Two round-1 follow-ups (owner):
- The overlay painted over the scroller's scrollbar track. The scroller
reserves symmetric gutters (scrollbar-gutter: stable both-edges), so the
overlay's right edge is now inset by the measured gutter — half of
offsetWidth − clientWidth; zero on macOS overlay scrollbars, the real
width on classic platforms. Re-measured on window resize. The headless
shell renders classic scrollbars (11px), so the path is live-verified.
- Wheel over the dock did nothing: the overlay is a SIBLING of the
scroller, so native scroll chaining stops at the overlay (the old
in-flow dock chained for free — a real regression, not a preference).
A non-passive wheel listener on the dock block forwards the delta to
the transcript ONLY when nothing under the cursor can take it: the
question area, textarea overflow, and the model menus keep their
native scrolling, including their own edge chaining; an unhandled
wheel (over a fully-visible box, or at an internal scroller's edge)
scrolls the transcript — so wheel-up over the quiz itself now folds
the box under the cursor.
Verified live: wheel-up over the question box scrolled the transcript
(9460→9100) and folded to the bar; wheel-down handled cleanly; the
overlay inset matches the measured gutter exactly.
The question box folded far too late because the old dock was IN FLOW
inside the scroller: folding it shrank the scroll content, so the fold
had to wait until the learner was a full dock-height past the bottom —
any closer and the browser's scrollTop clamp pinned them back to the
bottom, re-expanding the box (the oscillation the old hysteresis
comments worked around).
The dock now overlays the column bottom (absolute — a sibling of the
scroller, the use-stick-to-bottom pattern, like the scroll button), and
a spacer at the transcript's end reserves its height (DockOverlay
measures its settled block and reports upward; ChatView's dockReserve
state feeds the spacer). At-rest geometry is identical: the spacer
reserves exactly the overlay footprint.
The fold is now VISUAL ONLY: COLLAPSE_DIST 64px (was dock-height+64),
EXPAND_DIST 16px, position-based, no velocity. The reserved tail GROWS
immediately (typing, new variants) but never drops while a question is
open — a shrinking tail would reintroduce the clamp; in chat mode drops
apply (the composer typing-shrink is the old harmless behavior).
expandIntent (2s bar-click suppression) and the teach reveal gate are
unchanged.
Verified live (headless chromium, dev 3001, pending-quiz chat): teach
invitation mounts collapsed; click → expand + bottom; scroll up 200px →
folds to the bar at d=200 with the spacer held (no geometry change, no
yank); back to bottom → re-expands; spacer ≡ overlay footprint.
Note: the overlay spans the column's full width, so on platforms with
classic (non-overlay) scrollbars its gradient/background strip covers
the bottom of the scrollbar track — invisible on macOS overlay
scrollbars; flagged for master.
Model + thinking control refinements (components/model-picker.tsx):
- Drop the '·' separator between the two buttons.
- The thinking button shows just the level ('High', 'Max') — same
doctrine as the model button showing just the model name; '—' when
unknown. Pinned in render-stats-bar (label + no-dot assertions).
- The dropdowns now close on any pointerdown outside the control
(the toggle buttons stay inside the container and keep their own
toggling) and on Esc — the intuitive dropdown contract, instead of
the button being the only way out.
Note: file is graph-stream-owned per SCOPE; touching it per the owner's
direct request (same as cd518a4) — flagged for master.
The owner measured the arrowhead in Photoshop: past ~40° off the
perpendicular of the box edge, one barb crosses the tip's level and the
silhouette breaks. edgePathD now takes the target box and clamps the
final approach: the entry direction may tilt at most ±ENTRY_ANGLE_LIMIT
(40°) off the entry face's outward normal — a steeper natural approach
swings to the cone edge (same rotation side as the route), the shaft
stub and head follow the clamped direction.
Docking refined: the tip floats DOCK_GAP past the border along the FACE
NORMAL (constant gap at any tilt; perpendicular entries are bit-identical
to before — pinned). Faces are derived from where dagre's border point
sits on the target box (top/bottom/left/right all supported). Edges
without a known target box keep the unclamped geometry. Pinned: cone-edge
swing at 40° exactly, normal-gap docking, unclamped shallow entries.
