Root cause of the live-test bugs (2026-09-19): pi passes every run's
AbortSignal to tool execute, but toPiTool dropped it — agent.abort()
only landed at the next step boundary, so a long local grep held its
turn hostage until the dev server's memory restart killed the run
mid-tool, freezing the record at its write-ahead state (the 'went
back in time' rollback).
- lh-tool: execute ctx gains signal (parent + timeout combined); every
call races a 30s system-level timeout with a 250ms cooperative grace —
abort kills a stuck tool but lets a poll-the-signal tool deliver its
partial results. Pre-aborted signals never fire listeners: checked
upfront. toolTimeoutMs is the test seam (pdfModelOverride idiom).
- fs-shared: walkFiles returns {timedOut, aborted} instead of throwing —
abort yields partial results.
- grep: the scan phase was unbounded (the walk's 10s deadline only
bounded collection) — same budget now checked per file; find/grep
check the signal and return partials with a notice.
- pending/chat: streamingToolActivity names the ACTUAL running tool —
a part sits input-streaming through its whole execution, so the old
shimmer called minutes-long greps 'Writing a question…'.
Tests: toPiTool guard (forward, abort, timeout, grace, question
bypass, ToolInputError passthrough), walk mid-abort, find/grep
pre-abort + mid-run abort, activity labels. Gates: 617/617.
Learner source files, on the teacher's own terms (nothing auto-injects):
- read tool (src/agent/tools/read.ts): pi-read port — path/offset/limit,
2000-line/50KB truncation with use-offset continuation notices, image
files (jpg/png/gif/webp/bmp) returned as native image content parts
(new LhToolImageResult escape hatch in lh-tool.ts). Always-on in both
the prober and teacher registries (user-decided 2026-09).
- pdf-reader sub-agent (JobKind 'pdf', toggleable, ack-and-steer): triage
via pdf_info (pdfjs text layer: per-page text/image/math-density),
targeted pdf_extract, regex pdf_search (hits count as consultation),
and pdf_render — pages painted by pdfjs onto a canvas shim in headless
Chrome (the pdfium viewer doesn't paint headless), screenshot fed back
image-in-the-loop (visual-maker idiom, providerSafePng applied). The
unread-finish gate bounces a zero-consultation brief once. Teacher gets
a text brief with page citations only; renders never leave the loop.
- MODELS.pdfReader = qwen/qwen3.8-flash (user override 2026-09-19: the
look-at-renders workload wants the vision-proven model; catalog listed
glm-5.3-flash as image-capable but qwen is the known-good daily driver).
- UI: PdfCompleteBadge + PdfBriefCard (steer-card, job-card), dispatch
card + activity-dock kind entries (type-forced, flagged for UI review).
- Prompts: 'Learner source files' section in teacher/30-jobs.md; read
line in prober/10-probe-doctrine.md. Doctrine rides tool descriptions
per the self-contained-tools principle.
Part 1 — tool availability (the settings seam, machinery generic):
- core/pending.ts: TOGGLEABLE_TOOLS (today: research, demo, visual —
the model-facing tool names), EnabledTools = the JSON-native seam
shape Partial<Record<string, boolean>> (absent key = enabled, so old
records behave exactly as today), filterEnabledTools shared helper.
quiz/ask_user_question are always-on FOR NOW (user-decided); the
filtering is generic so future toggleables (question-tool variants,
the read tool, …) need no factory change.
- createProberTools/createTeacherTools take enabled?: EnabledTools; the
composition root (chat-handler, Platform's lane this wave) passes the
resolved settings through. probe_done is re-added after the filter —
always-on by construction (it ends the phase; a prober that can never
conclude is a lockout).
- Prompts stay STATIC under toggles this wave (user-decided); standing
principle recorded: tools ship self-contained usage guidance in their
descriptions; system prompts reference only always-on tools.
Part 2 — phase-based question-writing principles (user-locked wording,
drafted with the user over three iterations):
- NEW teacher/21-question-principles.md (composes after 20-teaching):
answerable-by-reasoning-from-the-session (the fairness rule that
filters unanswerable questions without forbidding visible answers),
never-just-reciting-your-words-back, PREFER problems that apply the
knowledge (preference, not hard rule — honest recall is the fallback,
not the habit), catch-confusion-early-and-nameable (wrong options as
real misconceptions), answer-is-signal-not-verdict.
- prober/10-probe-doctrine.md: 'no rush' retired; quick-measurement
directive in — every question moves an edge (bracket, narrow, or
characterize); a bracketed strand is done.
Tests: TOGGLEABLE_TOOLS pin, prober/teacher filtering (absent = full,
false drops exactly that tool, probe_done survives), prompt pins (the
principles compose verbatim into the teacher prompt, the no-rush
framing is gone from the prober). 547/547, typecheck/lint/build green.
Live observation (modal-realism run): the teacher batched two quizzes in
one turn; the dock showed the SECOND question first, so the answering
order was reversed vs the saved record and the transcript cards.
Cause: findPending scanned parts BACKWARD within a message, so with two
input-available quizzes it picked the last one first. No overwriting was
involved — parts are created in call order on both wire and record; the
scan direction alone reversed the interaction order. (pi executes a
parallel batch with starts + result messages in call order; the
execution-end order is completion-racy, which also allowed a millisecond
dock-subject flip mid-turn.)
