the clean
+			// end, the answer arrives as a real toolResult.
+			part.state = "input-available";
+		} else {
+			part.state = "output-available";
+			part.output = exec.output;
+		}
+	} else {
+		// No execution yet: mid-turn it is input-streaming (the shimmer
+		// activity); in a settled record a dangling call is the dock's
+		// pending question (input-available).
+		part.state = live?.inFlight ? "input-streaming" : "input-available";
+	}
+	return part;
+}
+
+function applyResultTo(part: LhViewPart | undefined, result: ToolResultMessage): void {
+	if (part) applyResult(part, result);
+	// A result with no visible part (e.g. a call whose assistant message
+	// rendered nothing) is dropped from the view — the record keeps it.
+}
+
+/** A marker's projected message (verdict/handoff/steers) or splice (plan block). */
+function projectMarker(marker: LhMarkerMessage, out: LhViewMessage[]): void {
+	switch (marker.customType) {
+		case "verdict": {
+			const data = marker.details as PlanVerdictData;
+			out.push({
+				id: marker.id,
+				role: "user",
+				parts: [{ type: "data-verdict", id: `vd_${marker.id}`, data }],
+			});
+			return;
+		}
+		case "probe-handoff": {
+			out.push({
+				id: marker.id,
+				role: "user",
+				parts: [{ type: "data-probe-handoff", id: `ph_${marker.id}`, data: marker.details }],
+			});
+			return;
+		}
+		case "subagent-steer": {
+			const meta = marker.details as SubagentResultMeta;
+			const artifact = marker.details as {
+				kind: SubagentArtifactData["kind"];
+				result: SubagentArtifactData["result"];
+			};
+			out.push({
+				id: marker.id,
+				role: "user",
+				metadata: { subagentResult: meta },
+				parts: [
+					textPart(marker.content),
+					{ type: "data-subagent-artifact", id: `sa_${marker.id}`, data: { kind: artifact.kind, result: artifact.result } },
+				],
+			});
+			return;
+		}
+		case "subagent-abort": {
+			const meta = marker.details as SubagentResultMeta & { aborted: true };
+			out.push({
+				id: marker.id,
+				role: "user",
+				metadata: { subagentResult: meta },
+				parts: [textPart(marker.content), { type: "data-subagent-aborted", id: `sa_${marker.id}`, data: meta }],
+			});
+			return;
+		}
+		case "plan-block": {
+			// No message of its own: splice into the preceding assistant
+			// group, IN FLOW after the last successful plan_init call.
+			const group = out[out.length - 1];
+			if (!group || group.role !== "assistant") return;
+			const idx = lastSuccessfulPlanInit(group.parts);
+			if (idx === -1) return;
+			const { block } = marker.details as { planCallId: string; block: PlanBlockData };
+			group.parts.splice(idx + 1, 0, { type: "data-plan-block", id: marker.id, data: block });
+			return;
+		}
+	}
+}

src/server/chat-handler.ts

@@ -322,7 +322,6 @@ export async function handleChatInput(
 		// late views reattach via GET /api/chats/[id]/stream.
 		const handle = runPiTurn({
 			agent: constructed,
-			base: merged,
 			// Stream-level failures are provider/transport internals — masked,
 			// like the SDK defaults. Tool-input validation messages never
 			// route through here: they ride the error-result path and reach the

src/server/chats.ts

@@ -1,9 +1,7 @@
 import { abortExtraction, extractionStatuses } from "../agent/subagents/extraction.ts";
 import { coverageOf } from "../core/coverage.ts";
-import { projectNative } from "../core/projection.ts";
 import { deleteChatTraces, type ChatDeleteMode, type ChatDeletionStats } from "../core/graph-deletion.ts";
 import { titleFromPrompt } from "../core/contracts.ts";
-import type { UIMessage } from "ai";
 import { graphRevision } from "./graph.ts";
 import { stores } from "./stores.ts";
 import { isTurnRunning } from "./turn-liveness.ts";
@@ -144,7 +142,7 @@ export function sessionCoverage(chatId: string): SessionCoverage | null {
 	const entries = stores.graphStore.readCoverage();
 	// Wave B: the record is native history — coverage reads through the
 	// projection (same view the extraction pipeline folds).
-	const cov = coverageOf(projectNative(record.messages), entries, chatId);
+	const cov = coverageOf(record.messages, entries, chatId);
 	let lastFoldedAt: string | null = null;
 	for (const entry of entries) {
 		if (entry.chatId !== chatId) continue;

src/server/pi/ui-stream.ts

@@ -1,67 +1,66 @@
 import type { Agent, AgentEvent, AgentMessage } from "@earendil-works/pi-agent-core";
-import type { UIMessage } from "ai";
-import { QUESTION_TOOLS } from "../../core/pending.ts";
-import type { LhMessage } from "../../core/native-messages.ts";
-import { projectNative } from "../../core/projection.ts";
 
 /**
- * The UI-stream bridge (docs/pi-engine-swap.md §5): runs one agent turn and
+ * THE NATIVE WIRE (Wave D, owner-approved 2026-09-26): one agent turn, streamed
- * re-emits it as an AI-SDK-shaped UIMessageStream SSE response — the wire
+ * to every view as VERBATIM pi AgentEvents — no translation, no remapping, no
- * protocol `useChat` (Band D, unchanged) parses:
+ * part assembly. The SSE body is the agent's own event stream:
  *
- *   data: {"type":"text-delta", …}\n\n … data: [DONE]
+ *   data: {"type":"agent_start"}\n\n
+ *   data: {"type":"message_update","message":{…},"assistantMessageEvent":{…}}\n\n
+ *   …
+ *   data: {"type":"agent_end","messages":[…]}\n\n
+ *   data: [DONE]\n\n
  *
- * Load-bearing protocol rules (verified against ai@7's client):
+ * Load-bearing protocol facts (verified against pi-agent-core):
- * - The `start` chunk's messageId MUST be the transcript's last message id
+ * - Every `message_update` carries the FULL partial assistant message (each
- *   when it is an assistant message (the continuation id rule) — a fresh id
+ *   AssistantMessageEvent ships `partial`) — a view REPLACES its streaming
- *   there duplicates every prior part client-side.
+ *   assistant with `event.message`; no delta assembly exists anywhere.
- * - Question tools (quiz / ask_user_question): emit tool-input-available and
+ * - Question tools (quiz / ask_user_question): NO suppression (owner ruling,
- *   NEVER tool-output-available (the client throws on an output for a call it
+ *   Wave D fork 2) — the closing `tool_execution_end` with the in-memory
- *   owns). The synthetic "question posed" result lives in memory only; the
+ *   "Question posed" placeholder flows to the client, which ignores it
- *   record's part stays output-less until the learner answers.
+ *   (question tools are client-owned; the answer arrives as the input event).
- * - tool-input-available is emitted ONLY after the call cleared validation
+ *   The placeholder still never persists and never reaches the model.
- *   at the boundary (clean tool_execution_end) — the SDK's
+ * - Tool-call validation gating survives by construction: `tool_execution_end`
- *   InvalidToolInputError gating. An invalid call surfaces as
+ *   only fires after the call cleared the boundary (an invalid call arrives
- *   tool-input-error instead and NEVER becomes input-available, so a
+ *   as isError — never a clean end), so the dock's mount rule keys on it.
- *   malformed question can never mount the dock (the live modal-realism
+ * - App-level events beyond pi's taxonomy: the reattach terminal
- *   crash: GLM double-encoded `options` as a JSON string).
+ *   {"type":"turn-settled"} (served by turn-runs when no turn is running) and
- * - The WIRE assembly (below) is the client view only (Wave B): the
+ *   the loop-level {"type":"error","errorText"} (agent.continue() threw
- *   record write is the agent's own native final state — the projection
+ *   outside the event stream — mapped through the handler's masking policy).
- *   (core/projection.ts) reproduces the client's part shapes for every
+ * - Stream-level provider errors ride their message verbatim
- *   consumer (phases, plan blocks, pending, extraction) until Wave D.
+ *   (message_end with stopReason "error" + errorMessage).
  *
  * BACKGROUND-TURN RUNNER (owner ruling, 2026-09-25): a turn executes as a
  * server-side background object decoupled from any single HTTP response —
  * clients are VIEWS. runPiTurn returns the handle (event log + subscriber
  * set); piTurnResponse wraps a subscriber stream as the SSE response, and
  * the POST /api/chat response and GET /api/chats/[id]/stream are BOTH
- * exactly that. Cancelling a view detaches it ONLY — never aborts the
+ * exactly that. Cancelling a view detaches it ONLY — never aborts the turn;
- * turn; the handle's abort() is the one remote kill.
+ * the handle's abort() is the one remote kill.
  */
 
-/** SSE chunk — the flat JSON line shape the client's zod schema accepts. */
+/** The wire event: a verbatim pi AgentEvent, or one of the two app events. */
-type Chunk = Record<string, unknown> & { type: string };
+export type TurnWireEvent =
+	| AgentEvent
+	| { type: "error"; errorText: string };
 
 export interface PiTurnSpec {
 	/** The per-turn agent (stateless engine — built from the record). */
 	agent: Agent;
-	/** The turn's base: the native record + this turn's appends
+	/** The handler's error-masking policy for LOOP-level failures (the
-	 *  (write-ahead persisted already). Wave B: the agent's context and
+	 *  event stream itself is verbatim — provider errors ride their
-	 *  the record share one shape; onFinish receives the full agent state
+	 *  message's stopReason/errorMessage through untouched). */
-	 *  and the caller persists the slice past base.length. */
-	base: LhMessage[];
-	/** The handler's error-unmasking policy (validation messages stay readable). */
 	onError?: (error: unknown) => string;
 	/** The canonical end-of-turn seam: persist + usage + extraction + endTurn.
-	 *  Receives the agent's FULL final state (native messages) — the turn's
+	 *  Receives the agent's FULL final state (native messages) — the caller
-	 *  new messages are the slice past base.length. */
+	 *  persists the slice past its own base count. */
 	onFinish: (messages: AgentMessage[], isAborted: boolean) => void;
 }
 
-/** One view onto a running turn: push() replays/streams a chunk, done()
+/** One view onto a running turn: push() replays/streams an event, done()
  *  terminates the view ([DONE] + close). Cheap to be a Set member. */
 export interface PiTurnSubscriber {
-	push(chunk: Chunk): void;
+	push(event: TurnWireEvent): void;
 	done(): void;
 }
 
@@ -75,109 +74,35 @@ export interface PiTurnHandle {
 	 *  view only — the turn runs on. */
 	subscribe(): ReadableStream<Uint8Array>;
 	/** The remote kill — the ONLY way a running turn dies early: sawAbort +
-	 *  agent.abort(); the loop's agent_end does the rest (finish chunk to
+	 *  agent.abort(); the loop's agent_end does the rest (the settle seam
-	 *  every subscriber, then onFinish persists the partial transcript). */
+	 *  persists the partial transcript to every listener). */
 	abort(): void;
 }
 
-/** Map a pi stop reason to the finish chunk's enum. */
-function finishReasonOf(stopReason: string | undefined): string {
-	switch (stopReason) {
-		case "stop":
-			return "stop";
-		case "length":
-			return "length";
-		case "toolUse":
-			return "tool-calls";
-		case "error":
-			return "error";
-		default:
-			return "other";
-	}
-}
-
-function resultErrorText(result: unknown): string {
-	const content = (result as { content?: Array<{ type: string; text?: string }> } | undefined)?.content;
-	const text = content?.find((c) => c.type === "text")?.text;
-	return text ?? "The tool call failed.";
-}
-
-/** Loose part shapes the bridge assembles (the client's part shapes). */
-interface TextPart {
-	type: "text";
-	text: string;
-	state: string;
-}
-interface ReasoningPart {
-	type: "reasoning";
-	id: string;
-	text: string;
-	state: string;
-}
-interface ToolPart {
-	type: string;
-	toolCallId: string;
-	state: string;
-	input?: unknown;
-	rawInput?: unknown;
-	output?: unknown;
-	errorText?: string;
-}
-
 export function runPiTurn(spec: PiTurnSpec): PiTurnHandle {
-	const { agent, base, onFinish } = spec;
+	const { agent, onError, onFinish } = spec;
-
-	// The streaming target: continuation (same id, parts appended to a copy
-	// of the existing message) or a fresh assistant message. The continuation
-	// rule is the CLIENT's: it keys on the last UIMessage's id — read through
-	// the projection (the record is native; a turn after an answer continues
-	// the group that carries the answered quiz, which is the projection's
-	// last message).
-	const projected = projectNative(base);
-	const last = projected[projected.length - 1] as
-		| { id?: string; role?: string; parts?: UIMessage["parts"] }
-		| undefined;
-	const continuing = last?.role === "assistant" && typeof last.id === "string";
-	const messageId =
-		(continuing && last.id) || `msg_${Date.now().toString(36)}_${crypto.randomUUID().slice(0, 8)}`;
-	const baseParts: UIMessage["parts"] = continuing && last.parts ? [...last.parts] : [];
-	const parts: UIMessage["parts"] = [...baseParts];
-
-	// Stable, collision-free part ids across continuation runs (reasoning
-	// parts persist their id; text ids only need per-run uniqueness).
-	let textN = baseParts.filter((p) => p.type === "text").length;
-	let reasoningN = baseParts.filter((p) => p.type === "reasoning").length;
-	const textIds = new Map<number, string>(); // contentIndex → chunk id
-	const reasoningIds = new Map<number, string>();
-	const toolInputs = new Map<string, { toolName: string; input: unknown }>();
-	/** The streaming text/reasoning parts of THIS run, keyed by chunk id. */
-	const activeParts = new Map<string, TextPart | ReasoningPart>();
-	/** Tool parts created this run, keyed by toolCallId. */
-	const newToolParts = new Map<string, ToolPart>();
 
 	let settled = false;
 	let sawAbort = false;
-	let finishReason: string | undefined;
-	let errorMessage: string | undefined;
 
