Two user-reported problems from the muscle-protein-synthesis session,
both diagnosed from the live data (chat_mtxeml9e_6118e59f).
The freeze: after probe_done, the open session never showed the
teacher's wake turn — only navigating away and back loaded it. Not an
agent-switch problem: the server-side handoff worked perfectly
(write-ahead persist, liveness registry, the sidebar's busy dot saw the
wake the whole time). The gap was client attach: the only mid-session
wake detector the open page has is the jobs poll, and probe-phase
sessions have zero job traffic, so that poll is OFF; the post-turn
adoption poll read turnInFlight=false at ~0ms (the wake fires at
~0.75-1.2s, after the pump's grace), adopted, and stopped. Fix: the
post-turn adoption effect now keeps a short WATCH WINDOW after adopting
(~4s, 8x500ms) — if turnInFlight flips true, it hands off to the
existing wake-adoption poll (busy shimmer, adopt at finish).
The mermaid failures: four failed plan_init calls, all caused by ONE
hidden rule — underscores are forbidden in node ids (mps_translation,
net_balance, per_meal). The id regex rejected '_' BEFORE the slug check,
so the specific, actionable message never fired; the generic 'can't read
the edge — expected a[title] --> b[title]' lied (the syntax was exactly
that), and the teacher spent three attempts hunting punctuation it
suspected (colons, commas, multi-word labels) — visible in its own
reasoning trail. Per the user's principle — enforce only the slugs,
free everything else: underscores are first-class slug chars; the id
CANDIDATE class is broad so every bad id reaches the specific slug
message; chained edges (a --> b --> c) expand to pairs; ==>, -.-> are
edge synonyms; any direction header accepted (we layout ourselves);
id["title"] quote pairs are stripped. The tool description, zod
describe, and the teacher's planning prompt now state the one strict
rule instead of the old subset liturgy.
Verified: tsc, 305/305 (the exact muscle-protein mermaid is now a
passing test), lint, build; real-browser puppeteer smoke on live GLM —
the teacher's wake turn appeared in the open page ~50s after the probe
concluded with ZERO navigation, verdict gate + first teach check +
plan window all live in-page (the single console 404 is the app's
missing favicon, environmental noise).
The accept-plan UI lands (docs/probe-teach-split.md §4, formerly the
interim rule): presenting the plan no longer moves the phase — the
learner's ACCEPT does. While the presented plan awaits its verdict, the
composer's send square becomes two stacked buttons (green Accept plan /
red Decline); typed text rides the verdict as his note, Enter keeps
sending plain side questions (phase unmoved), decline asks for a
revision and the buttons stay.
The seam: accept sends a user message carrying a data-plan-accepted
part (PlanAcceptedData in contracts/chat.ts, TeacherUIDataParts
'plan-accepted'); phaseAt3 reads it as the plan→teach event (plan_init
demoted to presentation-only); convertToModelMessages drops data parts,
so the teacher sees only the text. Mid-lesson rewrites need no fresh
verdict — acceptance happens once, at the first presentation.
The teacher's planning prompt now ends the turn at presentation and
waits for the verdict (never teaches into an unaccepted plan).
Gates: tsc clean, 304/304 (phases tests rewritten to the acceptance
seam incl. decline/side-question/failed-init cases; composer verdict
render tests), lint, build. Live GLM smoke ALL PASS: phase pinned at
'plan' after plan_init, presentation turn carried NO check, accept
click flipped the summary to 'teach', then tagged checks moved the map
(current=rates, visited=composite); zero console errors.
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).
Wave 4 (agent-side observability + control), reconciled with the dock
contract (master relay 2026-09):
- abortJob in src/server/jobs.ts, exported through the server barrel for
POST /api/activity/abort (master's thin route): fires the job's
AbortController (real kill — the in-flight model call aborts, not just
the next loop check), drops any queued result steer, flips the record
to error "Aborted by learner" + activity "aborted" synchronously, and
queues a one-line abort steer (buildAbortSteer) through the same wake
pipeline so the teacher's promise doesn't hang in the lesson. Shares
the result steer's deterministic id space, so abort + racing result
can never both deliver (steerDelivered dedupe).
- Abort registry (src/agent/subagents/aborts.ts) on globalThis, mirroring
the wake pipeline's module-state pattern; runner registers/unregisters
per run, so the route's module instance sees the loop's controller.
- Runner usage capture: LoopContext (abortSignal + onUsage) threads into
all three kind loops; per-step usage folds via accumulateJobUsage into
JobRecord.usage (model from config MODELS, absent-never-zero), flushed
in runJob's finally so an aborted job's money record survives the kill.
- Contracts: JobUsage token fields optional + lastStep (most recent
call's accounting, for the dock's per-blob context line vs
CONTEXT_LIMITS); SubagentResultMeta.aborted; AbortSteerMessage;
buildAbortSteer; SteerMessage union generalized.
