practice slice 3A — runSubmitAgent extracted to shared module; practice model roles
runSubmitAgent moved verbatim from extraction.ts to src/agent/subagents/ submit-agent.ts (exported; extraction imports it — zero behavior change, extraction suites green untouched). New optional prepareArguments seam: extraction wires its healStringifiedArrays heal; default is identity pass-through. ModelRole += partitioner | problemGenerator | grader; ROLE_SLOT pins all three to strong (quality-critical trio, plan default). Note: BACKGROUND_MODEL_ROLES (per-role settings overrides) deliberately NOT extended — settings-surface decision, flagged for the planning session.
This commit is contained in:
amosblomqvist
@@ -29,6 +29,7 @@ import { batchRuns, coverageOf, messageViewLines } from "../../core/coverage.ts"
|
||||
import { teacherProseBlocks } from "../../core/coverage.ts";
|
||||
import type { LhMessage } from "../../core/native-messages.ts";
|
||||
import { auditorGraphTools } from "../tools/graph.ts";
|
||||
import { runSubmitAgent } from "./submit-agent.ts";
|
||||
|
||||
/**
|
||||
* The extraction pipeline (vision §1 construction, 2026-09-02 redesign;
|
||||
@@ -344,75 +345,6 @@ field because it is empty.
|
||||
|
||||
Your entire deliverable is ONE submit_findings call.`;
|
||||
|
||||
/** One single-submit agent run — the shared shape of EVERY LLM stage
|
||||
* here (session + notes): the whole deliverable is ONE TypeBox-validated
|
||||
* submit call (a pi agent purely so the model can retry a schema-invalid
|
||||
* call). The submitted value is captured from the execute seam, which
|
||||
* only fires on a schema-valid call — the ??-fill (the old zod defaults)
|
||||
* applies at the capture. */
|
||||
async function runSubmitAgent<T>(opts: {
|
||||
role: ModelRole;
|
||||
toolName: string;
|
||||
toolDescription: string;
|
||||
parameters: TSchema;
|
||||
/** Weak-model tolerance: the old zod `.default()`s, applied to the
|
||||
* schema-valid raw input at the submit seam. */
|
||||
fill: (input: any) => T;
|
||||
systemPrompt: string;
|
||||
prompt: string;
|
||||
engine: PiEngine | undefined;
|
||||
signal: AbortSignal | undefined;
|
||||
onStep: (usage: Usage, model: string, step: number, finishReason: string, tools: string) => void;
|
||||
neverSubmitted: string;
|
||||
/** Extra tools beside the submit call (the auditor's graph-exploration
|
||||
* set, 2026-09-27): a small agentic loop instead of a bare retry. */
|
||||
extraTools?: Record<string, AgentTool<any>>;
|
||||
/** Turn budget for the loop (default 4 = retry-only; the auditor runs
|
||||
* an exploration loop instead — 16). */
|
||||
maxTurns?: number;
|
||||
}): Promise<T> {
|
||||
let submitted: T | undefined;
|
||||
const result = await runSubagent(
|
||||
[
|
||||
...(opts.extraTools ? Object.values(opts.extraTools) : []),
|
||||
{
|
||||
name: opts.toolName,
|
||||
label: opts.toolName,
|
||||
description: opts.toolDescription,
|
||||
parameters: opts.parameters,
|
||||
prepareArguments: (args) => healStringifiedArrays(args) as never,
|
||||
execute: async (_toolCallId, input) => {
|
||||
submitted = opts.fill(input);
|
||||
return toToolResult(submitted);
|
||||
},
|
||||
} satisfies AgentTool<any>,
|
||||
],
|
||||
{
|
||||
role: opts.role,
|
||||
engine: opts.engine,
|
||||
systemPrompt: opts.systemPrompt,
|
||||
prompt: opts.prompt,
|
||||
maxTurns: opts.maxTurns ?? 4,
|
||||
abortSignal: opts.signal,
|
||||
onAssistantMessage: (message, step) =>
|
||||
opts.onStep(
|
||||
message.usage,
|
||||
// Provider-qualified stamp (model slots wave, 2026-09-20):
|
||||
// pi reports the bare id; usage stamps carry provider:id to
|
||||
// match the price table's uniform keys.
|
||||
`${message.provider}:${message.model}`,
|
||||
step,
|
||||
message.stopReason,
|
||||
message.content.filter((c) => c.type === "toolCall").map((c) => (c as { name: string }).name).join(", ") || "(none)",
|
||||
),
|
||||
shouldStop: (_message, toolResults) => toolResults.some((r) => r.toolName === opts.toolName && !r.isError),
|
||||
},
|
||||
);
|
||||
void result;
|
||||
if (!submitted) throw new Error(opts.neverSubmitted);
|
||||
return submitted;
|
||||
}
|
||||
|
||||
/** The extractor over ONE session transcript (the chat front door). */
|
||||
function runExtractor(
|
||||
engine: PiEngine | undefined,
|
||||
@@ -432,6 +364,7 @@ function runExtractor(
|
||||
signal,
|
||||
onStep,
|
||||
neverSubmitted: "the extractor never submitted findings",
|
||||
prepareArguments: healStringifiedArrays,
|
||||
});
|
||||
}
|
||||
|
||||
@@ -455,6 +388,7 @@ function runNotesExtractor(
|
||||
signal,
|
||||
onStep,
|
||||
neverSubmitted: "the extractor never submitted findings",
|
||||
prepareArguments: healStringifiedArrays,
|
||||
});
|
||||
}
|
||||
|
||||
@@ -521,6 +455,7 @@ function runAuditor(
|
||||
neverSubmitted: "the auditor never submitted a review",
|
||||
extraTools: tools,
|
||||
maxTurns: 16,
|
||||
prepareArguments: healStringifiedArrays,
|
||||
});
|
||||
}
|
||||
|
||||
|
||||
@@ -0,0 +1,87 @@
|
||||
/**
|
||||
* runSubmitAgent — the shared single-submit agent seam (moved verbatim from
|
||||
* extraction.ts, practice slice 3, 2026-09-29): one TypeBox-validated submit
|
||||
* call is the whole deliverable (a pi agent purely so the model can retry a
|
||||
* schema-invalid call). The submitted value is captured from the execute
|
||||
* seam, which only fires on a schema-valid call — the ??-fill (the old zod
|
||||
* defaults) applies at the capture.