Round-12 regression: manually repeating the mouth point as a control
point made d3 emit a degenerate final path command (a zero-length L).
orient="auto" derives the marker angle from the direction of the LAST
path segment; a zero-length segment has none, so every head fell back
to 0° — pointing right. Fix: drop the manual repeat — curveBasis already
repeats the endpoint internally, so controls ending [Q, M] keep the
exact end tangent (the approach direction) AND a real final segment
(the straight Q→M stub). Left a comment in edgePathD so this never
regresses again.
Two edge-drawing refinements (owner), both pure post-processing in
edgePathD on dagre's waypoints:
- Docking gap (4 units ≈ half a head length): the head's tip floats
short of the target face instead of touching it — the mouth pulls
back by tip overhang + gap, so the head kisses nothing and tilted
entries stop biting into the box.
- Shaft stub (12 units ≈ 1.5 head lengths): from Q, every control
point sits exactly on the approach line, so the basis spline renders
that stretch perfectly straight and the head lands on a straight
shaft; the corner where the route squares up is rounded by the
curve, and the repeated mouth point makes the end tangent exactly
the approach direction. Collinear routes render identically
(pinned: on-axis start to mouth).
Angle alignment/perpendicular entry deliberately NOT touched yet —
owner wants to evaluate this first (round 13 candidate).
The old lines were our hand-rolled shortcut: every arrow from the
source's bottom-center to the target's top-center as a single S-bezier.
Dagre actually routes each edge as border-intersected waypoints that
leave/enter nodes on whatever side the route needs — mermaid's
behavior. We now keep that routing (owner: adopt it).
blockLayout also returns edgePts — dagre's per-edge waypoint lists.
edgePathD draws them through d3's curveBasis, the exact recipe
dagre-d3's own renderer uses (and mermaid's default flowchart curve):
endpoints repeat so the line starts at the source border and the END
pulls back by the tip overhang so the arrowhead's tip lands on the
target border with its mouth at the path's end. Pinned in tests.
Dangling/unroutable edges draw nothing. d3-shape added as a direct
dependency (^3.2.0, already in the tree via dagre-d3-es — lane touch,
flagged to master with dagre-d3-es).
Paint order inside a node group: box fill/stroke → yellow glow → text.
The round-9 background fill no longer dims the glow inside the current
node's box, edges behind boxes still dim under the fill (round 9
intact), and the title text stays crisp above the glow.
Arrows routed behind a node showed straight through the fill-none
boxes. The white (visited/current/monochrome) boxes now carry a 50%
opacity fill of the standard app background color (fill-background/50):
an edge passing behind composites UNDER the fill — dimmed to half, not
shown through — while the no-fill look survives on the background
surface. Upcoming (grayed) nodes keep their solid card fill.
The owner chose: adopt the same layout engine regular mermaid uses, keep
our SVG skin. That engine is dagre (Sugiyama-style layered layout) — and
mermaid 11 ships dagre-d3-es 7.0.14, so we import exactly that fork, via
the same deep subpath mermaid itself uses. No new code shipped; pinned
verbatim in package.json (lane touch — flagged to master).
blockLayout keeps its signature ({ x, y, nodeH, width, height }, all
consumers unchanged): it feeds dagre our geometry (NODE_W boxes, GAP_X/
GAP_Y separations, zero margin, per-node grown heights from round 7),
runs the layout (deterministic, synchronous, DOM-free — SSR and render
tests stay exact), and converts dagre's node centers back to top-lefts.
Dangling edges are ignored defensively (a ghost endpoint would spawn an
unmeasured node and break the measurer).
topoFor stays exported for order/depth semantics (tests); placement no
longer calls it. dagre's network-simplex ranker + crossing minimization
now handle multi-parent DAGs, so future plan shapes route like mermaid
diagrams do. On the existing tree fixtures dagre's numbers agree exactly
with the old hand-rolled layout — every render pin survives verbatim.
Fly-in, take two: round 4 kept settled OFF until the first fit, but the
flag flipped in the SAME commit as the transform change — the browser saw
the new transform + newly-added transition in one style update and
animated from the initial top-left zoom:1 state. The transition now
enables one paint AFTER the fit (requestAnimationFrame), so the popup
opens settled and only glides for user input afterwards.