- demo-builder/visual-maker accept abortSignal + onUsage (visual folds
generateText usage per iteration).
- stillRunning/finishJob/jobContext guards: an aborted job is never
overwritten by the runner (terminal no matter what the wake does);
failJob and the pump's finalizeJob keep the flip.
- UI: abort steer renders a quiet aborted chip (job-aborted-badge).
- Prompt 20-jobs.md: one line about aborted jobs.
Tests: accumulateJobUsage folds (undefined-preservation, lastStep),
runner usage capture, real-kill abort (sync flip, silent unwind,
abort steer wakes), usage-so-far survives the kill, aborting a
delivering job drops the queued result steer, outcome reasons,
abort-chip render. 330/330, typecheck/lint/build green.
Wave 3: the plan protocol collapses to one re-callable mermaid plan_init
(seven mutation tools deleted by design; statuses derive from quiz checks;
live quiz-nodeId validation with mid-turn retry). The steer-starvation arc:
steers deliver coalesced with the learner's answer (claimSteers), persist
dedupes same-id continuation copies, jobs settle-sync defers while a turn
is busy. Research steers gain the collapsed full-brief affordance. Model
reasoning effort set to minimal across roles.
- plan_init is now the ONLY plan tool: a mermaid 'graph TD' block parsed by a
constrained-subset parser (mermaid-plan.ts) with line-precise errors; the
seven mutation/tracking tools are gone by design (2026-09-03 design pass).
- Re-call plan_init = sanctioned wholesale rewrite; slugs are permanent.
- derivePlanStatuses (contracts/plan.ts): done = passed check tagged with a
quiz nodeId; current = open check's node else first topo-ready node — the
tracking the teacher no longer declares by hand.
- questionToolsFor(ctx) (tools.ts): the quiz schema chains a dynamic plan-
nodeId validation against the LIVE record — an invalid nodeId is rejected
at the call boundary and the teacher retries mid-turn, before any UI.
Static questionTools stay validator-less for the default path.
- Prompts rewritten for mermaid authoring + derived position.
- Out-of-lane surgical touches (flagged for master review):
src/core/question-core.ts (quizInputObject export + nodeId description),
src/core/contracts/plan.ts (derivePlanStatuses + doc).
Seam from the state-class strip: renderGraphContext tags nodes with raw
latest-quiz outcomes ([miss] / [dont-know] / [correct] / [never probed]);
50-graph-context.md still taught the deleted state classes ([missed] /
[known] / [candidate]). Re-aligned, guidance unchanged.
The minimal implementation of the 2026-09-02 graph vision (docs/vision.md §1):
- contracts/graph.ts: every node is a TRUE proposition (statement is both
rendering label and identity key; required description; no kind, no name).
Edges are follows (derivable) / supports (grounds without entailing) with
message-granular provenance points {chatId, messageId}. Events reduce to
correct/miss/dont-know plus observation (understanding notes); no taught
events, no misconception kind.
- extraction.ts: extractor is TREE-BLIND (transcript only, no plan tree) and
free-form; auditor sees findings + the current graph, owns faithfulness and
identity (link-or-create verdict per proposition, recorded as linkTo);
transcriptView tags every line with its message id (provenance anchors).
- core/graph.ts merger: applies verdicts deterministically — linkTo appends
a link line (phrasings join aliases, provenance unions; reads merge by id),
absent creates; NO string matching anywhere. Quiz events rebuilt from tool
parts with attribution-only nodeId resolution; understanding notes become
observation events (obs:<sourceId>); mapping lines are the verdict audit
trail and make re-merges idempotent.
- deriveNodeStates: transitional — latest quiz event wins (known/missed/
dont-know, else candidate); taught unreachable until the state-model
ideation lands.
- graph-context.ts + 50-graph-context.md: follows chains, hotspot grouping,
state vocabulary updated for the new event kinds.
- graph-view.tsx: node label is the statement; detail shows statement +
description.
- tests: merger verdicts (link, no-string-match, cross-chat assertions),
tree-blind pipeline end-to-end, follows vocabulary in context tests.
276/276 green.
composeInstructions loaded EVERY .md in src/agent/prompts — including
README.md, the directory's human documentation, which rode into the
system prompt as its own section (visible in OpenRouter logs, old POC
paths and all).
The loader now enforces the documented convention: only numeric-prefix
fragments (\d+-*.md) compose. README.md stays in the directory as docs
and can never reach the model again. README's rule text updated to
match; composeInstructions exported for tests; regression test asserts
the README never leaks and every numeric fragment composes verbatim.
The job tools (research/demo/visual) switch from client-side turn-breaking
calls to pi's model: execute resolves IMMEDIATELY with a fire-and-forget
ack, the teacher's turn continues, and the result arrives later as an
injected user-role steer message that wakes the teacher in a new
server-side turn.