 	// THE PER-TURN EVENT LOG (memory-only — the settle-only disk writes
-	// are unchanged): every emitted chunk, in order, so any late subscriber
+	// are unchanged): every forwarded event, in order, so any late
-	// gets a FULL replay. Unbounded BY HONEST DESIGN (owner ruling
+	// subscriber gets a FULL replay. Unbounded BY HONEST DESIGN (owner
-	// 2026-09-25): a turn's chunk log is finite by construction — the agent
+	// ruling 2026-09-25): a turn's event log is finite by construction —
-	// loop ends (or the 30-minute liveness TTL bounds the pathological
+	// the agent loop ends (or the 30-minute liveness TTL bounds the
-	// never-settling case), and concurrency is human-scale (one turn per
+	// pathological never-settling case), and concurrency is human-scale
-	// chat, a few chats). No cap, no truncation: replay must never
+	// (one turn per chat, a few chats). No cap, no truncation: replay must
-	// silently drop history. The log dies with the turn at settle.
+	// never silently drop history. The log dies with the turn at settle.
-	const log: Chunk[] = [];
+	const log: TurnWireEvent[] = [];
 	const subscribers = new Set<PiTurnSubscriber>();
 
 	// Send = append to the log + fan out to every live view. A view's own
 	// enqueue failure (a cancelled reader) is swallowed by the view — the
 	// turn must still settle; onFinish persists regardless of who watches.
-	const send = (chunk: Chunk) => {
+	const send = (event: TurnWireEvent) => {
 		if (settled) return;
-		log.push(chunk);
+		log.push(event);
-		for (const sub of subscribers) sub.push(chunk);
+		for (const sub of subscribers) sub.push(event);
 	};
 	const closeAll = () => {
 		if (settled) return;
@@ -185,175 +110,23 @@ export function runPiTurn(spec: PiTurnSpec): PiTurnHandle {
 		for (const sub of subscribers) sub.done();
 		subscribers.clear();
 	};
-
+	// The settle seam: the record write is the agent's OWN final state
-	send({ type: "start", messageId });
+	// (native history — signatures, usage, provider stamps all real).
-
+	// Copy the array: the agent may outlive this call's read.
 	const settleWith = () => {
-		// Wave B: the record write is the agent's OWN final state (native
-		// history — signatures, usage, provider stamps all real); the wire
-		// assembly above served the client view only. Copy the array: the
-		// agent may outlive this call's read.
 		onFinish([...agent.state.messages], sawAbort);
 	};
 
 	const handle = (event: AgentEvent) => {
-		switch (event.type) {
+		// VERBATIM: every pi AgentEvent forwards as-is — no remapping.
-			case "turn_start":
+		send(event);
-				send({ type: "start-step" });
+		if (event.type === "message_end") {
-				parts.push({ type: "step-start" });
+			const m = event.message as { role?: string; stopReason?: string } | undefined;
-				break;
+			if (m?.role === "assistant" && m.stopReason === "aborted") sawAbort = true;
-			case "message_update": {
+		}
-				const streamEvent = event.assistantMessageEvent;
+		if (event.type === "agent_end") {
-				switch (streamEvent.type) {
+			closeAll();
-					case "text_start": {
+			settleWith();
-						const id = `${messageId}-t${textN++}`;
-						textIds.set(streamEvent.contentIndex, id);
-						const part: TextPart = { type: "text", text: "", state: "streaming" };
-						activeParts.set(id, part);
-						parts.push(part as UIMessage["parts"][number]);
-						send({ type: "text-start", id });
-						break;
-					}
-					case "text_delta": {
-						const id = textIds.get(streamEvent.contentIndex);
-						const part = id !== undefined ? (activeParts.get(id) as TextPart | undefined) : undefined;
-						if (part) part.text += streamEvent.delta;
-						if (id !== undefined) send({ type: "text-delta", id, delta: streamEvent.delta });
-						break;
-					}
-					case "text_end": {
-						const id = textIds.get(streamEvent.contentIndex);
-						const part = id !== undefined ? (activeParts.get(id) as TextPart | undefined) : undefined;
-						if (part) part.state = "done";
-						activeParts.delete(id ?? "");
-						if (id !== undefined) send({ type: "text-end", id });
-						break;
-					}
-					case "thinking_start": {
-						const id = `${messageId}-r${reasoningN++}`;
-						reasoningIds.set(streamEvent.contentIndex, id);
-						const part: ReasoningPart = { type: "reasoning", id, text: "", state: "streaming" };
-						activeParts.set(id, part);
-						parts.push(part as UIMessage["parts"][number]);
-						send({ type: "reasoning-start", id });
-						break;
-					}
-					case "thinking_delta": {
-						const id = reasoningIds.get(streamEvent.contentIndex);
-						const part = id !== undefined ? (activeParts.get(id) as ReasoningPart | undefined) : undefined;
-						if (part) part.text += streamEvent.delta;
-						if (id !== undefined) send({ type: "reasoning-delta", id, delta: streamEvent.delta });
-						break;
-					}
-					case "thinking_end": {
-						const id = reasoningIds.get(streamEvent.contentIndex);
-						const part = id !== undefined ? (activeParts.get(id) as ReasoningPart | undefined) : undefined;
-						if (part) part.state = "done";
-						activeParts.delete(id ?? "");
-						if (id !== undefined) send({ type: "reasoning-end", id });
-						break;
-					}
-					case "toolcall_start": {
-						const block = (event.message as { content: any[] }).content[streamEvent.contentIndex];
-						if (block?.type !== "toolCall") break;
-						toolInputs.set(block.id, { toolName: block.name, input: undefined });
-						const part: ToolPart = {
-							type: `tool-${block.name}`,
-							toolCallId: block.id,
-							state: "input-streaming",
-							input: undefined,
-						};
-						newToolParts.set(block.id, part);
-						parts.push(part as unknown as UIMessage["parts"][number]);
-						send({ type: "tool-input-start", toolCallId: block.id, toolName: block.name });
-						break;
-					}
-					case "toolcall_delta": {
-						const block = (event.message as { content: any[] }).content[streamEvent.contentIndex];
-						if (block?.type !== "toolCall") break;
-						send({
-							type: "tool-input-delta",
-							toolCallId: block.id,
-							inputTextDelta: streamEvent.delta,
-						});
-						break;
-					}
-					case "toolcall_end": {
-						const toolCall = streamEvent.toolCall;
-						toolInputs.set(toolCall.id, { toolName: toolCall.name, input: toolCall.arguments });
-						// Input stays input-streaming here: it only becomes available
-					// once the call clears validation at the execution boundary
-						// (tool_execution_end) — an invalid input must never reach
-						// input-available, where the dock would mount on it.
-						const part = newToolParts.get(toolCall.id);
-						if (part) part.input = toolCall.arguments;
-						break;
-					}
-				}
-				break;
-			}
-			case "tool_execution_end": {
-				const known = toolInputs.get(event.toolCallId);
-				const toolName = event.toolName || known?.toolName || "";
-				const input = known?.input;
-				const part = newToolParts.get(event.toolCallId);
-				if (event.isError) {
-					const errorText = resultErrorText(event.result);
-					if (part) {
-						part.state = "output-error";
-						part.rawInput = input;
-						part.errorText = errorText;
-					}
-					send({
-						type: "tool-input-error",
-						toolCallId: event.toolCallId,
-						toolName,
-						input,
-						errorText,
-					});
-				} else {
-					// Validation cleared: only NOW does the input become available
-					// to the client (the InvalidToolInputError gating the SDK owned).
-					if (part) {
-						part.state = "input-available";
-						part.input = input;
-					}
-					send({ type: "tool-input-available", toolCallId: event.toolCallId, toolName, input });
-					if ((QUESTION_TOOLS as readonly string[]).includes(toolName)) {
-						// Client-owned: the answer arrives via addToolOutput;
-						// the synthetic placeholder result never reaches wire
-						// or record.
-						break;
-					}
-					const output = (event.result as { details?: unknown } | undefined)?.details;
-					if (part) {
-						part.state = "output-available";
-						part.output = output;
-					}
-					send({ type: "tool-output-available", toolCallId: event.toolCallId, output });
-				}
-				break;
-			}
-			case "turn_end":
-				send({ type: "finish-step" });
-				break;
-			case "message_end":
-				if (event.message && (event.message as { role?: string }).role === "assistant") {
-					const m = event.message as { stopReason?: string; errorMessage?: string };
-					finishReason = m.stopReason;
-					if (m.stopReason === "aborted") sawAbort = true;
-					if (m.stopReason === "error") errorMessage = m.errorMessage;
-				}
-				break;
-			case "agent_end": {
-				if (errorMessage !== undefined) {
-					send({ type: "error", errorText: spec.onError?.(errorMessage) ?? errorMessage });
-				}
-				send({ type: "finish", finishReason: finishReasonOf(finishReason) });
-				closeAll();
-				settleWith();
-				break;
-			}
 		}
 	};
 	agent.subscribe(handle);
@@ -363,9 +136,12 @@ export function runPiTurn(spec: PiTurnSpec): PiTurnHandle {
 			await agent.continue();
 		} catch (err) {
 			// A loop-level failure (not a provider error — those ride the
-			// stream as stopReason "error"): surface, persist, close.
+			// stream as stopReason "error"): surface, close, persist. If the
-			send({ type: "error", errorText: spec.onError?.(err) ?? "An error occurred." });
+			// agent already settled (agent_end reached before the throw),
-			send({ type: "finish", finishReason: "error" });
+			// the seam has run — never settle twice.
+			if (settled) return;
+			console.error("[turn] agent loop failed:", err);
+			send({ type: "error", errorText: onError?.(err) ?? "An error occurred." });
 			closeAll();
 			settleWith();
 		}
@@ -379,10 +155,10 @@ export function runPiTurn(spec: PiTurnSpec): PiTurnHandle {
 					const encoder = new TextEncoder();
 					let closed = false;
 					sub = {
-						push(chunk) {
+						push(event) {
 							if (closed) return;
 							try {
-								controller.enqueue(encoder.encode(`data: ${JSON.stringify(chunk)}\n\n`));
+								controller.enqueue(encoder.encode(`data: ${JSON.stringify(event)}\n\n`));
 							} catch {
 								/* reader cancelled — stop emitting, keep running */
 								closed = true;
@@ -399,11 +175,11 @@ export function runPiTurn(spec: PiTurnSpec): PiTurnHandle {
 							}
 						},
 					};
-					// Attach THEN replay — synchronous, so no chunk can fall
+					// Attach THEN replay — synchronous, so no event can fall
 					// between replay and live (single-threaded: the agent's
 					// handle() only runs after this stack returns).
 					subscribers.add(sub);
-					for (const chunk of log) sub.push(chunk);
+					for (const event of log) sub.push(event);
 					// A subscriber that attached after settle (the unregister
 					// race) still gets the full history + [DONE].
 					if (settled) sub.done();
@@ -426,8 +202,8 @@ export function runPiTurn(spec: PiTurnSpec): PiTurnHandle {
 }
 
 /** The subscriber-shaped SSE response — the POST /api/chat response and
- *  GET /api/chats/[id]/stream are BOTH this: same headers, same wire
+ * GET /api/chats/[id]/stream are BOTH this: same headers, same wire
- *  dialect, one protocol for the client to consume. */
+ * dialect, one protocol for the client to consume. */
 export function piTurnResponse(handle: PiTurnHandle): Response {
 	return new Response(handle.subscribe(), {
 		headers: {

src/server/turn-service.ts

@@ -21,7 +21,6 @@ import type { LhMessage } from "../core/native-messages.ts";
 import { newMessageId } from "../core/native-messages.ts";
 import { QUESTION_TOOLS } from "../core/pending.ts";
 import { planBlocksFor } from "../core/plan-blocks.ts";
-import { projectNative } from "../core/projection.ts";
 import { coverageOf } from "../core/coverage.ts";
 import type { ChatStore, GraphStore } from "../core/ports.ts";
 import type { ExtractionPipeline } from "../agent/subagents/extraction.ts";
@@ -72,12 +71,10 @@ export function persistTurn(
 			stamped.push(message.id ? message : ({ ...message, id: newMessageId() } as LhMessage));
 		}
 		const tree = record.tree;
-		const blocks = planBlocksFor(
+		const blocks = planBlocksFor(record.messages, stamped, tree, [
-			record.messages,
+			...record.messages,
-			stamped,
+			...stamped,
-			tree,
+		]);
-			projectNative([...record.messages, ...stamped]),
-		);
 		chatStore.saveMessages(chatId, [...record.messages, ...stamped, ...blocks]);
 		return stamped;
 	} catch (err) {
@@ -110,7 +107,7 @@ export function triggerExtraction(
 	try {
 		const record = deps.chatStore.getChat(chatId);
 		if (!record) return;
-		const covered = coverageOf(projectNative(record.messages), deps.graphStore.readCoverage(), chatId);
+		const covered = coverageOf(record.messages, deps.graphStore.readCoverage(), chatId);
 		if (covered.uncoveredTokens < chunkTokens) return;
 		void extraction.runExtraction(chatId).catch((err) => {
 			// BYOK (owner, 2026-09-22) + key presence (provider-key interlude,

src/store/chats.ts

@@ -13,9 +13,8 @@ import { PATHS } from "../../config.ts";
 import type { ChatStore } from "../core/ports.ts";
 import type { ChatRecord, ChatSettings, ChatSummary, ExtractionRunUsage, PlanTree, TurnUsage } from "../core/contracts.ts";
 import type { LhMessage } from "../core/native-messages.ts";
-import { findPendingToolCall } from "../core/pending.ts";
+import { danglingQuestionOf } from "../core/native-messages.ts";
-import { phaseAt3 } from "../core/phases.ts";
+import { phaseAt3Lh } from "../core/phases.ts";
-import { projectNative } from "../core/projection.ts";
 
 function chatFile(id: string): string {
 	if (!/^[A-Za-z0-9_-]+$/.test(id)) throw new Error(`Invalid chat id: ${id}`);
@@ -58,7 +57,6 @@ function listChats(): ChatSummary[] {
 				// zero-back-compat wipe makes these "corrupt") must not blank the
 				// whole sidebar — same isolation getChat's per-file catch gives.
 				try {
-					const projected = projectNative(messages);
 					return {
 						// turns and extractions stay out of the list payload (list =
 						// summaries; usage aggregates ride the full record, the picker
@@ -66,13 +64,14 @@ function listChats(): ChatSummary[] {
 						...summary,
 						messageCount: messages.length,
 						// Record-derived status for the session picker: phase from the
-						// transcript's explicit markers (probe_done / plan_init — the
+						// transcript's explicit events (probe_done concluded / verdict
-						// event-driven phases, docs/probe-teach-split.md §4), pending
+						// accepted — docs/probe-teach-split.md §4), pending question
-						// question from the last unanswered quiz/ask part — both read
+						// from the native dangling-call scan. Wave D: both folds read
-						// through the one-way projection. Liveness (busy) is NOT
+						// the NATIVE record (core/phases.ts phaseAt3Lh +
-						// record-derived — the API route merges it (ChatListItem).
+						// native-messages.ts danglingQuestionOf). Liveness (busy) is
-						phase: phaseAt3(projected),
+						// NOT record-derived — the API route merges it (ChatListItem).
-						pendingQuestion: findPendingToolCall(projected) !== null,
+						phase: phaseAt3Lh(messages),
+						pendingQuestion: danglingQuestionOf(messages) !== null,
 					};
 				} catch (err) {
 					console.error(`[chat ${summary.id}] summary projection failed:`, err);

tests/chat-flow.test.ts

@@ -33,7 +33,7 @@ import { abortTurn, runningTurn, settledTurnResponse } from "../src/server/turn-
 import { chatStore } from "../src/store/chats.ts";
 import { buildAnswer, coerceCorrectAnswer, normalizeOptions, resolveCorrect, type QuestionData } from "../src/core/question-core.ts";
 import { danglingQuestionOf, newAnswerResult } from "../src/core/native-messages.ts";
-import { projectNative } from "../src/core/projection.ts";
+import { projectView } from "../src/core/view.ts";
 import type { PiEngine } from "../src/agent/models.ts";
 import type { LhMessage } from "../src/core/native-messages.ts";
 import { PATHS } from "../config.ts";
@@ -108,6 +108,18 @@ function gradedAnswer(input: typeof QUIZ_INPUT, values: string[]) {
 	return buildAnswer(qd, { values });
 }
 