Nodes show their FULL content (owner): wrapTitle no longer caps lines
or ellipsizes — oversized words are hard-chunked, nothing is dropped.
Boxes grow with their wrapped lines (NODE_H + LINE_H per extra line);
layout rows take their tallest node, edges leave from the source's real
bottom, the glow rect and the plan-window's centering follow the grown
height, and the text block centers in the grown box.
- Arrows into not-yet-visited nodes are grayed out like the nodes
(stroke-border line + a fill-border copy of the arrowhead); the
in-chat block stays all-white.
- The yellow glow is static — the pulse is gone (owner).
- Smaller blur: stdDeviation 7 → 5, glow rect margin 10 → 8.
- The glow clipped at the frame on narrow graphs: its visible reach
(GLOW_MARGIN + 3·GLOW_BLUR = 23 units) far exceeded GRAPH_PAD (8).
Surfaces that show the glow now pad by GLOW_REACH — the popup's
viewBox AND its fit box both use POPUP_PAD = GLOW_REACH, so the blur
never clips; the monochrome in-chat block stays tight at GRAPH_PAD.
arrowhead_1.svg (the CM/MJX long-arrow head) baked in — transform chain
collapsed to a clean scale 0.2 + y-flip: viewBox 0 0 50.2 104, tip at
(50.2, 52), notch mouth at x 27.2, tail opening 9.6 tall (≈10 at the
owner's full height 104).
Sizing per the owner's spec: line width stays 1.5; head height 16 makes
the mouth opening tuck the line flush (the LaTeX proportion). markerUnits
userSpaceOnUse, refX at the mouth (27.2) so the line terminates at the
notch — never at the narrow tip, where it would show through. Edges
shorten by the tip overhang (≈3.54 units) so the head's tip lands
exactly on the target node's border.
Look (rectangular, white language — owner spec):
- Boxes: no corner rounding (rx dropped everywhere).
- Visited: fill-none, white stroke, white text. Upcoming: today's gray
(fill-card, border stroke, muted text) — kept per owner. Current:
white stroke/text + a faint yellow highlight — a blurred yellow fill
(feGaussianBlur) that bleeds out over the node's edges, still breathing.
- Edges + interim arrowhead: white (marker swaps to the owner's SVG when
it lands: right-pointing, solid fill, tight viewBox, tip at right edge).
- The IN-CHAT plan block renders monochrome — every node white, no
status language, no glow; the popup/plan-window show the progression.
Popup entrance fix: the transition from round 3 animated the first fit
(graph flew in from the top-left zoom:1 state). The transition now stays
off until the first fit has been applied — the popup paints settled.
- ZOOM_MAX 2× → 4×: node titles reach ~50px rendered, comfortable for
close reading. Single tunable constant.
- Smooth zoom: a 180ms ease-out CSS transition on the transform wrapper —
CSS retargets from the current computed value, so chained wheel ticks
and the fit/button zooms all glide instead of snapping. Disabled while
dragging (per-frame pans must not rubber-band). panning is React state
toggled at drag start/end.
Root cause of the popup's fit/center failure: PlanDiagram rendered
w-full inside the popup's shrink-to-fit absolute wrapper — the svg
resolved to the popup's width regardless of its viewBox extent, so the
transform math (1 svg unit = 1 px at zoom 1) was off and the 'fit floor'
neither fitted nor centered.
- naturalSize: the popup's diagram renders at the viewBox width in px,
so fit/pan/zoom math holds.
- The popup's fit box is now the PADDED ink box (GRAPH_PAD all around):
centered strokes and the glow halo never clip at the frame edge.
- fitZoomFor lost its 0.15 artificial floor — for wide graphs it sat
ABOVE the true fit and broke 'fully zoomed out = the whole graph
visible, whichever axis binds' (owner's rule). The fit itself is the
floor; a 0.02 epsilon only guards degenerate viewports.
In-chat full plan block:
- TEXT_CAP_SCALE (16.5/12.5): node text renders at text-xs in svg units;
the block now carries maxWidth = extent × cap + mx-auto, so it still
fills the column whenever the column is narrower than the cap, but
never scales node text past the chat prose size (--lh-prose-size).