|
||||
*
|
||||
* Extraction passes its own weak-model argument heal (healStringifiedArrays)
|
||||
* through `prepareArguments`; the default is identity pass-through.
|
||||
*/
|
||||
|
||||
import type { Usage } from "@earendil-works/pi-ai";
|
||||
import type { AgentTool } from "@earendil-works/pi-agent-core";
|
||||
import type { TSchema } from "typebox";
|
||||
import { toToolResult } from "../tool-execute.ts";
|
||||
import type { ModelRole } from "../../../config.ts";
|
||||
import type { PiEngine } from "../models.ts";
|
||||
import { runSubagent } from "./pi-run.ts";
|
||||
|
||||
export async function runSubmitAgent<T>(opts: {
|
||||
role: ModelRole;
|
||||
toolName: string;
|
||||
toolDescription: string;
|
||||
parameters: TSchema;
|
||||
/** Weak-model tolerance: the old zod `.default()`s, applied to the
|
||||
* schema-valid raw input at the submit seam. */
|
||||
fill: (input: any) => T;
|
||||
systemPrompt: string;
|
||||
prompt: string;
|
||||
engine: PiEngine | undefined;
|
||||
signal: AbortSignal | undefined;
|
||||
onStep: (usage: Usage, model: string, step: number, finishReason: string, tools: string) => void;
|
||||
neverSubmitted: string;
|
||||
/** Extra tools beside the submit call (the auditor's graph-exploration
|
||||
* set, 2026-09-27): a small agentic loop instead of a bare retry. */
|
||||
extraTools?: Record<string, AgentTool<any>>;
|
||||
/** Turn budget for the loop (default 4 = retry-only; the auditor runs
|
||||
* an exploration loop instead — 16). */
|
||||
maxTurns?: number;
|
||||
/** Weak-model argument heal at pi's prepareArguments seam (runs BEFORE
|
||||
* TypeBox validation — the sanctioned shim point). Identity by
|
||||
* default; extraction wires healStringifiedArrays here. */
|
||||
prepareArguments?: (args: unknown) => unknown;
|
||||
}): Promise<T> {
|
||||
let submitted: T | undefined;
|
||||
const result = await runSubagent(
|
||||
[
|
||||
...(opts.extraTools ? Object.values(opts.extraTools) : []),
|
||||
{
|
||||
name: opts.toolName,
|
||||
label: opts.toolName,
|
||||
description: opts.toolDescription,
|
||||
parameters: opts.parameters,
|
||||
prepareArguments: (args) =>
|
||||
(opts.prepareArguments ? opts.prepareArguments(args) : args) as never,
|
||||
execute: async (_toolCallId, input) => {
|
||||
submitted = opts.fill(input);
|
||||
return toToolResult(submitted);
|
||||
},
|
||||
} satisfies AgentTool<any>,
|
||||
],
|
||||
{
|
||||
role: opts.role,
|
||||
engine: opts.engine,
|
||||
systemPrompt: opts.systemPrompt,
|
||||
prompt: opts.prompt,
|
||||
maxTurns: opts.maxTurns ?? 4,
|
||||
abortSignal: opts.signal,
|
||||
onAssistantMessage: (message, step) =>
|
||||
opts.onStep(
|
||||
message.usage,
|
||||
// Provider-qualified stamp (model slots wave, 2026-09-20):
|
||||
// pi reports the bare id; usage stamps carry provider:id to
|
||||
// match the price table's uniform keys.
|
||||
`${message.provider}:${message.model}`,
|
||||
step,
|
||||
message.stopReason,
|
||||
message.content.filter((c) => c.type === "toolCall").map((c) => (c as { name: string }).name).join(", ") || "(none)",
|
||||
),
|
||||
shouldStop: (_message, toolResults) => toolResults.some((r) => r.toolName === opts.toolName && !r.isError),
|
||||
},
|
||||
);
|
||||
void result;
|
||||
if (!submitted) throw new Error(opts.neverSubmitted);
|
||||
return submitted;
|
||||
}
|
||||
@@ -226,6 +226,12 @@ export const ROLE_SLOT: Readonly<Record<ModelRole, ModelSlotName>> = {
|
||||
auditor: "cheap",
|
||||
visualMaker: "vision",
|
||||
pdfReader: "vision",
|
||||
// The practice trio (2026-09-29): problem engineering is the product's
|
||||
// quality-critical component — strong tier for partitioner/generator/
|
||||
// grader alike (owner-approved plan default; one edit away from cheap).
|
||||
partitioner: "strong",
|
||||
problemGenerator: "strong",
|
||||
grader: "strong",
|
||||
};
|
||||
|
||||
/** General settings — app-level behavior. Version DISPLAY and plans login
|
||||
|
||||
Reference in New Issue
Block a user