+/** Parse an SSE body into its wire events (verbatim pi AgentEvents +
+ *  the app-level error/turn-settled terminals). Structural beats substring:
+ *  key order never matters. */
+function eventsOf(body: string): Array<Record<string, any>> {
+	const out: Array<Record<string, any>> = [];
+	for (const line of body.split("\n")) {
+		if (!line.startsWith("data: ") || line === "data: [DONE]") continue;
+		out.push(JSON.parse(line.slice(6)));
+	}
+	return out;
+}
+
 async function drain(res: Response): Promise<string> {
 	if (res.status !== 200) {
 		assert.fail(`expected 200, got ${res.status}: ${await res.text()}`);
@@ -213,7 +225,7 @@ describe("teacher chat flow (scripted model)", () => {
 
 		// The projection pins the client view: reasoning part + a pending
 		// (input-available) quiz part — the dock's mount state.
-		const view1 = projectNative(saved1.messages);
+		const view1 = projectView(saved1.messages);
 		const reasoningPart = view1[1]!.parts.find((p) => p.type === "reasoning");
 		assert.ok(reasoningPart && reasoningPart.type === "reasoning", "reasoning part projected");
 		const quizPart = view1[1]!.parts.find((p) => p.type === "tool-quiz");
@@ -241,7 +253,7 @@ describe("teacher chat flow (scripted model)", () => {
 		// Native: user, assistant, answer toolResult, reaction assistant.
 		assert.equal(saved2.messages.length, 4, "the answer + reaction append natively");
 		assert.equal((saved2.messages[2] as { role?: string }).role, "toolResult", "the graded answer is the call's tool result");
-		const view2 = projectNative(saved2.messages);
+		const view2 = projectView(saved2.messages);
 		assert.equal(view2.length, 2, "the reaction continues the same assistant message (one group)");
 		const finalParts = view2[1]!.parts;
 		const textPart = finalParts.find((p) => p.type === "text");
@@ -298,20 +310,28 @@ describe("teacher chat flow (scripted model)", () => {
 		await drain(ok);
 	});
 
-	it("streams tool-input-start before tool-input-available (protocol pin for the 'Writing a question…' indicator)", async () => {
+	it("streams the toolCall partial before tool_execution_start/end (protocol pin for the 'Writing a question…' shimmer)", async () => {
-		// The client's writing indicator keys off tool parts in state
+		// Wave D: the wire is VERBATIM pi AgentEvents. The client's writing
-		// 'input-streaming', which the stream processor sets from the
+		// indicator derives 'input-streaming' from a toolCall block seen in a
-		// tool-input-start chunk. If a transport change ever stopped emitting
+		// message_update with no execution yet; the dock mounts on a CLEAN
-		// it, the indicator would silently die — pin the protocol.
+		// tool_execution_end. Pin the ordering that drives both — if the
+		// passthrough ever stopped emitting one of these, the indicator or
+		// the dock would silently die.
 		const { engine } = scriptedTeacher(defaultLesson);
 		const chat = createConfiguredChat("Test: protocol pin");
 		createdChats.push(chat.id);
 		const body = await drain(await handleChatRequest(post(chat.id, { type: "text", text: "quiz me" }), agentFactory(engine)));
-		const startIdx = body.indexOf('"type":"tool-input-start"');
+		const events = eventsOf(body);
-		const availIdx = body.indexOf('"type":"tool-input-available"');
+		const callSeen = events.findIndex(
-		assert.ok(startIdx !== -1, "tool-input-start chunk present in the stream");
+			(e) => e.type === "message_update" && JSON.stringify(e.message?.content ?? []).includes('"type":"toolCall"'),
-		assert.ok(availIdx !== -1, "tool-input-available chunk present in the stream");
+		);
-		assert.ok(startIdx < availIdx, "start precedes available");
+		const execStart = events.findIndex((e) => e.type === "tool_execution_start" && e.toolCallId === "call-1");
+		const execEnd = events.findIndex((e) => e.type === "tool_execution_end" && e.toolCallId === "call-1");
+		assert.ok(callSeen !== -1, "a message_update carries the toolCall block (the shimmer source)");
+		assert.ok(execStart !== -1, "tool_execution_start present in the stream");
+		assert.ok(execEnd !== -1, "tool_execution_end present in the stream");
+		assert.ok(callSeen < execStart && execStart < execEnd, "partial → execution start → execution end");
+		assert.equal(events[execEnd]!.isError, false, "the clean end is what mounts the dock");
 	});
 
 	it("a cancelled view NEVER aborts the turn — it runs to completion (owner ruling 2026-09-25)", async () => {
@@ -375,22 +395,17 @@ describe("teacher chat flow (scripted model)", () => {
 		createdChats.push(chat.id);
 		const body = await drain(await handleChatRequest(post(chat.id, { type: "text", text: "quiz me" }), agentFactory(engine)));
 
-		// WIRE pin (the live modal-realism crash): the invalid call must NEVER
+		// WIRE pin (the live modal-realism crash): the invalid call's execution
-		// become input-available — the dock mounts on input-available parts and
+		// end is an ERROR — the dock only mounts on a CLEAN tool_execution_end,
-		// crashed on a malformed quiz before tool-input-error could replace it.
+		// so a malformed quiz can never reach it. The retry's end is clean.
-		// The boundary gates this on clean tool_execution_end.
+		const events = eventsOf(body);
-		assert.ok(
+		const badEnd = events.find((e) => e.type === "tool_execution_end" && e.toolCallId === "call-bad");
-			!body.includes('"type":"tool-input-available","toolCallId":"call-bad"'),
+		assert.ok(badEnd, "the invalid call's execution end is on the wire");
-			"the invalid call never reached input-available on the wire",
+		assert.equal(badEnd.isError, true, "the invalid call surfaced as an error execution");
-		);
+		assert.match(String(badEnd.result?.content?.[0]?.text ?? ""), /does not match any option value/);
-		assert.ok(
+		const goodEnd = events.find((e) => e.type === "tool_execution_end" && e.toolCallId === "call-good");
-			body.includes('"type":"tool-input-error","toolCallId":"call-bad"'),
+		assert.ok(goodEnd, "the valid retry's execution end is on the wire");
-			"the invalid call surfaced as tool-input-error",
+		assert.equal(goodEnd.isError, false, "the valid retry's end is CLEAN — the dock mounts on it");
-		);
-		assert.ok(
-			body.includes('"type":"tool-input-available","toolCallId":"call-good"'),
-			"the valid retry DID become input-available",
-		);
 
 		// The model was re-prompted (the error did not end the turn)…
 		assert.equal(calls.length, 2, "the invalid call triggered a second model prompt");
@@ -405,7 +420,7 @@ describe("teacher chat flow (scripted model)", () => {
 		// model's retry context — errored question calls DO persist), the
 		// retry as the dangling quiz. The projection renders the error part.
 		const saved = chatStore.getChat(chat.id)!;
-		const view = projectNative(saved.messages);
+		const view = projectView(saved.messages);
 		const quizParts = view[1]!.parts.filter((p) => p.type === "tool-quiz");
 		const failed = quizParts.find(
 			(p) => p.type === "tool-quiz" && (p as { state?: string }).state === "output-error",
@@ -446,17 +461,14 @@ describe("teacher chat flow (scripted model)", () => {
 		createdChats.push(chat.id);
 		const body = await drain(await handleChatRequest(post(chat.id, { type: "text", text: "quiz me" }), agentFactory(engine)));
 
-		assert.ok(
+		const events = eventsOf(body);
-				!body.includes('"type":"tool-input-available","toolCallId":"call-stringy"'),
+		const stringyEnd = events.find((e) => e.type === "tool_execution_end" && e.toolCallId === "call-stringy");
-				"the structurally invalid quiz never became input-available on the wire",
+		assert.ok(stringyEnd, "the invalid quiz's execution end is on the wire");
-			);
+		assert.equal(stringyEnd.isError, true, "the structural reject surfaced as an error execution");
-		assert.ok(
+		assert.match(String(stringyEnd.result?.content?.[0]?.text ?? ""), /options/);
-				body.includes('"type":"tool-input-error","toolCallId":"call-stringy"'),
-				"the invalid quiz surfaced as tool-input-error",
-			);
 
 		const saved = chatStore.getChat(chat.id)!;
-		const view = projectNative(saved.messages);
+		const view = projectView(saved.messages);
 		const failed = view[1]!.parts.find(
 				(p) => (p as { toolCallId?: string }).toolCallId === "call-stringy",
 			);
@@ -488,15 +500,18 @@ describe("teacher chat flow (scripted model)", () => {
 		createdChats.push(chat.id);
 		const body = await drain(await handleChatRequest(post(chat.id, { type: "text", text: "quiz me twice" }), agentFactory(engine)));
 
-		// (B) wire determinism: the batches' input-available chunks in call order.
+		// (B) wire determinism: question tools execute sequentially, so the
-		const availA = body.indexOf('"type":"tool-input-available","toolCallId":"call-a"');
+		// quizzes' CLEAN execution ends arrive in call order (the dock mounts
-		const availB = body.indexOf('"type":"tool-input-available","toolCallId":"call-b"');
+		// on clean ends — the mount order is the wire order).
-		assert.ok(availA !== -1 && availB !== -1, "both quizzes became available");
+		const events = eventsOf(body);
-		assert.ok(availA < availB, "call A's input became available before call B's (sequential execution)");
+		const endA = events.findIndex((e) => e.type === "tool_execution_end" && e.toolCallId === "call-a" && e.isError === false);
+		const endB = events.findIndex((e) => e.type === "tool_execution_end" && e.toolCallId === "call-b" && e.isError === false);
+		assert.ok(endA !== -1 && endB !== -1, "both quizzes' clean ends are on the wire");
+		assert.ok(endA < endB, "call A's end precedes call B's (sequential execution)");
 
 		// Saved in call order (the projection preserves the content order).
 		const saved = chatStore.getChat(chat.id)!;
-		const quizIds = projectNative(saved.messages)[1]!.parts
+		const quizIds = projectView(saved.messages)[1]!.parts
 			.filter((p) => p.type === "tool-quiz")
 			.map((p) => (p as { toolCallId: string }).toolCallId);
 		assert.deepEqual(quizIds, ["call-a", "call-b"], "parts persisted in call order");
@@ -576,7 +591,7 @@ describe("mid-turn navigation (turn liveness + write-ahead)", () => {
 		const midTurn = chatStore.getChat(chat.id)!;
 		assert.equal(midTurn.messages.length, 3, "write-ahead appended the answer tool result before finish");
 		assert.equal((midTurn.messages[2] as { role?: string }).role, "toolResult");
-		const midQuiz = projectNative(midTurn.messages)[1]!.parts.find((p) => p.type === "tool-quiz");
+		const midQuiz = projectView(midTurn.messages)[1]!.parts.find((p) => p.type === "tool-quiz");
 		assert.ok(midQuiz && midQuiz.type === "tool-quiz");
 		assert.equal((midQuiz as { state?: string }).state, "output-available", "write-ahead persisted the answer before finish");
 