- GRAPH_PAD (8 svg units) around the full-extent viewBox: centered
strokes (half outside the geometry) and the −4 glow halo were clipping
at the frame edges — the padded origin (-8,-8) keeps all ink visible.
Only the full-extent render pads; the plan window's override viewport
and the popup's fit margins are untouched.
Test pins updated: mx-auto + max-width:232.32px + viewBox='-8 -8 176 152'.
\ding{55} = U+2717 BALLOT X — I'd extracted U+2715 (\ding{53}, a
different multiplication cross) last round by mistake. Marks now match
pifont exactly: check \ding{51} (U+2713), x \ding{55} (U+2717).
components/ui/glyph-mark.tsx: extracted vector outlines from the system
Zapf Dingbats font — U+2713 (LaTeX \ding{51}, the hand-drawn check) and
U+2715 (\ding{55}, the bulky cross) — as static SVG geometry on an
ink-bounded viewBox with a y-flip path transform (inverse of font axes).
This matches the pifont shape LaTeX renders, not a random glyph. Colors
stay: green/red like the verdict chips; the span keeps data-mark pins.
- Option rows: items-start → items-center — the icon span's top-misalign
came from the row alignment; now marks center across the row.
- The literal ✓/✗ look (owner preference over the round-3 lucide set)
re-rendered as hand-rolled SVG with thick rounded strokes (strokeWidth
3, round caps/joins) — the glyph shape, as a true vector, sized like
the row text. Test pins updated to data-mark=check|x.
The muted 'Why' box's px-3 double-inset with card p-4 — the W-label sat
deeper than the option-box edges. px removed (py-2 stays for vertical
air); the label now aligns with everything else on the left edge. The
rounded bg-muted box is bg-matched so its padding was invisible anyway.
- ✓/✗ in the answered card: replaced literal glyphs with lucide CheckIcon/
XIcon (strokeWidth 2.5) inside a fixed size-4 flex box — text elements
never carried the right size/vertical alignment, vectors do. Test pins
updated (lucide-check/lucide-x classes).
- All user-typed INPUT text now matches markdown output: lh-prose class
replaces the earlier font-prose (family-only) band-aid on the composer,
question-dock text/thinking/Other fields, home topic input, model search
box, and answered-card 'your thinking'/free-text records. lh-prose sets
family+size (16.5px) so placeholders inherit correctly.
- Home view: StatsBar under the composer (model+thinking control) — the
DOCK geometry matches the chat view (max-w-[46rem] px-4) so nothing
shifts when a session starts; pre-chat selection persists via deferred
save (ModelControl chatId-empty) + PUT after POST /api/chats.
- Answer marks (✓/✗): same visual size as row text, vertically centered;
colored green-500/red-500 matching the verdict chips.
- All user-typed inputs (composer, question-dock text/thinking/Other,
answered-card 'your thinking' and free-text records) now render
Source Sans 3 via the font-prose utility. Placeholders inherit the
family via the element.
- Answered-card explanation: markdown wrapper made truly inline so the
'Why:' label + the paragraph sit on one line — removes the oversized
leading gap.
- New session: literal '+' glyph removed — one lucide PlusIcon (the
duplicated glyph was PlusIcon + text '+').
- Top bar: a small pointer-events-none gradient under the header eases
the scroller into the bar exactly like the dock's bottom fade.
- Option subtext (opt.description): .lh-prose — Source Sans 3; 'I don't
know' copy stays JetBrains Mono.
- 'I don't know': now layout-matched with a radio circle; dashed border
preserved as the deliberate non-answer.
- Separator spacing: options group pb-1 → pb-2 so the gap before the
divider equals the 8px inter-option gap.
The plain user-text div renders Source Sans 3 too (.lh-prose) — the
teaching surface reads in one family, chrome labels stay JetBrains Mono.
Build green; gates clean. The reskin's final live-browser pass runs on
this tree (dev on 3001) — no screenshot tooling in this session, SSR
checked.