- contracts.ts: THE ASYNC PROTOCOL rewritten (ack now, steer later); steer
types (SubagentSteerMessage, metadata.subagentResult + data-subagent-
artifact part), pure buildSubagentSteer/steerDelivered/isSubagentFailure
- tools/jobs.ts: jobTools(ctx) factory — server-side execute creates the
job (deduped by toolCallId), fires spawn with the dispatch-time input
(the part isn't persisted yet at execute — a transcript scan would race)
- runner.ts: runJob(jobId, input?) builds the steer and hands it to the
injected deliver seam; pipeline owns the done/error flip (a polling
client re-syncs once, at the end); failures still steer
- server/jobs.ts (new pipeline): per-chat serialized queues on globalThis,
idle gates (turn liveness + pending question, 1.5s grace), 409 retry
with backoff, idempotent per jobId, stale-job recovery sweep (re-deliver
or re-spawn); wake turns run through the same handleChatRequest
composition via setTurnRunner injection
- chat-handler.ts: wires runner+deliver+tools ctx; drops spawnTurnJobs
from onFinish; teacher default registry = question tools only
- plan.ts: PlanToolContext split out (plan tools need no job wiring)
- visual-maker: edit_svg added (pi parity — exact-match single
replacement, fuzzy context on a miss); researcher: pi's kept/dropped
source accounting restored
- prompts 20-jobs/40-demos: async discipline rewritten (announce and keep
teaching, no waiting/polling/fabricating, parallel dispatch, teach with
the artifact when it arrives)
- tests: jobs.test.ts rewritten (dispatch ack mid-turn, runner steers,
wake pipeline: deferral/409/idempotency/failure), demos.test.tsx runner
tests ported to runJob + fake deliver; 222 green
FLAGGED surgical touches outside the stream's owned files, for master
review at merge: src/core/contracts.ts (protocol doc + steer types),
src/core/ui-message.ts (job tool + data-part types), src/server/
chat-handler.ts (wiring), src/server/turn-service.ts (spawnTurnJobs
retirement), src/agent/teacher-agent.ts (default tools), src/agent/tools/
plan.ts (context type split).
Verified live on PORT=3003 with real models: teacher dispatched research
mid-turn, researcher ran real searches/fetches, steer injected on quiz
answer, wake turn wove the brief in citing its sources, job flipped done;
a client/wake race correctly 409'd.
- src/store/jobs.ts: per-job JSON store in data/jobs (create/get/update/list,
dedup by toolCallId)
- src/agent/tools/jobs.ts: research tool — client-side (no execute), turn ends
at the call; input per contracts ResearchToolInput
- src/agent/tools/web.ts: server-side web_search (Google CSE) + web_fetch
(Readability/Turndown/linkedom, Jina fallback) — errors return as strings,
never thrown into the agent loop. Deps: @mozilla/readability, linkedom,
turndown (+@types/turndown dev)
- src/agent/subagents/researcher.ts: isolated ToolLoopAgent (web tools only,
12-step cap), researcher prompt ported from pi
- src/agent/subagents/runner.ts: spawnPendingJobs hook + runResearchJob —
fire-and-forget, activity line per tool call, fulfills the part in the
LATEST chat record (output-available) before flipping job status; failure
still fulfills the part so the teacher learns the outcome
- src/server/chat-handler.ts: onFinish spawn hook after saveMessages
- app/api/jobs/route.ts: GET ?chatId= polling endpoint
- components/artifacts/job-card.tsx + registry entry: polling activity card
(2s) → syncFromServer on finish; research-brief card (markdown + sources)
- src/agent/prompts/20-jobs.md: research doctrine for the teacher
- tests/jobs.test.ts: store CRUD, spawn hook via scripted mock teacher,
runner completion/failure with fake model + stubbed fetch
- src/core/plan-tree.ts: pure tree ops (createTree/addNode/updateNode/removeNode/
addEdge/removeEdge/setCurrent/markDone), DAG invariants (unique ids, existing
endpoints, no self-loops/duplicates/cycles), cascade remove, version bumps on
state changes only (true no-ops for idempotent setCurrent/markDone).
- src/agent/tools/plan.ts: plan_init (refuses re-init), plan_add_node (optional
parents → requires edges), plan_update_node, plan_remove_node, plan_add_edge,
plan_remove_edge, plan_set_current, plan_mark_done. Execute = getChat → op →
saveTree → PlanToolResult with the full updated tree; PlanTreeError and missing
chats become {ok:false, error, tree}.
- src/agent/prompts/10-plan-tree.md: tree-authoring doctrine (when: post-probe;
grain: one reasoning step per node; discipline: set_current on start,
mark_done on demonstrated mastery; freedom: mid-session restructure).
- tests/plan-tree.test.ts: every validation error path, version semantics,
cascade remove, no-mutation, tool round-trips against the real chat store.
amosblomqvist