@@ -585,7 +600,7 @@ describe("mid-turn navigation (turn liveness + write-ahead)", () => {
 		// Native: user, assistant, answer result, reaction — one append-only
 		// history; the projection still renders ONE assistant group.
 		assert.equal(saved2.messages.length, 4, "full turn persisted as usual at finish");
-		assert.equal(projectNative(saved2.messages).length, 2, "the projection groups the answered quiz + reaction");
+		assert.equal(projectView(saved2.messages).length, 2, "the projection groups the answered quiz + reaction");
 		assert.equal(isTurnRunning(chat.id), false, "registry cleared at finish");
 	});
 });
@@ -732,14 +747,15 @@ describe("background-turn runner (owner ruling 2026-09-25: every turn is a serve
 		// Let the turn finish; both views close at settle.
 		release();
 		const text = await secondText;
-		assert.ok(text.startsWith('data: {"type":"start"'), "replay starts from the very first chunk (the start event)");
+		const replay = eventsOf(text);
+		assert.equal(replay[0]!.type, "agent_start", "replay starts from the very first event");
 		assert.ok(text.includes("Gated reply."), "replay carries the full streamed content");
 		assert.ok(text.endsWith("data: [DONE]\n\n"), "live tail ends with the SSE terminal, same dialect as POST");
-		// Wire honesty: a normal turn settles via agent_end with NO error chunk
+		// Wire honesty: a normal turn settles with NO error event (the live
-		// (the live smoke caught a close()-throw masked into the error path
+		// smoke caught a close()-throw masked into the error path — unit
-		// — unit asserts never looked; these pins would have).
+		// asserts never looked; these pins would have).
-		assert.ok(!text.includes('"type":"error"'), "a clean turn never emits an error chunk");
+		assert.ok(!text.includes('"type":"error"'), "a clean turn never emits an error event");
-		assert.ok(text.includes('"finishReason":"stop"'), "the finish chunk carries the stop reason");
+		assert.ok(replay.some((e) => e.type === "message_end" && e.message?.stopReason === "stop"), "message_end carries the stop reason");
 		// And the first view also closes cleanly (no hang, no drop): read its
 		// remaining chunks through the reader we already hold.
 		const decoder = new TextDecoder();

tests/chat-transport.test.ts

@@ -1,118 +0,0 @@
-/**
- * 409 identity-retry transport tests (components/chat-transport.ts).
- *
- * Wave B (input events): a 409 means a turn is in flight — the event was
- * NEVER persisted (write-ahead happens only on accept), so the retry waits
- * out the turn and re-POSTs the SAME event body. Nothing merges, nothing
- * derives: the event is the unit, and re-POSTing it is idempotent-safe.
- */
-import assert from "node:assert/strict";
-import { afterEach, describe, it } from "node:test";
-import { retryingFetch } from "../components/chat-transport.ts";
-
-const realFetch = globalThis.fetch;
-afterEach(() => {
-	globalThis.fetch = realFetch;
-});
-
-interface Call {
-	url: string;
-	body: string | undefined;
-}
-
-function installFetch(routes: {
-	/** POST /api/chat → Response; receives every attempt. */
-	chatPOST: (call: Call, attempt: number) => Response;
-	/** GET /api/jobs?chatId=… → turnInFlight value per poll (poll index). */
-	jobsGET?: (poll: number) => boolean;
-}): Call[] {
-	const calls: Call[] = [];
-	let poll = 0;
-	globalThis.fetch = (async (input: RequestInfo | URL, init?: RequestInit) => {
-		const url = String(input);
-		calls.push({ url, body: init?.body as string | undefined });
-		if (url === "/api/chat" && init?.method === "POST") {
-			return routes.chatPOST(calls.at(-1)!, calls.filter((c) => c.url === "/api/chat").length);
-		}
-		if (url.startsWith("/api/jobs")) return Response.json({ turnInFlight: routes.jobsGET?.(++poll) ?? false });
-		throw new Error(`unexpected fetch: ${url}`);
-	}) as typeof globalThis.fetch;
-	return calls;
-}
-
-/** One input event — the POST body's whole unit. */
-function outgoing(input: Record<string, unknown>): string {
-	return JSON.stringify({ chatId: "chat_x", input });
-}
-
-describe("retryingFetch (409 merge-retry)", { skip: process.env.NO_SLOW ? true : undefined }, () => {
-	it("passes non-409 responses through untouched", async () => {
-		const calls = installFetch({
-			chatPOST: () => new Response("stream", { status: 200 }),
-		});
-		const fetch = retryingFetch("chat_x", { pollMs: 10, waitMaxMs: 100 });
-		const res = await fetch("/api/chat", { method: "POST", body: outgoing({ type: "text", text: "hi" }) });
-		assert.equal(res.status, 200);
-		assert.equal(calls.length, 1, "no polling, no merge, single call");
-	});
-
-	it("waits out the turn, then re-POSTs the SAME event body (identity retry)", async () => {
-		const event = { type: "text", text: "my question while the wake turn ran" };
-		const calls = installFetch({
-			chatPOST: (_call, n) => (n === 1 ? new Response("busy", { status: 409 }) : new Response("stream", { status: 200 })),
-			jobsGET: (poll) => poll <= 1, // first poll: still running; second: done
-		});
-		const fetch = retryingFetch("chat_x", { pollMs: 10, waitMaxMs: 100 });
-		const res = await fetch("/api/chat", { method: "POST", body: outgoing(event) });
-		assert.equal(res.status, 200);
-
-		// The retry carries the IDENTICAL body — the event was never
-		// persisted pre-accept, so re-sending it is the whole protocol.
-		const posts = calls.filter((c) => c.url === "/api/chat");
-		assert.equal(posts.length, 2, "one 409, one retry");
-		assert.equal(posts[1]!.body, posts[0]!.body, "the event body is byte-identical");
-		assert.deepEqual(JSON.parse(posts[1]!.body!), { chatId: "chat_x", input: event });
-	});
-
-	it("does not poll the record or merge — the event is the only payload", async () => {
-		const calls = installFetch({
-			chatPOST: (_call, n) => (n === 1 ? new Response("busy", { status: 409 }) : new Response("stream", { status: 200 })),
-		});
-		const fetch = retryingFetch("chat_x", { pollMs: 10, waitMaxMs: 100 });
-		const res = await fetch("/api/chat", { method: "POST", body: outgoing({ type: "text", text: "q" }) });
-		assert.equal(res.status, 200);
-		assert.equal(
-			calls.filter((c) => c.url.startsWith("/api/chats/")).length,
-			0,
-			"the retry never reads the record — the server owns it",
-		);
-	});
-
-	it("loops on repeated 409s, re-POSTing the same event each time", async () => {
-		const event = { type: "answer", toolCallId: "call-1", answer: { correct: true } };
-		const calls = installFetch({
-			chatPOST: (_call, n) => (n <= 2 ? new Response("busy", { status: 409 }) : new Response("stream", { status: 200 })),
-		});
-		const fetch = retryingFetch("chat_x", { pollMs: 10, waitMaxMs: 100 });
-		const res = await fetch("/api/chat", { method: "POST", body: outgoing(event) });
-		assert.equal(res.status, 200);
-		const posts = calls.filter((c) => c.url === "/api/chat");
-		assert.equal(posts.length, 3);
-		assert.ok(posts.every((c) => c.body === posts[0]!.body), "every attempt carries the same event");
-	});
-
-	it("gives up after TURN_WAIT_MAX_MS and returns the last 409 (banner as before)", async () => {
-		// turnInFlight stuck true; two polls (~1s) then force the deadline path
-		// by asserting we eventually get the 409 back — bounded by poll count.
-		let polls = 0;
-		const calls = installFetch({
-			chatPOST: () => new Response("busy", { status: 409 }),
-			jobsGET: () => true,
-		});
-		const fetch = retryingFetch("chat_x", { pollMs: 10, waitMaxMs: 100 });
-		const res = await fetch("/api/chat", { method: "POST", body: outgoing({ type: "text", text: "q" }) });
-		assert.equal(res.status, 409);
-		polls = calls.filter((c) => c.url.startsWith("/api/jobs")).length;
-		assert.ok(polls <= 15, `expected bounded polling (~10×10ms), got ${polls}`);
-	});
-});

tests/coverage.test.ts

@@ -1,7 +1,9 @@
 import assert from "node:assert/strict";
 import { describe, it } from "node:test";
 import * as fs from "node:fs";
-import type { UIMessage } from "ai";
+import type { LhMessage } from "../src/core/native-messages.ts";
+import { newAnswerResult, newPlanBlockMessage, newUserMessage } from "../src/core/native-messages.ts";
+import type { ToolResultMessage } from "@earendil-works/pi-ai";
 
 import {
 	batchRuns,
@@ -11,20 +13,47 @@ import {
 	messageFingerprint,
 	messageViewLines,
 } from "../src/core/coverage.ts";
-import type { CoverageEntry } from "../src/core/contracts.ts";
+import type { CoverageEntry } from "../src/core/contracts/graph.ts";
 
 /**
  * Extraction coverage — the pure core (vision-extraction-coverage.md,
- * settled 2026-09-15). THE MESSAGE IS THE UNIT OF COVERAGE; the fingerprint
+ * settled 2026-09-15; Wave D native rebase 2026-09-26). THE NATIVE MESSAGE
- * and the LLM prompt share ONE line builder (messageViewLines), so "covered"
+ * IS THE UNIT OF COVERAGE (owner ruling); the fingerprint and the LLM prompt
- * is by construction exactly "what the extractor saw".
+ * share ONE line builder (messageViewLines), so "covered" is by construction
+ * exactly "what the extractor saw". Batches obey the ONE INSEPARABILITY RULE
+ * (owner fork 3): a tool-call message and its tool-result messages always
+ * appear in the same chunk.
  */
 
-function user(id: string, text: string): UIMessage {
+function user(id: string, text: string): LhMessage {
-	return { id, role: "user", parts: [{ type: "text", text }] } as unknown as UIMessage;
+	return { ...newUserMessage(text), id };
 }
 
-function slot(messages: UIMessage[], id: string) {
+function teacher(id: string, text: string): LhMessage {
+	return {
+		id,
+		role: "assistant",
+		content: [{ type: "text", text }],
+	} as unknown as LhMessage;
+}
+
+/** An assistant message calling `calls` (each {id,name}) — the bond anchor. */
+function assistantWithCalls(
+	id: string,
+	calls: Array<{ id: string; name: string; args?: unknown }>,
+): LhMessage {
+	return {
+		id,
+		role: "assistant",
+		content: calls.map((c) => ({ type: "toolCall", id: c.id, name: c.name, arguments: c.args ?? {} })),
+	} as unknown as LhMessage;
+}
+
+function result(id: string, callId: string, name: string, details: unknown): LhMessage {
+	return { ...newAnswerResult(callId, name, details), id } as LhMessage;
+}
+
+function slot(messages: ReadonlyArray<LhMessage>, id: string) {
 	const found = coverageSlots(messages).find((s) => s.message.id === id);
 	assert.ok(found, `slot ${id}`);
 	return found;
@@ -37,9 +66,8 @@ describe("messageFingerprint — what changes it, what never does", () => {
 		const same = messageFingerprint(user("m1", "hello world"));
 		assert.notEqual(a, b);
 		assert.equal(a, same);
-		// Role is hashed too — a user and an assistant line never collide.
+		// Role is hashed too — a learner and a teacher line never collide.
-		const teacher = { id: "m1", role: "assistant", parts: [{ type: "text", text: "hello world" }] } as unknown as UIMessage;
+		assert.notEqual(a, messageFingerprint(teacher("m1", "hello world")));
-		assert.notEqual(a, messageFingerprint(teacher));
 	});
 
 	it("the message id is the provenance tag — part of exactly what the extractor saw", () => {
@@ -48,39 +76,61 @@ describe("messageFingerprint — what changes it, what never does", () => {
 		// extraction view byte-for-byte. (Ids are stable in practice — the
 		// write-ahead persist owns them.)
 		assert.notEqual(messageFingerprint(user("m1", "same")), messageFingerprint(user("m2", "same")));
-		// The id is the ONLY tag: the same id + content re-fingerprints equal.
+		// The same id + content re-fingerprints equal.
 		assert.equal(messageFingerprint(user("m1", "same")), messageFingerprint(user("m1", "same")));
 	});
 
-	it("data parts are invisible — plan-block attach never re-opens a covered message", () => {
+	it("marker messages are invisible — a plan-block attach never re-opens anything", () => {
-		const plain = user("m1", "teach me the chain rule");
+		// Wave D: the marker is its OWN message (not a part on another one);
-		const withData = {
+		// its view is empty, so it neither re-opens others nor itself
-			...plain,
+		// carries any learning evidence.
-			parts: [...plain.parts, { type: "data-plan-blocks", planBlocks: [] }],
+		const marker = newPlanBlockMessage("call_p1", {
-		} as unknown as UIMessage;
+			kind: "full",
-		assert.equal(messageFingerprint(plain), messageFingerprint(withData));
+			at: "2026-09-26T00:00:00.000Z",
+			current: null,
+			nodes: [],
+			edges: [],
+		}) as LhMessage;
+		assert.deepEqual(messageViewLines(marker, marker.id), []);
+		// And the message it rides beside keeps its fingerprint untouched:
+		const before = messageFingerprint(teacher("t1", "the chain rule"));
+		const withMarker = [teacher("t1", "the chain rule"), marker];
+		assert.equal(
+			messageFingerprint(withMarker[0]!),
+			before,
+			"the neighbour's fingerprint is unchanged by the attach",
+		);
 	});
 
-	it("a quiz answer merging in changes the fingerprint — and only that message's", () => {
+	it("an answered quiz is its OWN message — the asking message never re-opens", () => {
-		const asking = {
+		// Wave D: the answer arrives as the call's toolResult message; the
-			id: "a1",
+		// asking assistant message's fingerprint is STABLE (the old wire
-			role: "assistant",
+		// merged the answer into the asking group and re-opened it).
-			parts: [{ type: "tool-quiz", toolCallId: "q1", state: "input-available", input: { question: "Q?" } }],
+		const asking = assistantWithCalls("a1", [{ id: "q1", name: "quiz", args: { question: "Q?" } }]);
-		} as unknown as UIMessage;
+		const answer = result("r1", "q1", "quiz", { questionText: "Q?", correct: true });
-		const answered = {
+		const fpAsking = messageFingerprint(asking);
-			id: "a1",
+		assert.equal(fpAsking, messageFingerprint(assistantWithCalls("a1", [{ id: "q1", name: "quiz", args: { question: "Q?" } }])));
-			role: "assistant",
+		// The answer's own line carries the graded outcome:
-			parts: [
+		const lines = messageViewLines(answer, "r1");
-				{ type: "tool-quiz", toolCallId: "q1", state: "output-available", input: { question: "Q?" }, output: { questionText: "Q?", correct: true } },
+		assert.deepEqual(lines, ["QUIZ [q1] Q? — the learner answered correctly"]);
-			],
+		assert.ok(slot([asking, answer], "r1").tokens > 0, "the quiz outcome counts toward coverage tokens");
-		} as unknown as UIMessage;
+	});
-		const fpBefore = messageFingerprint(asking);
+
-		assert.notEqual(fpBefore, messageFingerprint(answered));
+	it("quiz outcome wording — dontKnow and incorrect each get their line", () => {
-		// An output-bearing line's mass counts toward coverage tokens.
+		const dk = result("r1", "q1", "quiz", { questionText: "Q?", dontKnow: true });
-		assert.ok(
+		assert.deepEqual(messageViewLines(dk, "r1"), ["QUIZ [q1] Q? — the learner did not know"]);
-			messageViewLines(answered, answered.id).join("").length >
+		const wrong = result("r2", "q2", "quiz", { questionText: "W?", correct: false, note: "mixed up signs" });
-				messageViewLines(asking, asking.id).join("").length,
+		assert.deepEqual(messageViewLines(wrong, "r2"), ["QUIZ [q2] W? — the learner answered incorrectly (note: mixed up signs)"]);
-		);
+	});
+
+	it("other clean tool results get one generic line; errors are invisible", () => {
+		const demo = result("r1", "j1", "demo", { title: "T" });
+		assert.deepEqual(messageViewLines(demo, "r1"), ["(demo tool result delivered)"]);
+		const failed = {
+			...newAnswerResult("j2", "visual", { oops: true }),
+			isError: true,
+		} as LhMessage;
+		assert.deepEqual(messageViewLines(failed, "r9"), [], "a failed call is retry context, not learning evidence");
 	});
 });
 