--text-xxs (10px) joins the ladder; every component-level text-[9|10|11px]
literal in owned components now pulls from the ladder (graph-view.tsx and
home.tsx untouched — graph stream). Cosmetic pins updated where touched.
COLUMN now spans the session area with a thin margin (px-6, no max
width); DOCK (the composer/question shell) is a deliberate narrower
46rem centered column — same elements, same places, width polish only.
- Session title: font-title (CMU Serif bold) at text-2xl, per the mock.
- MessageResponse's markdown container now carries .lh-prose (Source
Sans 3 + emphasis tokens) on every markdown surface; labels stay
JetBrains Mono. Unlayered plain CSS beats the text-sm utility.
- PhaseChip + PhaseDivider sizes now pull from the declared text ladder
(text-xs) instead of pixel literals.
- 'Learning Harness' wordmark → 'LH' (the owner's header in the mockup;
flagged as a copy change — it's visually the same brand mark).
- New session: + NEW SESSION row wired with one POST /api/chats then
navigate (the mock drew it as visual-only; POST already existed).
- Search: a SESSIONS label + inline input filtering the polled snapshot
client-side (title substring, no server round trip).
- Skin: sidebar stays bg-background with the edge #434343 border-r; the
selected row now carries the rounding + borders from the raised
surface #1C1C1C via accent tokens.
Live-driving the real browser caught it (the user's report: "they don't
work as buttons"): the stats row sits OUTSIDE the composer shell's
pointer-events-auto div, inside the sticky scroll wrapper's
pointer-events-none — every control on the row was unclickable, including
the pre-existing stats-detail popover button. The row now re-enables
pointer-events itself.
Also, per the user's sizing note: bar text 10px → 11px with a touch more
breathing room, and the model + thinking controls gain quiet pill
affordance (rounded-full border, card bg, ▾ chevron) — still muted by
design, but visibly buttons.
Verified live end-to-end in the browser: click opens the model menu ABOVE
(366 rows, search autofocused), "qwen flash" narrows to 5, picking
qwen3.8-flash persists + flips the label, the thinking menu shows exactly
that model's supported levels (Off/Minimal/Low/Medium/High), Medium
persists on the record via PUT, labels update. Test session settings
restored after the drive. Gates 350/350.
The model-control wave (user-decided design, 2026-09-12): ONE controller
for the conversation agents (prober/teacher are functionally one agent);
background sub-agents stay config-pinned. pi-ai owns the registry — the
shipped 366-model OpenRouter catalog (cost, context, modality,
thinkingLevelMap), getSupportedThinkingLevels/clampThinkingLevel for the
per-model thinking truth, and streamSimple's automatic clamp. No second
registry to drift out of sync.
Seams:
- config.ts: ReasoningEffort retired for pi's ThinkingLevel
(off|minimal|low|medium|high|xhigh|max). GLM roles go honest at "low" —
pi's catalog says GLM-5.3-flash accepts exactly low|high|max, and the old
"minimal" was silently clamped to low by streamSimple since the swap;
nothing changes behaviorally. visualMaker → qwen/qwen3.8-flash (fixes
the dead qwen/qwen3.8-max slug — the catalog only ships -0902; also the
user's cost pick, 0.15/0.47 per-1M).
- src/agent/models.ts: piEngineFor(role, settings?) — provider+id resolved
from the catalog, thinking level from settings ?? config, CLAMPED to the
model's supported list (a settings change between models can never send
an invalid level). PiEngine gains `level` (observability/tests).
- ChatRecord.settings {model?: {provider, id}, thinkingLevel?} — written
by PUT /api/chats/[id]/settings (replace-whole, no updatedAt churn),
read by chat-handler at turn composition: agents are stateless per turn,
so the change is next-turn-effective by construction. The turns log
still records the ACTUAL billed model (the honest witness).
- GET /api/models: the configured-provider catalog as client-safe DTOs
(core/contracts/models.ts) + config defaults; a second provider joins
the picker by registering it in piModels() — nothing else changes.
- PUT settings validation gate (src/server/model-catalog.ts): unknown
provider/model or unsupported level → 400 carrying the supported list,
never a silent clamp.