Fix A — src/core/pending.ts: parts now scan FORWARD within a message
(messages stay newest-first): the FIRST unanswered question is the dock
subject; the second appears only after the first is answered.
Fix B — src/agent/lh-tool.ts: question tools opt into sequential
execution (pi runs the batch in call order when any called tool opts
in), so tool-input-available chunks always arrive in call order and the
dock subject can never flip mid-turn. Question executes are instant.
Pinned in chat-flow.test.ts: a parallel two-quiz batch asserts wire
order (A's available chunk before B's), part persistence order, and the
dock selection sequence (A first, then B after A is answered).
Gates: tsc, lint, build, 307/307.
- components/chat-transport.ts (NEW): createChatTransport(chatId) — the
chat.tsx seam is now a one-line swap. Its fetch turns a 409 on POST
/api/chat (a learner send racing the wake pipeline's ~1s adoption window)
into: wait out the in-flight turn (poll turnInFlight, 30s cap), refetch
the record, MERGE the outgoing messages onto the fresh server list, re-POST.
A bare retry would clobber the steer + wake messages via the write-ahead
persist (steerDelivered makes that loss permanent) — merging is the whole
point. Unit-tested: pass-through, wait+merge+reorder, dedupe by id,
repeated-409 loop, bounded give-up.
- src/core/pending.ts: findPendingJobCall deleted (dead since wave 1 —
job tools resolve server-side with an ack, nothing pends client-side).
JOB_TOOLS kept (runner.ts still uses it). Header doc rewritten.
- src/server/jobs.ts: GRACE_MS 1500 → 750, with the evidence in the
comment: the race window is only the learner's answer→POST flight
(sendAutomaticallyWhen dispatches synchronously; ~10ms localhost,
<300ms realistic RTT — 750 is 2.5× worst case) and misses self-heal
via the pump's gate recheck, the 409 backoff ramp, and the new
transport merge-retry. POLL_MS=400 kept (sampling granularity only).
- tests/jobs.test.ts: prompt-size guard — every fragment ≤8k chars,
composed prompt ≤60k — the README-leak class (815 bytes ≈ 200 tokens
per teacher turn before wave 1's fix) can't silently recur.
Gates: typecheck, 279/279 tests, lint, build — all green.
Sidebar (replaces the modal drawer):
- Inline, toggleable, no backdrop/focus theft — can stay open while working
- Open state persists across pages (provider in root layout) and reloads
(localStorage; starts closed server-side to avoid hydration mismatch,
stored value applied post-mount; default open on first run)
- Hamburger in page headers only when closed; close button in sidebar header
Floating composer blob (per reference screenshot):
- Floats over the transcript (absolute overlay), persists at bottom while
scrolling, content fades behind it (gradient mask); pb-44 scroll clearance
- Concentric corner radii: blob rounded-2xl(16px) − p-1.5(6px) = send
rounded-[10px] = 25% of the 40px button; CornerDownLeft icon per reference
- Stop moved from header into the blob (busy → square icon)
- prompt-input.tsx + 7 orphan ui files pruned (attachments/command palette
unused); hand-rolled Composer is ~100 lines and render-tested
Pending-question morph (one text field on screen, never two):
- Quiz/ask-select pending → blob becomes the note field (send hidden, hint
line above); card's internal note input removed; Submit consumes it
- Ask free-text pending → blob becomes the answer field; its send submits
via the card's registered submit api (ref bridge); Enter works
- Chat owns the single text state; card clears it on submit
- canSend for answer mode mirrors the card predicate directly — reading the
ref during render lagged one commit (found in browser testing)
Widths: one shared column (max-w-3xl + px-4) for transcript, cards, and
composer — all three measured pixel-identical (left 416, width 736).
Assistant Message max-w-full/MessageContent w-full (was max-w-[95%]/w-fit —
the 'cards are a bit shorter' bug).
User messages render plain text again — no math pipeline on learner-authored
text (currency hazard accepted for the model only, guarded by the prompt).
Tests: 58 pass (5 new composer pins; card tests updated to controlled
props). Verified live: note flow, answer flow, sidebar persistence, float,
fade, zero console errors — plus the live model spontaneously honoring the
$ currency escape contract in a continuation.
The teaching agent no longer runs on pi. Stack:
- ToolLoopAgent (ai@7) + OpenRouter (OPENROUTER_API_KEY), model via LH_MODEL,
default anthropic/claude-sonnet-4.5 with reasoning tokens
- quiz/ask_user_question = client-side tools (no server execute): the call
lands as a typed tool part in input-available state; the card grades via
src/core/question-core.ts; addToolOutput sends the verdict; the teacher's
reaction continues the same assistant message
- resume-mid-quiz is free: the pending tool part persists in the transcript
- persistence: one JSON per chat (data/chats/), full UIMessage transcripts,
saved onFinish (also on abort)
- UI: AI Elements (owned copy-paste components, pruned to the five in use)
+ button-based QuizCard/AnsweredCard (ask answers never graded)
- pi fully removed: TUI, extensions, Fastify+WS server, bridge, custom
protocol all deleted (recoverable from git history)
- tests: question-core units carry over; chat-flow integration drives
handleChatRequest with MockLanguageModelV4 (48 tests, zero tokens);
render smoke tests guard the cards (renderToStaticMarkup)
Carried over unchanged: teacher system prompt, question-core, schema/
ledger/derive/extract pipeline, data model.
amosblomqvist