@@ -96,7 +146,9 @@ describe("coverageOf — uncovered is DERIVED, never dual-tracked", () => {
 			{ chatId: "c2", fp: slot(messages, "m3").fp, tokens: 1, foldedAt: "2026-09-15T00:00:00.000Z" },
 		];
 		const cov = coverageOf(messages, entries, "c1");
-		assert.deepEqual(cov.uncovered.map((s) => s.message.id), ["m3"]);
+		// `uncovered` is INDEXES (the batcher needs positions for bonding):
+		assert.deepEqual(cov.uncovered, [2]);
+		assert.deepEqual(cov.slots[cov.uncovered[0]!]!.message.id, "m3");
 		assert.equal(cov.coveredCount, 2);
 		assert.equal(cov.uncoveredCount, 1);
 		assert.equal(cov.uncoveredTokens, slot(messages, "m3").tokens);
@@ -114,7 +166,8 @@ describe("coverageOf — uncovered is DERIVED, never dual-tracked", () => {
 		assert.equal(coverageOf(messages, entries, "c1").uncoveredCount, 0);
 		const mutated = [user("m1", "aaaa!"), messages[1]!];
 		const cov = coverageOf(mutated, entries, "c1");
-		assert.deepEqual(cov.uncovered.map((s) => s.message.id), ["m1"]);
+		assert.deepEqual(cov.uncovered, [0]);
+		assert.equal(cov.slots[0]!.message.id, "m1");
 	});
 });
 
@@ -128,19 +181,106 @@ describe("batchRuns — token-budgeted, consecutive, never splits a message", ()
 			user("m3", "c".repeat(40)),
 			user("m4", "d".repeat(40)),
 		]);
-		const runs = batchRuns(slots, 45);
+		const runs = batchRuns(slots, [0, 1, 2, 3], 45);
-		assert.deepEqual(runs.map((r) => r.map((s) => s.message.id)), [["m1", "m2", "m3"], ["m4"]]);
+		assert.deepEqual(
+			runs.map((r) => r.map((s) => s.message.id)),
+			[["m1", "m2", "m3"], ["m4"]],
+		);
 	});
 
 	it("a single message larger than the budget becomes its own over-budget run", () => {
 		const slots = coverageSlots([user("m1", "x".repeat(400)), user("m2", "y".repeat(20))]);
-		const runs = batchRuns(slots, 10);
+		assert.deepEqual(
-		assert.deepEqual(runs.map((r) => r.map((s) => s.message.id)), [["m1"], ["m2"]]);
+			batchRuns(slots, [0, 1], 10).map((r) => r.map((s) => s.message.id)),
+			[["m1"], ["m2"]],
+		);
 	});
 
 	it("an empty or non-positive budget never drops messages into a void", () => {
-		assert.deepEqual(batchRuns(coverageSlots([user("m1", "hi")]), 0), []);
+		assert.deepEqual(batchRuns(coverageSlots([user("m1", "hi")]), [0], 0), []);
-		assert.deepEqual(batchRuns([], 4096), []);
+		assert.deepEqual(batchRuns([], [], 4096), []);
+	});
+
+	it("THE BOND RULE (owner fork 3): a call and its results never split, even under a tiny budget", () => {
+		// The call's assistant message is COVERED; only the (new) answer is
+		// uncovered — the asking message rides into the chunk anyway, and
+		// the two NEVER split despite a budget of 1.
+		const messages = [
+			user("u1", "topic"),
+			assistantWithCalls("a1", [{ id: "q1", name: "quiz", args: { question: "Q?" } }]),
+			result("r1", "q1", "quiz", { questionText: "Q?", correct: true }),
+		];
+		const slots = coverageSlots(messages);
+		const cov = coverageOf(
+			messages,
+			coverageSlots([messages[0]!, messages[1]!]).map((s) => ({
+				chatId: "c1",
+				fp: s.fp,
+				tokens: s.tokens,
+				foldedAt: "2026-09-26T00:00:00.000Z",
+			})),
+			"c1",
+		);
+		assert.deepEqual(cov.uncovered, [2], "only the answer is uncovered");
+		const runs = batchRuns(slots, cov.uncovered, 1);
+		assert.deepEqual(
+			runs.map((r) => r.map((s) => s.message.id)),
+			[["a1", "r1"]],
+			"the covered asking message rides along as context — one chunk, over budget, never split",
+		);
+	});
+
+	it("the reverse bond: an uncovered CALL pulls its covered results into the chunk", () => {
+		const messages = [
+			{
+				id: "a1",
+				role: "assistant",
+				content: [
+					{ type: "text", text: "Let me research that and build a demo." },
+					{ type: "toolCall", id: "j1", name: "research", arguments: { topic: "x" } },
+					{ type: "toolCall", id: "j2", name: "demo", arguments: { title: "T" } },
+				],
+			} as unknown as LhMessage,
+			result("r1", "j1", "research", { summary: "s" }),
+			result("r2", "j2", "demo", { done: true }),
+		];
+		const slots = coverageSlots(messages);
+		const cov = coverageOf(
+			messages,
+			coverageSlots([messages[1]!, messages[2]!]).map((s) => ({
+				chatId: "c1",
+				fp: s.fp,
+				tokens: s.tokens,
+				foldedAt: "2026-09-26T00:00:00.000Z",
+			})),
+			"c1",
+		);
+		assert.deepEqual(cov.uncovered, [0], "only the calling message is uncovered");
+		const runs = batchRuns(slots, cov.uncovered, 10_000);
+		assert.deepEqual(
+			runs.map((r) => r.map((s) => s.message.id)),
+			[["a1", "r1", "r2"]],
+			"all three travel together (checkpoint idempotency makes re-marking the results a no-op)",
+		);
+	});
+
+	it("adjacent messages WITHOUT a bond still batch greedily (only calls bond)", () => {
+		// Two plain teacher messages after a bonded pair: the pair is one
+		// range, the plain messages group by budget alone.
+		const messages = [
+			assistantWithCalls("a1", [{ id: "q1", name: "quiz" }]),
+			result("r1", "q1", "quiz", { questionText: "Q?", correct: true }),
+			teacher("t1", "x".repeat(40)),
+			teacher("t2", "y".repeat(40)),
+		];
+		const slots = coverageSlots(messages);
+		assert.equal(slots[0]!.tokens, 0, "a call-only assistant message has no view lines");
+		// budget 30: the pair (~12) + t1 (~14) = 26 fits; t2 overflows.
+		const runs = batchRuns(slots, [0, 2, 3], 30);
+		assert.deepEqual(
+			runs.map((r) => r.map((s) => s.message.id)),
+			[["a1", "r1", "t1"], ["t2"]],
+		);
 	});
 });
 
@@ -158,6 +298,7 @@ describe("batchTag — deterministic, ref-stable, mutation-sensitive", () => {
 		assert.notEqual(batchTag(slots), batchTag(mutated));
 	});
 });
+
 describe("the all-sessions coverage GET is dead (extraction-UX close-out, 2026-09-24)", () => {
 	it("the route file is gone — the only coverage endpoint is the Extract POST", () => {
 		// UI confirmation (ff1ec7a): nothing client-side consumes the

tests/demos.test.tsx

@@ -57,7 +57,7 @@ import { chatStore } from "../src/store/chats.ts";
 import { jobStore } from "../src/store/jobs.ts";
 import { handleChatRequest } from "../src/server/chat-handler.ts";
 import { createTeacherAgent } from "../src/agent/teacher-agent.ts";
-import { projectNative } from "../src/core/projection.ts";
+import { projectView as projectNative } from "../src/core/view.ts";
 import type { PiEngine } from "../src/agent/models.ts";
 
 /**

tests/extraction.test.ts

@@ -26,7 +26,7 @@ import { createFauxCore, fauxAssistantMessage, fauxToolCall, type FauxResponseSt
 import type { PiEngine } from "../src/agent/models.ts";
 import type { LhMessage } from "../src/core/native-messages.ts";
 import { newPlanBlockMessage } from "../src/core/native-messages.ts";
-import { projectNative } from "../src/core/projection.ts";
+import { projectView as projectNative } from "../src/core/view.ts";
 import type {
 	AuditResult,
 	CoverageEntry,
@@ -42,7 +42,7 @@ import type {
 import type { GraphStore } from "../src/core/ports.ts";
 import { abortExtraction, createExtractionPipeline, extractionStatuses, transcriptView } from "../src/agent/subagents/extraction.ts";
 import { ModelsNotConfiguredError, ProviderKeyMissingError, providerKeyMissing, type ModelSlotName } from "../src/core/settings.ts";
-import { batchTag, coverageOf, coverageSlots } from "../src/core/coverage.ts";
+import { batchTag, coverageOf, coverageSlots, messageViewLines } from "../src/core/coverage.ts";
 import { chatStore } from "../src/store/chats.ts";
 import { PATHS } from "../config.ts";
 
@@ -399,7 +399,7 @@ describe("the extraction pipeline (extractor → auditor → merger)", () => {
 		// a new event (vision-extraction-coverage.md).
 		const events = ledgers.readEvents();
 		const record = chatStore.getChat(chatId)!;
-		const expectedTag = batchTag(coverageSlots(projectNative(record.messages)));
+		const expectedTag = batchTag(coverageSlots(record.messages));
 		assert.deepEqual(
 			events.map((e) => [e.kind, e.ref, e.nodeId]),
 			[
@@ -415,12 +415,18 @@ describe("the extraction pipeline (extractor → auditor → merger)", () => {
 		// one entry per transcript message, fingerprinted content each —
 		// the same turns are never extracted twice.
 		const covered = ledgers.readCoverage();
-		assert.equal(covered.length, projectNative(record.messages).length);
+		// Wave D: the coverage unit is the VISIBLE native message — bonded
+		// riders with no view lines (a textless call message) are context,
+		// never checkpointed.
+		const visible = record.messages.filter(
+			(m) => messageViewLines(m, m.id).length > 0,
+		).length;
+		assert.equal(covered.length, visible);
 		assert.ok(covered.every((c) => c.chatId === chatId && c.fp.length === 64 && c.tokens > 0));
 		assert.equal(ledgers.readCheckpoints().length, 0, "the chat path writes coverage, not the legacy checkpoint");
 
 		assert.equal(
-			coverageOf(projectNative(record.messages), covered, chatId).uncoveredCount,
+			coverageOf(record.messages, covered, chatId).uncoveredCount,
 			0,
 			"every message is now covered",
 		);
@@ -1020,9 +1026,9 @@ describe("the chunked pipeline (extraction coverage)", () => {
 
 		// Per-batch coverage: every message checkpointed, nothing legacy.
 		const coverage = ledgers.readCoverage();
-		assert.equal(coverage.length, projectNative(record.messages).length);
+		assert.equal(coverage.length, record.messages.length);
 		assert.equal(ledgers.readCheckpoints().length, 0);
-		assert.equal(coverageOf(projectNative(record.messages), coverage, chatId).uncoveredCount, 0);
+		assert.equal(coverageOf(record.messages, coverage, chatId).uncoveredCount, 0);
 
 		// A second run is a NO-OP: nothing uncovered → no model calls, no
 		// new usage entry (no money spent on covered ground).
@@ -1075,7 +1081,7 @@ describe("the chunked pipeline (extraction coverage)", () => {
 		// Batch 1's emissions are kept and its messages stay covered.
 		assert.deepEqual(stats, { nodes: 1, edges: 0, events: 1 });
 		assert.equal(ledgers.readCoverage().length, 3, "batch 1 checkpointed its three messages");
-		const cov = coverageOf(projectNative(record.messages), ledgers.readCoverage(), chatId);
+		const cov = coverageOf(record.messages, ledgers.readCoverage(), chatId);
 		assert.equal(cov.uncoveredCount, 2, "the in-flight batch's messages stay uncovered");
 
 		// The re-run — the manual extract button — picks up ONLY the
@@ -1109,9 +1115,9 @@ describe("the chunked pipeline (extraction coverage)", () => {
 			auditor: fauxEngine([passThroughAuditStep()]),
 		});
 		await pipeline.runExtraction(chatId);
-		// Wave B: the record is native — coverage reads the projection.
+		// Wave D: the record IS native history — coverage reads it directly.
 		const record = () => chatStore.getChat(chatId)!.messages;
-		const messages = () => projectNative(record());
+		const messages = () => record();
 		assert.equal(coverageOf(messages(), ledgers.readCoverage(), chatId).uncoveredCount, 0);
 
 		// Data parts are invisible to the extraction view — a plan-block

tests/graph.test.ts

@@ -20,7 +20,7 @@ import { after, describe, it } from "node:test";
 import type { AuditResult } from "../src/core/contracts.ts";
 import type { QuestionAnswerData } from "../src/core/question-core.ts";
 import type { LhMessage } from "../src/core/native-messages.ts";
-import { projectNative } from "../src/core/projection.ts";
+import { projectView as projectNative } from "../src/core/view.ts";
 import { getGraph, graphSnapshot, mergeFindings, type GraphDeps } from "../src/core/graph.ts";
 import { coverageSlots } from "../src/core/coverage.ts";
 import { graphStore } from "../src/store/graph.ts";
@@ -33,12 +33,12 @@ const graphDeps = { chatStore, graphStore } as const satisfies GraphDeps;
 /** The session merge source for a chat — what the pipeline passes: the live
  *  transcript + its updatedAt checkpoint stamp. (Notes imports pass
  *  messages: null — see the notes-merger test below.) */
-function sessionSource(chatId: string): { messages: ReturnType<typeof projectNative>; foldedAt: string } {
+function sessionSource(chatId: string): { messages: LhMessage[]; foldedAt: string } {
 	// Wave B: the record is native history — the merger's transcript view
 	// reads through the one-way projection (same shape the extraction
 	// pipeline passes: its batch slots are projected messages).
 	const record = chatStore.getChat(chatId)!;
-	return { messages: projectNative(record.messages), foldedAt: record.updatedAt };
+	return { messages: record.messages, foldedAt: record.updatedAt };
 }
 
 // ─── ledger isolation (backup-and-clear of data/graph/*.jsonl) ──────────────
@@ -1036,7 +1036,7 @@ describe("session coverage read (extraction-UX redesign, 2026-09-24)", () => {
 
