per-session models: every session owns its model + thinking level from birth

The reported bug (2026-09-28): every chat ran glm-5.3-flash regardless of
any pick, while composer and settings UIs agreed with each other and lied
about the run. Root cause: a fossil default-settings record (settings.json
chat section, frozen pre-slots-wave) was stamped verbatim into every new
chat at birth by the new-session flow, and resolveConversationSpec prefers
the per-chat record over the strong slot forever.

Owner redesign — models are per session, always; the strong slot is the
new-session default:

- POST /api/chats stamps the newborn verbatim from the strong slot
  (model + thinking level); no client settings PUT on create anymore
- the composer control is dual-bound: in-session it reads/writes the
  session's own record (GET /api/chats/:id, PUT .../settings); on the
  draft surfaces it binds the strong slot (picking there = changing the
  default)
- thinking levels follow the identical mechanics (confirmed by owner)
- the zombie default-settings system dies whole: useDefaultSettings,
  /api/settings/defaults, src/store/default-settings.ts, the chat
  section of UserSettings (sanitizer strips fossil copies), the legacy
  default-settings.json fold, and DEFAULT_SETTINGS_FILE
- catalogDefaultsFor collapses to the strong slot alone
- thinking display for a legacy unstamped session resolves exactly like
  the engine (session ?? strong ?? model ladder)

Tests: new chat-birth-stamp suite pins the stamp, the unconfigured boot,
the unset level, and the copy-not-live-view regression; defaults pins
rewritten strong-only; legacy-migration suite deleted.
This commit is contained in:
amosblomqvist
2026-09-28 23:30:11 +03:00
parent 1f5e21a674
commit 3e9a85d60f
27 changed files with 336 additions and 491 deletions
+4
View File
@@ -11,6 +11,10 @@ import type { ChatSettings } from "@/src/core";
/**
* The per-chat model control (model-control wave): replace-whole settings —
* the picker sends the complete desired object ({} clears all overrides).
* PER-SESSION MODELS (owner, 2026-09-28): this route is THE in-session
* writer — the composer's control binds the session's own record; the
* strong slot is only the birth default and never reaches an existing
* session again.
* Validation is the catalog gate (src/server/model-catalog.ts): unknown
* provider/model is a 400. Thinking levels never reject (owner,
* interlude 2026-09-22): a pinned level clamps to the named model's
+15 -1
View File
@@ -1,4 +1,4 @@
import { createChat, isTurnRunning, listChats, titleFromPrompt } from "@/src/server";
import { createChat, isTurnRunning, listChats, readSettingsRecord, stores, titleFromPrompt } from "@/src/server";
import type { ChatListItem } from "@/src/core";
/** The session picker's list: record-derived summaries (phase, pending
@@ -20,5 +20,19 @@ export async function POST(req: Request) {
const body = (await req.json().catch(() => ({}))) as { title?: unknown };
const title = typeof body.title === "string" ? body.title : "";
const chat = createChat(titleFromPrompt(title || "Untitled session"));
// PER-SESSION MODELS (owner, 2026-09-28): a session owns its model +
// thinking level from birth — stamped verbatim from the STRONG slot,
// the new-session default. From here the strong slot never reaches
// this session again: in-session picks write the session's own record
// (PUT /api/chats/:id/settings). No strong slot configured (fresh BYOK
// boot) → no stamp; the turn gate's 503 models_not_configured envelope
// refuses honestly until the setup flow fills the slot.
const strong = readSettingsRecord()?.models?.strong;
if (strong?.model) {
stores.chatStore.saveSettings(chat.id, {
model: { provider: strong.model.provider, id: strong.model.id },
thinkingLevel: strong.thinkingLevel,
});
}
return Response.json({ id: chat.id, title: chat.title });
}
-53
View File
@@ -1,53 +0,0 @@
/**
* GET/PUT /api/settings/defaults — server persistence for the DEFAULT
* settings: THE default model + thinking level + knowledge-context
* budget. The new-session surfaces' controls (the home graph view's
* topic dock and /chat/new) ARE the settings page for the default model
* (owner, 2026-09-18): there is no baked-in default — BYOK first boot
* renders "no model" until the user picks one, and the last choice IS
* the default from then on. Both surfaces share ONE record; the file
* lives at data/default-settings.json (config.DEFAULT_SETTINGS_FILE).
*
* GET → 200 {settings} once saved, 200 null on first boot ("no model").
* PUT → 200 {settings} — the VALIDATED copy, level-clamped, persisted.
* The exact per-chat contract (parseSettingsBody + the catalog
* gate): an unknown provider/model is a 400. Thinking levels
* never reject (owner, interlude 2026-09-22): a pinned level
* clamps to the named model's nearest supported and the clamped
* value is stored — model swaps just work.
*
* Per-session settings (PUT /api/chats/:id/settings) are a separate record
* untouched by this — the default still rides along on session creation
* exactly as before.
*/
import {
clampChatLevel,
effectiveSettings,
parseSettingsBody,
readDefaultSettings,
validateChatSettings,
writeDefaultSettings,
} from "@/src/server";
export const dynamic = "force-dynamic";
export async function GET() {
return Response.json(readDefaultSettings());
}
export async function PUT(request: Request) {
const body = await request.json().catch(() => null);
const parsed = parseSettingsBody(body);
if (!parsed.ok) return Response.json({ error: parsed.error }, { status: 400 });
const verdict = validateChatSettings(parsed.settings);
if (!verdict.ok) {
return Response.json({ error: verdict.error, supported: verdict.supported }, { status: 400 });
}
// Clamp-on-write (owner, interlude 2026-09-22): a pinned thinking
// level never rejects the write — it clamps to the named model's
// nearest supported level (body model, else the record's strong
// slot) and the clamped value is stored.
const settings = clampChatLevel(parsed.settings, effectiveSettings().models.strong);
writeDefaultSettings(settings);
return Response.json({ settings });
}