- scripts/sync-pricing.ts: regenerated PRICES/CONTEXT_LIMITS from pi-ai's
shipped openrouter.json (offline; same source as the engine + picker).
All 366 models priced; glm's stale 0.15/0.5 → pi's 0.075/0.25, limit
1310720 → 1048576 — cost is inferred at read time, so history re-prices.
- UI (components/model-picker.tsx): stats-bar right side becomes model +
thinking buttons; menus open ABOVE; fuzzy search over "provider id
name" (core/fuzzy.ts — pi's /model pattern); model rows show name/id/
price/context; the thinking menu lists exactly the chosen model's
supported levels (GLM: low|high|max — no dead options, no "off").
Model switch drops an incompatible pinned level (falls back to config,
which clamps honestly) rather than carrying a 400.
Tests (350/350, +15): catalog projection + defaults, the settings gate,
pi thinking truth pinned against upgrades (glm low|high|max; minimal→low;
off→low — GLM cannot disable thinking), the engine factory's
override/fallback/clamp/throw matrix, fuzzy ranking, store round-trip,
and the re-priced folds (usage, stats bar, activity dock).
Live smoke (disposable driver, deleted): catalog route 366 + defaults;
glm+minimal → 400 + supported; turn 1 billed qwen/qwen3.8-flash under
the override; mid-session switch flipped turn 2 to glm; clear-overrides
restored the config pin. ALL CHECKS PASS on real OpenRouter.
The custom-message scheme (user-decided 2026-09-12): system-authored
user-role messages are NOT human messages — they carry marker data parts
only, the adapter owns their model-facing wording, and the UI renders
them under control. Nothing fake is ever persisted as human speech.
The inventory (all migrated):
- probe→teacher handoff wake: was a text-bearing user message rendering
as a fake user bubble; now data-probe-handoff marker-only — the
adapter flattens it to HANDOFF_TEXT (core/contracts/chat.ts), the UI
renders a quiet 'Preparing your plan…' system line.
- plan accept/decline: was boilerplate text ('The plan is accepted…',
'I'd like the plan revised…') masquerading as learner speech — the
decline wasn't even marked; now data-verdict {accepted, at} on its own
message, adapter flattens to verdictText(), UI renders the quiet
verdict marker ('✓ Plan accepted' / '↻ Plan revision requested').
sendVerdict sends TWO REAL MESSAGES when the learner typed a note
(the note as a genuine user message, then the verdict) in ONE turn
via setMessages + sendMessage() — no 409 dance, no double teacher turn.
- sub-agent steers: the abort steer gains its data-subagent-aborted
marker part; the UI keys steer rendering off PARTS (artifact/aborted),
not metadata — metadata stays as pump mechanics only.
- record-adapter: systemTextOf() flattens marker-only messages; steers
keep their text (payload — briefs are content, not boilerplate).
phases.ts: plan→teach advances on data-verdict{accepted:true} only;
learner text alone can never sanction teaching (pinned). data-plan-
accepted retired, no back-compat (user-decided: wiped-data doctrine).
Pinned: record-adapter.test.ts (verdict/handoff flattening, note+verdict
order, steer payload kept, UI-only parts skipped), phases.test.ts
(verdict marker semantics, out-of-order, text-never-fakes), jobs.test.ts
(abort marker part), demos.test.tsx (badge renders off the part, not
metadata).
Live GLM smoke (disposable driver, deleted): probe→conclude→wake→plan→
decline-with-note (revised, still plan)→accept-with-note (teach) — ALL
CHECKS PASS, including marker-only handoff/verdict on the wire and zero
fake human speech in the transcript.
Gates: tsc, lint, build, 335/335.
The architectural answer to the validation-crash question: validation IS
first at the model-facing boundary (pi validates arguments before
execute; failures return to the model same-turn — all three live
incidents retried successfully). But the same bytes also exist UNVALIDATED
by design in two other places: the live wire streams raw tool-input
deltas DURING generation, and the transcript keeps the raw input of
REJECTED calls (honesty + the replay invariant). The consumers were
supposed to trust part STATES — but that trust was enforced by
TypeScript casts ('part.input as X'), which are compile-time lies with
no runtime check behind them. That is the hole both crashes came
through, and it should not have been possible.