 		// Fold the first two messages: the read derives the split + the
 		// newest fold timestamp from the ledger rows.
-		const slots = coverageSlots(projectNative(chatStore.getChat(chat.id)!.messages));
+		const slots = coverageSlots(chatStore.getChat(chat.id)!.messages);
 		graphStore.appendCoverage({ chatId: chat.id, fp: slots[0]!.fp, tokens: slots[0]!.tokens, foldedAt: "2026-09-24T10:00:00.000Z" });
 		graphStore.appendCoverage({ chatId: chat.id, fp: slots[1]!.fp, tokens: slots[1]!.tokens, foldedAt: "2026-09-24T11:30:00.000Z" });
 		// Another session's row must not leak into this read.

tests/jobs.test.ts

@@ -58,7 +58,7 @@ import { providerKeyMissing, type ModelSlotName } from "../src/core/settings.ts"
 import type { PiEngine } from "../src/agent/models.ts";
 import type { SearchProvider } from "../src/core/search.ts";
 import { PATHS } from "../config.ts";
-import { projectNative } from "../src/core/projection.ts";
+import { projectView as projectNative } from "../src/core/view.ts";
 import type { LhMessage } from "../src/core/native-messages.ts";
 import type { SteerInput, TurnInput } from "../src/core/input-events.ts";
 

tests/merge-adopted.test.ts

@@ -1,74 +0,0 @@
-/**
- * mergeAdopted (core/ui-message.ts) — the append-only record-adoption seam
- * (owner ground rules, 2026-09-15): server-added steer messages must become
- * visible mid-turn without the full snapshot swap that breaks the SDK's
- * continuation merge. Pure, no React.
- */
-
-import assert from "node:assert/strict";
-import { describe, it } from "node:test";
-import { mergeAdopted, type TeacherUIMessage } from "../src/core/index.ts";
-
-function userText(id: string, text: string): TeacherUIMessage {
-	return { id, role: "user", parts: [{ type: "text", text }] };
-}
-function steer(id: string): TeacherUIMessage {
-	// Wave 7 killed the graph-context injection; its marker part is gone.
-	// The steer-shaped user message here exercises mergeAdopted's
-	// non-text user parts path with the handoff marker instead.
-	return { id, role: "user", parts: [{ type: "data-probe-handoff", data: { chatId: "c" } }] };
-}
-function assistant(id: string): TeacherUIMessage {
-	return { id, role: "assistant", parts: [{ type: "text", text: "reply" }] };
-}
-
-describe("mergeAdopted — append-only adoption", () => {
-	it("inserts a server-added steer between the persisted core and the in-flight tail", () => {
-		const local = [userText("u1", "topic"), userText("u2", "answer"), assistant("a_inflight")];
-		const record = [userText("u1", "topic"), userText("u2", "answer"), steer("steer_ctx_c")];
-		const merged = mergeAdopted(local, record);
-		assert.ok(merged);
-		assert.deepEqual(
-			merged.map((m) => m.id),
-			["u1", "u2", "steer_ctx_c", "a_inflight"],
-			"marker between the answer and the streaming bubble",
-		);
-	});
-
-	it("adopts into an idle client (no in-flight tail)", () => {
-		const local = [userText("u1", "topic")];
-		const record = [userText("u1", "topic"), steer("s1"), assistant("a1")];
-		const merged = mergeAdopted(local, record);
-		assert.deepEqual(
-			merged?.map((m) => m.id),
-			["u1", "s1", "a1"],
-		);
-	});
-
-	it("nothing new → null (no needless re-render)", () => {
-		const local = [userText("u1", "topic")];
-		assert.equal(mergeAdopted(local, [userText("u1", "topic")]), null);
-		// The tail alone is not 'new' — nothing to adopt.
-		assert.equal(mergeAdopted([assistant("a_x")], []), null);
-	});
-
-	it("a missing core id (record rewrite) → null: defer to the full sync at idle", () => {
-		// A local message vanished from the record — mid-turn this means a
-		// rewrite, and the append-only shortcut must not paper over it.
-		const local = [userText("u1", "a"), userText("u2", "b"), assistant("tail")];
-		const record = [userText("u2", "b"), steer("s1")];
-		assert.equal(mergeAdopted(local, record), null);
-	});
-
-	it("only the LAST local message may be the absent in-flight tail", () => {
-		// Two absent local messages: not a simple append — bail.
-		const local = [userText("u1", "a"), userText("u2", "gone"), assistant("tail")];
-		const record = [userText("u1", "a"), steer("s1")];
-		assert.equal(mergeAdopted(local, record), null);
-	});
-
-	it("an empty local list adopts the whole record", () => {
-		const record = [steer("s1"), assistant("a1")];
-		assert.deepEqual(mergeAdopted([], record)?.map((m) => m.id), ["s1", "a1"]);
-	});
-});

tests/phases.test.ts

@@ -10,11 +10,11 @@
  */
 import assert from "node:assert";
 import test from "node:test";
-import type { UIMessage } from "ai";
+import type { LhViewMessage } from "../src/core/view.ts";
 import { phaseAt3, phaseTimeline, type Phase3 } from "../src/core/phases.ts";
 
 /** Text-only user message. */
-function user(text: string): UIMessage {
+function user(text: string): LhViewMessage {
 	return { id: `u_${text}`, role: "user", parts: [{ type: "text", text }] };
 }
 
@@ -23,7 +23,7 @@ function assistantTool(
 	name: string,
 	state: "input-available" | "output-available" | "output-error",
 	output?: unknown,
-): UIMessage {
+): LhViewMessage {
 	return {
 		id: `a_${name}_${state}_${Math.random().toString(36).slice(2, 6)}`,
 		role: "assistant",
@@ -37,7 +37,7 @@ const PLAN_OK = { ok: true, tree: { nodes: [{ id: "n1", title: "N" }], edges: []
 /** The verdict button's user message (the custom-message scheme): its own
  *  message carrying ONLY the data-verdict marker part — no human-authored
  *  text; the model-facing wording is toLlmMessages' (native-messages.ts). */
-function verdict(accepted: boolean): UIMessage {
+function verdict(accepted: boolean): LhViewMessage {
 	return {
 		id: `u_verdict_${Math.random().toString(36).slice(2, 6)}`,
 		role: "user",
@@ -48,7 +48,7 @@ function verdict(accepted: boolean): UIMessage {
 }
 
 /** The accept button's message. */
-function planAccepted(): UIMessage {
+function planAccepted(): LhViewMessage {
 	return verdict(true);
 }
 

tests/plan-blocks.test.ts

@@ -13,7 +13,7 @@ import test from "node:test";
 import type { PlanTree } from "../src/core/contracts.ts";
 import type { LhMessage } from "../src/core/native-messages.ts";
 import { planBlocksFor } from "../src/core/plan-blocks.ts";
-import { projectNative } from "../src/core/projection.ts";
+import { projectView as projectNative } from "../src/core/view.ts";
 import type { PlanBlockData } from "../src/core/contracts.ts";
 
 // 5-node fixture: a → b → c, b → d, d → e.
@@ -83,7 +83,8 @@ function resultOf(callId: string, toolName: string, details: unknown, isError =
  *  new messages are `turn`, run over `record`. Returns the projection of
  *  the whole record after the write. */
 function projectAfter(record: LhMessage[], turn: LhMessage[], tree: PlanTree | null) {
-	const blocks = planBlocksFor(record, turn, tree, projectNative([...record, ...turn]));
+	// Wave D: the status derivation reads the NATIVE transcript.
+	const blocks = planBlocksFor(record, turn, tree, [...record, ...turn]);
 	return projectNative([...record, ...turn, ...blocks]);
 }
 
@@ -128,8 +129,8 @@ test("idempotent — a call whose block the record already carries is never re-e
 	assert.equal(blocksOf(first).length, 1);
 	// Persist replay: the same turn's messages land again over a record
 	// that already emitted the marker → no duplicate.
-	const record: LhMessage[] = [...turn, ...planBlocksFor([], turn, TREE, projectNative(turn))];
+	const record: LhMessage[] = [...turn, ...planBlocksFor([], turn, TREE, turn)];
-	const blocks2 = planBlocksFor(record, turn, TREE, projectNative(record));
+	const blocks2 = planBlocksFor(record, turn, TREE, record);
 	assert.equal(blocks2.length, 0, "no duplicate block on persist replay");
 });
 
@@ -138,7 +139,7 @@ test("a plan REWRITE (a later turn's successful plan_init) gets its own full blo
 		assistantWithCall("m1", "plan_init", { tree: TREE }),
 		resultOf("m1", "plan_init", { ok: true, tree: TREE }),
 	];
-	const record = [...first, ...planBlocksFor([], first, TREE, projectNative(first))];
+	const record = [...first, ...planBlocksFor([], first, TREE, first)];
 	const second = [
 		assistantWithCall("m2", "plan_init", { tree: TREE2 }),
 		resultOf("m2", "plan_init", { ok: true, tree: TREE2 }),

tests/reattach-transport.test.ts

@@ -1,157 +0,0 @@
-/**
- * Reattach transport tests (components/chat-transport.ts, the
- * ReattachingChatTransport half): the SDK's native resume path —
- * reconnectToStream — repointed at GET /api/chats/[id]/stream with the
- * platform contract's two adapters:
- *
- * - turn-settled (the "no turn running" TERMINAL EVENT, not a 204) →
- *   cancel + onSettled + null (the SDK's no-active-stream signal);
- * - the continuation seam — a start chunk whose messageId matches the
- *   seeded last message is rewritten to a fresh id (the SDK seeds its
- *   resume streaming message EMPTY; a replace-last would drop the seeded
- *   message's parts for the whole stream).
- *
- * Plus the framing seam: the first SSE event is peeked byte-wise (a body
- * that arrives split across reads must still resolve its boundary, and
- * the leftover bytes must ride the head of the returned stream).
- */
-import assert from "node:assert/strict";
-import { afterEach, describe, it } from "node:test";
-import { createReattachingTransport } from "../components/chat-transport.ts";
-import type { ChatInput } from "../src/core/input-events.ts";
-
-/** A no-op input-events stash (the reattach tests never send). */
-const NO_INPUT = { takeInput: () => undefined };
-
-const realFetch = globalThis.fetch;
-afterEach(() => {
-	globalThis.fetch = realFetch;
-});
-
-/** SSE-frame each object the way the server's subscriber stream does. */
-function sse(...events: unknown[]): string {
-	return events.map((e) => `data: ${JSON.stringify(e)}\n\n`).join("");
-}
-
-/** A Response whose body drips `text` in `slice`-byte reads. */
-function dripped(text: string, slice: number, status = 200): Response {
-	const bytes = new TextEncoder().encode(text);
-	const body = new ReadableStream<Uint8Array>({
-		start(controller) {
-			for (let i = 0; i < bytes.length; i += slice) {
-				controller.enqueue(bytes.slice(i, i + slice));
-			}
-			controller.close();
-		},
-	});
-	return new Response(body, { status });
-}
-
-/** Collect a UIMessageChunk stream to an array. */
-async function collect(stream: ReadableStream): Promise<unknown[]> {
-	const out: unknown[] = [];
-	const reader = stream.getReader();
-	for (;;) {
-		const { done, value } = await reader.read();
-		if (done) return out;
-		out.push(value);
-	}
-}
-
-const RECONNECT_OPTS = {
-	chatId: "chat_x",
-	trigger: "resume-stream" as const,
-	messageId: undefined,
-};
-
-describe("reattach transport (reconnectToStream)", () => {
-	it("GETs the stream endpoint for the chat", async () => {
-		const urls: string[] = [];
-		globalThis.fetch = (async (input: RequestInfo | URL) => {
-			urls.push(String(input));
-			return dripped(sse({ type: "start", messageId: "msg_run_1" }), 8);
-		}) as typeof globalThis.fetch;
-		const transport = createReattachingTransport("chat_x", {}, NO_INPUT);
-		await collect((await transport.reconnectToStream(RECONNECT_OPTS))!);
-		assert.deepEqual(urls, ["/api/chats/chat_x/stream"]);
-	});
-
-	it("turn-settled: fires onSettled, resolves null (the body is released)", async () => {
-		globalThis.fetch = (async () =>
-			dripped(sse({ type: "turn-settled" }, "[DONE]"), 8)) as typeof globalThis.fetch;
-		let settled = 0;
-		const transport = createReattachingTransport("chat_x", { onSettled: () => settled++ }, NO_INPUT);
-		const stream = await transport.reconnectToStream(RECONNECT_OPTS);
-		assert.equal(stream, null);
-		assert.equal(settled, 1);
-		// (The reader.cancel hygiene is best-effort: undici wraps Response
-		// bodies, so the source-level cancel is not observable here.)
-	});
-
-	it("204 (the SDK's native no-stream signal) resolves null untouched", async () => {
-		globalThis.fetch = (async () => new Response(null, { status: 204 })) as typeof globalThis.fetch;
-		const transport = createReattachingTransport("chat_x", {}, NO_INPUT);
-		assert.equal(await transport.reconnectToStream(RECONNECT_OPTS), null);
-	});
-
-	it("a fresh start chunk passes through the SDK parse verbatim", async () => {
-		// slice 13: the boundary lands MID-read, so the peek's read-past
-		// remainder must ride the head (the bug this pins: dropped leftovers).
-		globalThis.fetch = (async () =>
-			dripped(
-				sse(
-					{ type: "start", messageId: "msg_run_1" },
-					{ type: "text-start", id: "msg_run_1-t0" },
-					{ type: "text-delta", id: "msg_run_1-t0", delta: "hi" },
-				),
-				13,
-			)) as typeof globalThis.fetch;
-		const transport = createReattachingTransport("chat_x", {}, NO_INPUT);
-		const chunks = (await collect((await transport.reconnectToStream(RECONNECT_OPTS))!)) as Array<
-			Record<string, unknown>
-		>;
-		assert.equal(chunks[0]!.type, "start");
-		assert.equal(chunks[0]!.messageId, "msg_run_1", "no rewrite when the id is fresh");
-		assert.equal(chunks[1]!.type, "text-start");
-		assert.equal(chunks[2]!.delta, "hi", "the leftover bytes after the peek ride the head");
-	});
-
-	it("a CONTINUATION start chunk (id === seeded last message) is rewritten fresh", async () => {
-		globalThis.fetch = (async () =>
-			dripped(
-				sse({ type: "start", messageId: "msg_seed" }, { type: "text-start", id: "msg_seed-t3" }),
-				64,
-			)) as typeof globalThis.fetch;
-		const transport = createReattachingTransport("chat_x", {
-			lastMessageId: () => "msg_seed",
-		}, NO_INPUT);
-		const chunks = (await collect((await transport.reconnectToStream(RECONNECT_OPTS))!)) as Array<
-			Record<string, unknown>
-		>;
-		assert.equal(chunks[0]!.type, "start");
-		assert.notEqual(chunks[0]!.messageId, "msg_seed");
-		assert.match(String(chunks[0]!.messageId), /^msg_[a-z0-9]+_[a-f0-9]{8}$/, "the server's id shape");
-		// Part ids stay server-prefixed — nothing keys parts by message id.
-		assert.equal(chunks[1]!.id, "msg_seed-t3");
-	});
-
-	it("survives a body split byte-by-byte across reads", async () => {
-		globalThis.fetch = (async () =>
-			dripped(sse({ type: "start", messageId: "msg_run_1" }, { type: "finish", finishReason: "stop" }), 1)) as typeof globalThis.fetch;
-		const transport = createReattachingTransport("chat_x", {}, NO_INPUT);
-		const chunks = (await collect((await transport.reconnectToStream(RECONNECT_OPTS))!)) as Array<
-			Record<string, unknown>
-		>;
-		assert.deepEqual(
-			chunks.map((c) => c.type),
-			["start", "finish"],
-		);
-		assert.equal(chunks[0]!.messageId, "msg_run_1");
-	});
-
-	it("an unhappy endpoint throws (the SDK surfaces the error banner)", async () => {
-		globalThis.fetch = (async () => new Response("boom", { status: 500 })) as typeof globalThis.fetch;
-		const transport = createReattachingTransport("chat_x", {}, NO_INPUT);
-		await assert.rejects(() => transport.reconnectToStream(RECONNECT_OPTS), /boom/);
-	});
-});

