amosblomqvist
5653c77a29
Diagnosis (2026-09-11 user testing, minimal-effort GLM): sessions stuck on 'plan' forever. Two causes: (1) the three-phase view read the stored tree's currentNodeId, which nothing ever writes post author-once (2026-09-03) — a structural dead seam; (2) the teacher stochastically left quiz.nodeId untagged, and done-by-quiz-result semantics would have falsely completed nodes anyway. The redesign (docs/probe-teach-split.md) abolishes derived occupancy in favor of explicit events: - Two agents over one transcript: createProberAgent (quiz without a nodeId field, ask, probe_done) and createTeacherAgent (full toolset, plan_init onward). The chat-handler construction branch keys on a probe_done part in the transcript. MODELS.prober added. - probe_done: server-side seam tool — carries the learner assessment (strands + free-form edge read), refuses while a question is open or the probe is already concluded, and queues a handoff steer through the generalized wake pipeline (steers now dedupe by id, not jobId; job finalization only for job steers). The teacher auto-wakes and authors the plan; the learner sees the assessment via the probe-done card. - phases: probe ends at a concluded probe_done, plan ends at a successful plan_init — strictly sequential, no skipping, no legacy compatibility (data wiped at the pivot per doctrine). - Plan map = position, not completion: derivePlanStatuses reads only tagged quiz CALLS (state/outcome feed nothing); current = latest tagged node with consecutive duplicates collapsed, visited = earlier distinct stays (a revisit stays in both — current wins at render). PlanTree is structure-only; currentNodeId/doneNodeIds and the dead mutation helpers are gone. No fallback current. - Prompts split per agent (prompts/prober/, prompts/teacher/ over the trimmed shared base); graph-context injection removed entirely (shelved, docs/probe-teach-split.md §5); knowledge-graph context no longer reaches either agent. - Plan blocks are full-only snapshots at plan_init (neighborhood machinery retired; the persistent DAG window is the future UI wave); usage capture unchanged. Gates: tsc clean, 294/294 tests (phases/plan-tree/plan-blocks/question- tools rewritten to the new semantics; probe-done + handoff-wake tests added), lint, build. Live GLM smoke PASS: prober concluded with an honest assessment, teacher auto-woke, authored a 10-node plan with the block attached, then tagged checks on slope-instinct → secant derived current/ visited exactly as designed; grading honest both ways (driver off-by-one proved a wrong answer grades false).
84 lines
3.4 KiB
TypeScript
84 lines
3.4 KiB
TypeScript
/**
|
|
* phases — the event-driven phase chain (docs/probe-teach-split.md §4).
|
|
*
|
|
* probe ends at the first CONCLUDED probe_done (output-available); planning
|
|
* ends at the first successful plan_init (a tree with nodes). No stored-tree
|
|
* seam: the phase is a pure fold over the transcript, so it can never stick
|
|
* (the old currentNodeId-based phaseOfTree read a field nothing ever set).
|
|
*/
|
|
import assert from "node:assert";
|
|
import test from "node:test";
|
|
import type { UIMessage } from "ai";
|
|
import { phaseAt3, phaseTimeline, type Phase3 } from "../src/core/phases.ts";
|
|
|
|
/** Text-only user message. */
|
|
function user(text: string): UIMessage {
|
|
return { id: `u_${text}`, role: "user", parts: [{ type: "text", text }] };
|
|
}
|
|
|
|
/** An assistant message carrying a tool part in the given state. */
|
|
function assistantTool(
|
|
name: string,
|
|
state: "input-available" | "output-available" | "output-error",
|
|
output?: unknown,
|
|
): UIMessage {
|
|
return {
|
|
id: `a_${name}_${state}_${Math.random().toString(36).slice(2, 6)}`,
|
|
role: "assistant",
|
|
parts: [{ type: `tool-${name}`, state, ...(output !== undefined ? { output } : {}) } as never],
|
|
};
|
|
}
|
|
|
|
const DONE_ASSESSMENT = { assessment: { strands: [{ strand: "s", assessment: "a" }] } };
|
|
const PLAN_OK = { ok: true, tree: { nodes: [{ id: "n1", title: "N" }], edges: [] } };
|
|
|
|
test("phaseAt3: probe until a concluded probe_done; plan until a successful plan_init; teach after", () => {
|
|
const probe = phaseAt3([user("teach me calculus"), assistantTool("quiz", "output-available")]);
|
|
assert.equal(probe, "probe", "quiz answers alone never end the probe");
|
|
|
|
const concluded = phaseAt3([
|
|
user("teach me"),
|
|
assistantTool("probe_done", "output-available", DONE_ASSESSMENT),
|
|
]);
|
|
assert.equal(concluded, "plan", "a concluded probe_done ends the probe (planning begins)");
|
|
|
|
const plan = phaseAt3([
|
|
user("teach me"),
|
|
assistantTool("probe_done", "output-available", DONE_ASSESSMENT),
|
|
assistantTool("plan_init", "output-available", PLAN_OK),
|
|
]);
|
|
assert.equal(plan, "teach", "a successful plan_init ends planning");
|
|
|
|
const failedPlan = phaseAt3([
|
|
user("teach me"),
|
|
assistantTool("probe_done", "output-available", DONE_ASSESSMENT),
|
|
assistantTool("plan_init", "output-error", { ok: false, error: "bad mermaid" }),
|
|
]);
|
|
assert.equal(failedPlan, "plan", "a failed plan_init does not end planning");
|
|
});
|
|
|
|
test("phaseAt3: a mere probe_done CALL (input-available, not yet concluded) keeps the probe open", () => {
|
|
const still = phaseAt3([user("teach me"), assistantTool("probe_done", "input-available")]);
|
|
assert.equal(still, "probe");
|
|
});
|
|
|
|
test("phaseAt3: legacy transcripts without probe_done read as probe (no backwards compat — data wiped)", () => {
|
|
const legacy = phaseAt3([user("teach me lacan"), assistantTool("plan_init", "output-available", PLAN_OK)]);
|
|
assert.equal(legacy, "probe", "plan_init alone never flips a phase post-split");
|
|
});
|
|
|
|
test("phaseTimeline: folds messages into phase spans for the dividers", () => {
|
|
const timeline = phaseTimeline([
|
|
user("teach me"),
|
|
assistantTool("quiz", "output-available"),
|
|
assistantTool("probe_done", "output-available", DONE_ASSESSMENT),
|
|
assistantTool("plan_init", "output-available", PLAN_OK),
|
|
user("go on"),
|
|
]);
|
|
assert.deepEqual(timeline satisfies Phase3[], ["probe", "probe", "plan", "teach", "teach"]);
|
|
assert.equal(timeline.length, 5);
|
|
});
|
|
|
|
test("phaseAt3: empty transcript is probe", () => {
|
|
assert.equal(phaseAt3([]), "probe");
|
|
}); |