The structural rule now: model-authored INPUT is consumed only through
parsers running the SAME zod schema the model faced — boundary-valid ⇔
consumer-parseable, mechanically. Server/learner-authored OUTPUT stays
trusted (our code authors it).
- src/core/tool-parts.ts (new): parseQuizInput (display parse over the
shared quizInputBase — strips the teacher-only nodeId, accepts both
prober and teacher boundary shapes), parseAskInput, parseProbeDoneInput.
- contracts/probe.ts: probeDoneInputSchema is now core-owned zod — the
probe_done tool declares it (single source of truth), the seam runs
the literal same schema.
- question-dock.tsx: the dock parses its input at entry; unparsable
renders the quiet malformed-question fallback, never a crash.
- probe-done-card.tsx: the input preview goes through the parse; the
old state reasoning and shape guards collapse into it.
Pinned in tests/tool-parts.test.ts: both exact live crash shapes
(stringified quiz options, stringified assessment), boundary-parity
cases for all three schemas, nodeId stripping, mid-stream partials;
render-cards pins the dock fallback markup.
Gates: tsc, lint, build, 328/328.
Two live bugs from the modal-realism stress runs:
1. Multi-select ask: picking the custom 'Other…' row cleared every
already-picked option, and picking any option cleared a typed custom
answer — single-select displacement logic applied to both modes. The
dock's selection is now a pure state machine in question-core
(toggleOption/toggleOther over DockSelection, unit-tested): single
displaces (unchanged), multi coexists — neither channel clears the
other; the typed text always survives.
2. ProbeDoneCard crashed on a REJECTED probe_done call: the model
stringified 'assessment' (the same double-encoding habit as the
options bug), zod rejected the call, the model retried — but the
card's input-preview fallback trusted the errored part's unvalidated
input shape and died on assessment.strands.map. The preview now reads
input only while the call is pending or streaming (guarded with
Array.isArray); an output-error part renders a quiet
malformed-and-retried note instead of the garbage.
Pinned: question-core.test.ts (3 dock-selection machine cases),
render-cards.test.tsx (4 probe-done card cases incl. the exact live
stringified shape, plus the multi-select ask markup pin).
Gates: tsc, lint, build, 315/315.
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.
The successor of the injected plan blocks (docs/probe-teach-split.md §6):
always on once a tree exists, a zoomed-in window (2:1, ~2¼ node widths)
centered on the CURRENT node, re-centering by pure derivation as tagged
checks move the lesson — statuses derive live from the transcript on
every render (derivePlanStatuses), glow on current, fallback focus on
the lesson's first node, click opens the existing full-DAG popup.
Bottom-left, mirroring the activity dock's bottom-right; hidden on
narrow viewports; the learner's visual polish pass comes later.
PlanDiagram gains a viewBoxOverride (exported); latestTree exported.
Tests: viewBox centering/clamping (pure) + render (hidden pre-plan,
windowed viewBox, glowing current, starting fallback).
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).
- src/core/plan-blocks.ts: pure emission builder — full block on successful
plan_init, neighborhood block when the derived plan state moved since the
newest existing block; the as-of state reads from the block's frozen
snapshot (transcript quiz parts mutate in place on answer, so
re-derivation is post-mutation blind). Idempotent by construction.
- turn-service persistTurn enriches with withPlanBlocks at the canonical
write; chat.tsx adopts the record once the server agrees the turn is
done (endTurn flips only after persist).
- chat-handler: per-request usage collector via agent-level onStepFinish
(StepResult.usage), appended at onFinish to ChatRecord.turns via
appendTurnUsage, idempotent per response message id.
POST /api/activity/abort landed on main (fde46c2). 404 not_found /
409 not_running mean the process reached the desired end state on its
own — refresh and clear the abort state, no error surfaced. Genuine
failures (network/400) show 'Abort failed — try again' on the blob.
The plan now lives IN the chat (plan blocks, wave-4 item 3) and background
processes live in the activity dock (item 1), so the right sidebar's
tree panel + sub-agent fleet column is retired from the chat view:
- delete components/tree-panel.tsx (three-mode rail/list/graph panel) and
components/sub-agents-panel.tsx; their graph interaction machinery lives
on in plan-block.tsx and fleet observability in activity-dock.tsx.