tests/session-core.test.ts

@@ -0,0 +1,552 @@
+/**
+ * SESSION MODEL tests (components/session-core.ts — Wave D): the pure state
+ * machine that replaced useChat + the AI-SDK transport. Everything the old
+ * transport/merge suites pinned survives here on the native wire:
+ *
+ * - the 409 IDENTITY-RETRY (postInputEvent): the event was never persisted
+ *   pre-acceptance, so the retry waits out the racing turn and re-POSTs
+ *   the SAME body — no merge, no record read;
+ * - the reattach path (resumeStream): record refetch, then the SAME
+ *   verbatim-AgentEvent consume as a POST; turn-settled is the terminal;
+ *   byte-split SSE frames never corrupt (the framing pin);
+ * - append-only ADOPTION (adoptRecord/absorbed): server-added steers show
+ *   mid-turn, the record rewrite bails to the idle full sync, and the
+ *   optimistic entries the write-ahead absorbed never double their
+ *   bubbles;
+ * - the view application (applyWireEvent → projectView): full-partial
+ *   replaces, the fork-2 question-placeholder ignore, execution states,
+ *   raw error display (owner ruling);
+ * - the continuation fold (needsContinuation): the mount/sync seam's
+ *   native replacement for lastAssistantMessageIsCompleteWithToolCalls.
+ */
+import assert from "node:assert/strict";
+import { afterEach, describe, it } from "node:test";
+import { absorbed, pendingCallTool, SessionModel } from "../components/session-core.ts";
+import type { ChatInput } from "../src/core/input-events.ts";
+import type { LhMessage } from "../src/core/native-messages.ts";
+import { newUserMessage, newAnswerResult, newVerdictMessage, needsContinuation } from "../src/core/native-messages.ts";
+
+const realFetch = globalThis.fetch;
+afterEach(() => {
+	globalThis.fetch = realFetch;
+});
+
+/** SSE-frame each object the way the server's subscriber stream does. */
+function sse(...events: unknown[]): string {
+	return events.map((e) => (e === "[DONE]" ? "data: [DONE]\n\n" : `data: ${JSON.stringify(e)}\n\n`)).join("");
+}
+
+/** A Response whose body drips `text` in `slice`-byte reads (the framing pin). */
+function dripped(text: string, slice: number, status = 200): Response {
+	const bytes = new TextEncoder().encode(text);
+	const body = new ReadableStream<Uint8Array>({
+		start(controller) {
+			for (let i = 0; i < bytes.length; i += slice) {
+				controller.enqueue(bytes.slice(i, i + slice));
+			}
+			controller.close();
+		},
+	});
+	return new Response(body, { status });
+}
+
+/** A plain single-read 200 body. */
+function sseResponse(...events: unknown[]): Response {
+	return new Response(sse(...events), { status: 200 });
+}
+
+function assistantMsg(id: string, text: string): LhMessage {
+	return {
+		id,
+		role: "assistant",
+		content: [{ type: "text", text }],
+		stopReason: "stop",
+		usage: { input: 0, output: 0, total: 0, cacheReads: 0, cacheWrites: 0 },
+		timestamp: 0,
+		provider: "openrouter",
+		model: "test",
+	} as unknown as LhMessage;
+}
+
+/** A streaming assistant partial (pi events carry the FULL partial). */
+function partial(id: string, blocks: unknown[], stopReason?: string): Record<string, unknown> {
+	return {
+		type: "message_update",
+		message: {
+			id,
+			role: "assistant",
+			content: blocks,
+			...(stopReason ? { stopReason } : {}),
+			usage: { input: 0, output: 0, total: 0, cacheReads: 0, cacheWrites: 0 },
+			provider: "openrouter",
+			model: "test",
+		},
+		assistantMessageEvent: { type: "text", textDelta: "" },
+	};
+}
+
+function model(opts: Partial<ConstructorParameters<typeof SessionModel>[0]> = {}): SessionModel {
+	return new SessionModel({ chatId: "chat_x", initialMessages: [], ...opts });
+}
+
+// ─── the 409 identity-retry ────────────────────────────────────────────────
+
+interface Call {
+	url: string;
+	body: string | undefined;
+}
+
+function installFetch(routes: {
+	chatPOST: (call: Call, attempt: number) => Response;
+	jobsGET?: (poll: number) => boolean;
+}): Call[] {
+	const calls: Call[] = [];
+	let poll = 0;
+	globalThis.fetch = (async (input: RequestInfo | URL, init?: RequestInit) => {
+		const url = String(input);
+		calls.push({ url, body: init?.body as string | undefined });
+		if (url === "/api/chat" && init?.method === "POST") {
+			return routes.chatPOST(calls.at(-1)!, calls.filter((c) => c.url === "/api/chat").length);
+		}
+		if (url.startsWith("/api/jobs")) return Response.json({ turnInFlight: routes.jobsGET?.(++poll) ?? false });
+		throw new Error(`unexpected fetch: ${url}`);
+	}) as typeof globalThis.fetch;
+	return calls;
+}
+
+describe("postInputEvent (the 409 identity-retry)", { skip: process.env.NO_SLOW ? true : undefined }, () => {
+	it("passes non-409 responses through untouched", async () => {
+		const calls = installFetch({ chatPOST: () => new Response("stream", { status: 200 }) });
+		const m = model({ pollMs: 10, waitMaxMs: 100 });
+		const res = await m.postInputEvent({ type: "text", text: "hi" });
+		assert.equal(res.status, 200);
+		assert.equal(calls.length, 1, "no polling, single call");
+		assert.deepEqual(JSON.parse(calls[0]!.body!), { chatId: "chat_x", input: { type: "text", text: "hi" } });
+	});
+
+	it("waits out the turn, then re-POSTs the SAME event body (identity retry)", async () => {
+		const event: ChatInput = { type: "text", text: "my question while the wake turn ran" };
+		const calls = installFetch({
+			chatPOST: (_c, n) => (n === 1 ? new Response("busy", { status: 409 }) : new Response("stream", { status: 200 })),
+			jobsGET: (poll) => poll <= 1, // first poll: still running; second: done
+		});
+		const m = model({ pollMs: 10, waitMaxMs: 100 });
+		const res = await m.postInputEvent(event);
+		assert.equal(res.status, 200);
+		const posts = calls.filter((c) => c.url === "/api/chat");
+		assert.equal(posts.length, 2, "one 409, one retry");
+		assert.equal(posts[1]!.body, posts[0]!.body, "the event body is byte-identical");
+	});
+
+	it("never reads the record — the event is the only payload", async () => {
+		const calls = installFetch({
+			chatPOST: (_c, n) => (n === 1 ? new Response("busy", { status: 409 }) : new Response("stream", { status: 200 })),
+		});
+		const m = model({ pollMs: 10, waitMaxMs: 100 });
+		assert.equal((await m.postInputEvent({ type: "text", text: "q" })).status, 200);
+		assert.equal(
+			calls.filter((c) => c.url.startsWith("/api/chats/")).length,
+			0,
+			"the retry never reads the record — the server owns it",
+		);
+	});
+
+	it("loops on repeated 409s, re-POSTing the same event each time", async () => {
+		const event: ChatInput = { type: "answer", toolCallId: "call-1", answer: { correct: true } as never };
+		const calls = installFetch({
+			chatPOST: (_c, n) => (n <= 2 ? new Response("busy", { status: 409 }) : new Response("stream", { status: 200 })),
+		});
+		const m = model({ pollMs: 10, waitMaxMs: 100 });
+		assert.equal((await m.postInputEvent(event)).status, 200);
+		const posts = calls.filter((c) => c.url === "/api/chat");
+		assert.equal(posts.length, 3);
+		assert.ok(posts.every((c) => c.body === posts[0]!.body), "every attempt carries the same event");
+	});
+
+	it("gives up after the deadline and surfaces the last 409", async () => {
+		const calls = installFetch({
+			chatPOST: () => new Response("busy", { status: 409 }),
+			jobsGET: () => true,
+		});
+		const m = model({ pollMs: 10, waitMaxMs: 100 });
+		const res = await m.postInputEvent({ type: "text", text: "q" });
+		assert.equal(res.status, 409);
+		const polls = calls.filter((c) => c.url.startsWith("/api/jobs")).length;
+		assert.ok(polls <= 15, `expected bounded polling (~10×10ms), got ${polls}`);
+	});
+});
+
+// ─── send: optimistic appends + wire consumption ────────────────────────────
+
+describe("send (one input event, verbatim consumption)", () => {
+	it("optimistically appends the text, POSTs {chatId, input}, consumes to [DONE]", async () => {
+		const posts: string[] = [];
+		globalThis.fetch = (async (input: RequestInfo | URL, init?: RequestInit) => {
+			const url = String(input);
+			if (url === "/api/chat" && init?.method === "POST") {
+				posts.push(init.body as string);
+				return sseResponse(
+					{ type: "agent_start" },
+					{ type: "message_start", message: { role: "assistant", id: "a1", content: [] } },
+					partial("a1", [{ type: "text", text: "hello there" }]),
+					{ type: "message_end", message: { role: "assistant", id: "a1", content: [{ type: "text", text: "hello there" }], stopReason: "stop" } },
+					{ type: "agent_end", messages: [] },
+					"[DONE]",
+				);
+			}
+			throw new Error(`unexpected fetch: ${url}`);
+		}) as typeof globalThis.fetch;
+
+		const m = model();
+		const pending = m.send({ type: "text", text: "my topic" });
+		// The optimistic bubble is INSTANT (the instant-answer UX seam):
+		assert.equal(m.getState().status, "submitted");
+		assert.match((m.getState().nativeMessages.at(-1) as { id?: string }).id ?? "", /^c_/);
+		assert.equal(m.getState().messages.at(-1)!.parts[0]!.text, "my topic");
+		await pending;
+		assert.deepEqual(JSON.parse(posts[0]!), { chatId: "chat_x", input: { type: "text", text: "my topic" } });
+		assert.equal(m.getState().status, "ready");
+		// The streamed assistant REPLACED through full partials (never deltas):
+		const view = m.getState().messages;
+		assert.equal(view.at(-1)!.role, "assistant");
+		assert.equal(view.at(-1)!.parts.at(-1)!.text, "hello there");
+	});
+
+	it("a 503 refusal surfaces the RAW body text (parseModelsRefusal's contract)", async () => {
+		globalThis.fetch = (async () =>
+			new Response(JSON.stringify({ code: "models_not_configured", missing: ["strong"] }), {
+				status: 503,
+			})) as typeof globalThis.fetch;
+		const m = model();
+		await m.send({ type: "text", text: "q" });
+		const state = m.getState();
+		assert.equal(state.status, "error");
+		assert.equal(
+			state.error!.message,
+			JSON.stringify({ code: "models_not_configured", missing: ["strong"] }),
+			"the banner carries the raw envelope — the refusal flow JSON-parses it",
+		);
+	});
+
+	it("the optimistic answer result attaches to the pending quiz part at once", async () => {
+		const quizCall = {
+			id: "call-1",
+			role: "assistant",
+			content: [
+				{ type: "text", text: "Q?" },
+				{ type: "toolCall", id: "call-1", name: "quiz", arguments: { question: "Q?", choices: ["a", "b"] } },
+			],
+			stopReason: "toolUse",
+			usage: { input: 0, output: 0, total: 0, cacheReads: 0, cacheWrites: 0 },
+			provider: "openrouter",
+			model: "test",
+			timestamp: 0,
+		} as unknown as LhMessage;
+		let postBody = "";
+		globalThis.fetch = (async (input: RequestInfo | URL, init?: RequestInit) => {
+			if (String(input) === "/api/chat") {
+				postBody = init?.body as string;
+				return sseResponse("[DONE]");
+			}
+			throw new Error(`unexpected fetch: ${String(input)}`);
+		}) as typeof globalThis.fetch;
+		const m = model({ initialMessages: [newUserMessage("hi"), quizCall] });
+		const pending = m.submitAnswer("quiz", "call-1", { correct: true } as never);
+		// The dock's AnsweredCard state — before any server round-trip:
+		const state = m.getState();
+		const quizPart = state.messages
+			.flatMap((v) => v.parts)
+			.find((p) => p.type === "tool-quiz")!;
+		assert.equal(quizPart.state, "output-available");
+		assert.deepEqual(quizPart.output, { correct: true });
+		await pending;
+		assert.deepEqual(JSON.parse(postBody), {
+			chatId: "chat_x",
+			input: { type: "answer", toolCallId: "call-1", answer: { correct: true } },
+		});
+	});
+
+	it("ignores the question placeholder's clean execution end (fork 2) and keeps errors", async () => {
+		// The dock's mount gate keys on the part STATE the live execution
+		// layer drives MID-STREAM: a clean question execution stays
+		// input-available (the in-memory "Question posed" placeholder
+		// never shows); an INVALID call is output-error and NEVER mounts
+		// (the modal-realism crash pin).
+		const enc = new TextEncoder();
+		let push: (e: unknown) => void = () => {};
+		const body = new ReadableStream<Uint8Array>({
+			start(controller) {
+				push = (e) => controller.enqueue(enc.encode(e === "[DONE]" ? "data: [DONE]\n\n" : `data: ${JSON.stringify(e)}\n\n`));
+			},
+		});
+		globalThis.fetch = (async () => new Response(body, { status: 200 })) as typeof globalThis.fetch;
+		const m = model();
+		const done = m.send({ type: "text", text: "q" });
+		const settle = () => new Promise((r) => setTimeout(r, 0));
+		push({
+			type: "message_start",
+			message: {
+				role: "assistant",
+				id: "a1",
+				content: [
+					{ type: "toolCall", id: "q1", name: "quiz", arguments: {} },
+					{ type: "toolCall", id: "bad1", name: "quiz", arguments: {} },
+				],
+			},
+		});
+		await settle();
+		push({ type: "tool_execution_end", toolCallId: "q1", toolName: "quiz", isError: false, result: { content: [{ type: "text", text: "Question posed" }], details: { placeholder: true } } });
+		push({ type: "tool_execution_end", toolCallId: "bad1", toolName: "quiz", isError: true, result: { content: [{ type: "text", text: "correctAnswer does not match any option value" }], details: undefined } });
+		await settle();
+		const parts = m.getState().messages.flatMap((v) => v.parts).filter((p) => p.type === "tool-quiz");
+		assert.equal(parts.length, 2);
+		assert.equal(parts[0]!.state, "input-available", "the clean placeholder is IGNORED — the dock mounts on it");
+		assert.equal(parts[1]!.state, "output-error");
+		assert.equal(parts[1]!.errorText, "correctAnswer does not match any option value");
+		push({ type: "agent_end", messages: [] });
+		push("[DONE]");
+		await done;
+	});
+
+	it("a provider error rides VERBATIM (owner ruling) — no masking anywhere", async () => {
+		globalThis.fetch = (async () =>
+			sseResponse(
+				partial("a1", [{ type: "text", text: "" }]),
+				{ type: "message_end", message: { role: "assistant", id: "a1", content: [{ type: "text", text: "" }], stopReason: "error", errorMessage: "OpenRouter 429: rate limited" } },
+				{ type: "agent_end", messages: [] },
+				"[DONE]",
+			)) as typeof globalThis.fetch;
+		const m = model();
+		await m.send({ type: "text", text: "q" });
+		assert.equal(m.getState().status, "ready");
+		assert.equal(m.getState().error!.message, "OpenRouter 429: rate limited");
+	});
+});
+
+// ─── reattach (resumeStream) ────────────────────────────────────────────────
+
+describe("resumeStream (the reattach path)", () => {
+	it("refetches the record, then GETs the stream endpoint", async () => {
+		const urls: string[] = [];
+		globalThis.fetch = (async (input: RequestInfo | URL) => {
+			const url = String(input);
+			urls.push(url);
+			if (url === "/api/chats/chat_x") return Response.json({ messages: [] });
+			if (url === "/api/chats/chat_x/stream")
+				return dripped(sse({ type: "message_start", message: { role: "assistant", id: "a1", content: [{ type: "text", text: "hi" }] } }, "[DONE]"), 8);
+			throw new Error(`unexpected fetch: ${url}`);
+		}) as typeof globalThis.fetch;
+		const m = model();
+		await m.resumeStream();
+		assert.deepEqual(urls, ["/api/chats/chat_x", "/api/chats/chat_x/stream"]);
+		assert.equal(m.getState().status, "ready");
+	});
+
+	it("turn-settled fires the settle hook and ends cleanly", async () => {
+		let settled = 0;
+		globalThis.fetch = (async (input: RequestInfo | URL) => {
+			const url = String(input);
+			if (url === "/api/chats/chat_x") return Response.json({ messages: [] });
+			if (url === "/api/chats/chat_x/stream") return dripped(sse({ type: "turn-settled" }, "[DONE]"), 8);
+			throw new Error(`unexpected fetch: ${url}`);
+		}) as typeof globalThis.fetch;
+		const m = model({ onReattachSettled: () => settled++ });
+		await m.resumeStream();
+		assert.equal(settled, 1);
+		assert.equal(m.getState().status, "ready");
+		assert.equal(m.getState().error, null);
+	});
+
+	it("a replayed CONTINUATION mounts BESIDE the seeded message (the SDK replace-last seam is dead)", async () => {
+		// The seeded record ends with an assistant message; the replayed
+		// turn continues with its OWN assistant message — the view groups
+		// them as one document with a step separator (pixel-identical to a
+		// live continuation; the settle sync lands the canon).
+		globalThis.fetch = (async (input: RequestInfo | URL) => {
+			const url = String(input);
+			if (url === "/api/chats/chat_x") return Response.json({ messages: [assistantMsg("seed", "the question")] });
+			if (url === "/api/chats/chat_x/stream")
+				return sseResponse(
+					{ type: "message_start", message: { role: "assistant", id: "run1", content: [{ type: "text", text: "the reaction" }] } },
+					{ type: "message_end", message: { role: "assistant", id: "run1", content: [{ type: "text", text: "the reaction" }], stopReason: "stop" } },
+					{ type: "agent_end", messages: [] },
+					"[DONE]",
+				);
+			throw new Error(`unexpected fetch: ${url}`);
+		}) as typeof globalThis.fetch;
+		const m = model({ initialMessages: [assistantMsg("seed", "the question")] });
+		await m.resumeStream();
+		const view = m.getState().messages;
+		assert.equal(view.length, 1, "one continuous assistant document");
+		assert.deepEqual(
+			view[0]!.parts.map((p) => p.type),
+			["step-start", "text", "step-start", "text"],
+			"seeded message keeps rendering; the run mounts with a step separator",
+		);
+	});
+
+	it("survives a body split byte-by-byte across reads (the framing pin)", async () => {
+		globalThis.fetch = (async (input: RequestInfo | URL) => {
+			const url = String(input);
+			if (url === "/api/chats/chat_x") return Response.json({ messages: [] });
+			if (url === "/api/chats/chat_x/stream")
+				return dripped(
+					sse(
+						{ type: "message_start", message: { role: "assistant", id: "a1", content: [] } },
+						partial("a1", [{ type: "text", text: "hi" }]),
+						{ type: "message_end", message: { role: "assistant", id: "a1", content: [{ type: "text", text: "hi" }], stopReason: "stop" } },
+					) + sse("[DONE]"),
+					1,
+				);
+			throw new Error(`unexpected fetch: ${url}`);
+		}) as typeof globalThis.fetch;
+		const m = model();
+		await m.resumeStream();
+		assert.equal(m.getState().messages.at(-1)!.parts.at(-1)!.text, "hi");
+	});
+
+	it("an unhappy endpoint surfaces the error banner", async () => {
+		globalThis.fetch = (async (input: RequestInfo | URL) => {
+			const url = String(input);
+			if (url === "/api/chats/chat_x") return Response.json({ messages: [] });
+			if (url === "/api/chats/chat_x/stream") return new Response("boom", { status: 500 });
+			throw new Error(`unexpected fetch: ${url}`);
+		}) as typeof globalThis.fetch;
+		const m = model();
+		await m.resumeStream();
+		const state = m.getState();
+		assert.equal(state.status, "error");
+		assert.match(state.error!.message, /boom/);
+	});
+});
+
+// ─── adoption (append-only ground rules, owner 2026-09-15) ──────────────────
+
+function userText(id: string, text: string): LhMessage {
+	return { ...newUserMessage(text), id };
+}
+
+describe("adoptRecord — append-only adoption", () => {
+	it("splices a server-added steer between the record and the live tail", () => {
+		const m = model({ initialMessages: [userText("u1", "topic")] });
+		m.applyWireEvent({ type: "message_start", message: { role: "assistant", id: "a_live", content: [{ type: "text", text: "streaming" }] } });
+		const steer = { ...newUserMessage("steered content"), id: "steer_1" } as LhMessage & { role: "user" };
+		const record = [userText("u1", "topic"), steer];
+		m.adoptRecord(record);
+		assert.deepEqual(
+			m.getState().nativeMessages.map((x) => x.id),
+			["u1", "steer_1", "a_live"],
+			"the steer shows NOW, the live tail stays last",
+		);
+	});
+
+	it("a missing base id (record rewrite) bails — the idle full sync is the net", () => {
+		const m = model({ initialMessages: [userText("u1", "a"), userText("u2", "b")] });
+		m.adoptRecord([userText("u2", "b")]); // u1 vanished: a rewrite, not an append
+		assert.deepEqual(
+			m.getState().nativeMessages.map((x) => x.id),
+			["u1", "u2"],
+			"the adoption is a no-op on a non-append",
+		);
+	});
+
+	it("absorbs the optimistic entries the record has taken over (no doubled bubbles)", () => {
+		// A text send's optimistic bubble (local c_ id): the write-ahead's
+		// server copy absorbs it by CONTENT — the bubble never doubles.
+		const optimistic = { ...newUserMessage("my question"), id: "c_local_1" } as LhMessage;
+		const serverCopy = { ...newUserMessage("my question"), id: "srv_9" } as LhMessage;
+		assert.ok(absorbed(optimistic, [userText("u1", "topic"), serverCopy]), "content match absorbs the local copy");
+		assert.ok(!absorbed(serverCopy, [optimistic]), "server stamps are never absorbed");
+		assert.ok(!absorbed({ ...newUserMessage("my question") }, [serverCopy]), "an id-less message is never absorbed");
+		// The answer result absorbs by toolCallId (ids differ, content may too —
+		// the toolCallId is the identity of an answer).
+		const answer = { ...newAnswerResult("call-1", "quiz", { correct: true }), id: "c_local_2" } as LhMessage;
+		assert.ok(
+			absorbed(answer, [{ ...newAnswerResult("call-1", "quiz", { correct: true }), id: "srv_10" } as LhMessage]),
+			"the answer result absorbs by toolCallId",
+		);
+		// A verdict marker absorbs by accepted (the timestamps differ by design).
+		const verdict = { ...newVerdictMessage(true, "2026-09-26T00:00:00.000Z"), id: "c_local_3" } as LhMessage;
+		const serverVerdict = { ...newVerdictMessage(true, "2026-09-26T00:00:01.000Z"), id: "srv_11" } as LhMessage;
+		assert.ok(absorbed(verdict, [serverVerdict]), "the verdict absorbs by accepted");
+	});
+
+	it("setMessages is the full snapshot swap (the idle settle sync)", () => {
+		const m = model({ initialMessages: [userText("u1", "old")] });
+		m.setMessages([userText("u1", "old"), userText("u2", "new"), assistantMsg("a1", "reply")]);
+		assert.deepEqual(
+			m.getState().nativeMessages.map((x) => (x as { id?: string }).id),
+			["u1", "u2", "a1"],
+		);
+	});
+});
+
+// ─── the continuation fold (mount/sync seam) ────────────────────────────────
+
+function assistantWithCall(id: string, callId: string, name = "research"): LhMessage {
+	return {
+		id,
+		role: "assistant",
+		content: [{ type: "toolCall", id: callId, name, arguments: { topic: "x" } }],
+		stopReason: "toolUse",
+		usage: { input: 0, output: 0, total: 0, cacheReads: 0, cacheWrites: 0 },
+		provider: "openrouter",
+		model: "test",
+		timestamp: 0,
+	} as unknown as LhMessage;
+}
+
+function resultMsg(callId: string, name = "research"): LhMessage {
+	return {
+		...newAnswerResult(callId, name, { ok: true }),
+		id: `r_${callId}`,
+	} as unknown as LhMessage;
+}
+
+describe("needsContinuation (the native lastAssistantMessageIsCompleteWithToolCalls)", () => {
+	it("a record ending on a fully-answered tool call continues", () => {
+		const messages = [userText("u1", "topic"), assistantWithCall("a1", "call-r1"), resultMsg("call-r1")];
+		assert.equal(needsContinuation(messages), true);
+	});
+
+	it("text-tailed records never continue", () => {
+		const messages = [userText("u1", "topic"), assistantMsg("a1", "done talking")];
+		assert.equal(needsContinuation(messages), false);
+	});
+
+	it("a dangling question never fires a turn (the dock owns it)", () => {
+		const messages = [userText("u1", "topic"), assistantWithCall("a1", "call-q1", "quiz")];
+		assert.equal(needsContinuation(messages), false);
+	});
+
+	it("parallel quizzes: only ALL-answered continues (the dock still owns the open one)", () => {
+		const messages = [
+			userText("u1", "topic"),
+			{
+				...assistantWithCall("a1", "call-a"),
+				content: [
+					{ type: "toolCall", id: "call-a", name: "quiz", arguments: {} },
+					{ type: "toolCall", id: "call-b", name: "quiz", arguments: {} },
+				],
+			} as unknown as LhMessage,
+			resultMsg("call-a", "quiz"),
+		];
+		assert.equal(needsContinuation(messages), false);
+		const both = [...messages, resultMsg("call-b", "quiz")];
+		assert.equal(needsContinuation(both), true, "once the dock answers B, the agent owes the reaction");
+	});
+
+	it("an orphan result (no asking assistant) never continues", () => {
+		assert.equal(needsContinuation([userText("u1", "topic"), resultMsg("call-x")]), false);
+	});
+});
+
+// ─── the pending-call tool scan ─────────────────────────────────────────────
+
+describe("pendingCallTool", () => {
+	it("finds the dangling call's tool name (the optimistic answer's stamp)", () => {
+		const messages = [assistantWithCall("a1", "call-q1", "ask_user_question")];
+		assert.equal(pendingCallTool(messages, "call-q1"), "ask_user_question");
+		assert.equal(pendingCallTool(messages, "call-zz"), null);
+	});
+});

tests/stream-activity.test.ts

@@ -7,11 +7,11 @@
  */
 import assert from "node:assert/strict";
 import { describe, it } from "node:test";
-import type { UIMessage } from "ai";
+import type { LhViewMessage } from "../src/core/view.ts";
 import { streamingToolActivity } from "../src/core/pending.ts";
 
-function msg(role: string, parts: unknown[]): UIMessage {
+function msg(role: string, parts: unknown[]): LhViewMessage {
-	return { id: "m1", role, parts } as unknown as UIMessage;
+	return { id: "m1", role, parts } as unknown as LhViewMessage;
 }
 
 function toolPart(tool: string, state: string) {