- app/layout.tsx drops TreePanelModeProvider (the mode no longer exists).
- chat.tsx drops the xl:flex right column; the middle column is the view.
- tests: render-tree-panel.test.tsx deleted; the sub-agents panel suite
removed from demos.test.tsx (dispatch cards' jobs-context stays — the
central poll drives transcript dispatch cards).
- components/plan-block.tsx: hand-built SVG (user decision) — deterministic
layered layout (Kahn topo + longest-path depth, rows centered, authoring
tie-break), bezier edges, the retired panel's status color language.
- PlanBlock renders data-plan-block parts AS CARRIED (frozen snapshots,
never re-derived): kind full = borderless, viewBox = layout extent so
nodes hug the column edges, natural height; kind neighborhood = own
small labeled 'Plan fragment' container with the current node glowing
(pulsing primary halo).
- PlanPopup: full-screen overlay with the LIVE plan — latest tree from the
transcript, statuses derived at click time via derivePlanStatuses
(contracts/plan.ts). Esc / backdrop / X close; drag pan, wheel
zoom-to-cursor, zoom buttons; fit-floor semantics harvested from the
retired tree panel (zoom floor = the fit, pan locks at it).
- chat.tsx: data-plan-block case in the part switch + popup mount.
- src/core/ui-message.ts: TeacherUIDataParts gains "plan-block": PlanBlockData
(additive, type-only — lane touch flagged to master).
- tests/render-plan-block.test.tsx: pure layout (layering, centering,
dangling-edge tolerance), wrap, frozen-snapshot render, live derivation.
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.
One quiet row beneath the composer/question shell (not flush to the
screen bottom; static height, outside the morphing shell):
- LEFT: context window + session cost, clickable → per-turn detail
popover (context, total, unpriced count, one row per turn with
model/steps/cost). All folds from core/prices.ts — sessionCost,
contextWindow, turnCost — no cost math in components.
- RIGHT: model + "Thinking —" — a placeholder slot for the coming
model/thinking-level CONTROL (later work with its contract changes);
display-only today, model read from the last turn's usage,
pretty-printed (Glm 5.3 Flash).
- Data: the record's turns log, seeded on mount, refreshed when the
local stream settles, kept fresh by the existing record re-syncs.
No turns yet → "—" everywhere; unpriceable turns never fold into a
fake $0 (the doctrine).
- formatUsd/formatTokens moved to components/usage-format.ts (shared
with the activity dock); prettyModel joins them, pure + tested.
One reserved corner of the shell, one blob per running BACKGROUND
process — session sub-agents (research/demo/visual) and global
extraction runs. Teaching turns are deliberately NOT here: the dock
exists to surface what the learner can't otherwise see; a running
teaching turn is already fully visible where it happens (user decision).
- Data: ONE poll of GET /api/activity (jobs across ALL chats,
title-joined, oldest-first + extraction runs), folded client-side
with core/prices.ts (jobCost). Adaptive interval — 2.5s while
anything runs, 10s idle, 4× slower on hidden tabs; a failed fetch
keeps the last feed (no flicker to empty on a blip).
- Per blob: kind icon, elapsed (ticking), cost, activity line, and an
explicit END on the blob's face. Click → detail popover (context
line, model, steps, open-session link). The context line is the LAST
model call's tokens vs the model's limit (JobUsage.lastStep — pending
agents-stream contract; "—" until it populates), NEVER a cumulative
sum. Absent data renders "—", never zero (the doctrine).
- Header: total running cost (sum of priced blobs, "—" when nothing
is priced). Narrow viewports collapse to a count badge → drawer.
- END posts to /api/activity/abort {kind, id} — the backend is being
composed by master over the streams' mechanisms; until it exists a
click locks the button and the popover says "Abort not available
yet" (never a dead control).
- Idle renders NOTHING. Mounted once in the root layout; every view
gets it for free.
toBlobs / formatElapsed / formatUsd / formatTokens are pure + tested;
SSR render tests pin blobs, the "—" doctrine, the face END, and the
priced total.
amosblomqvist