`<CopilotKit agent="beautiful-chat">` routes the chat to agent id
"beautiful-chat", but ExampleCanvas called `useAgent()` with no args and
fell back to DEFAULT_AGENT_ID ("default"). The frontend's agent registry
tracks state per id, so `manage_todos` state-deltas from the chat run
landed on "beautiful-chat" and never reached the canvas's "default"
subscription — the Task Manager pill auto-flipped the panel to App mode
but the To Do column stayed empty. Drop the unused "default" alias from
the runtime route and pin the canvas to `useAgent({ agentId:
"beautiful-chat" })` so both halves share one ProxiedCopilotRuntimeAgent
instance. Adds a Playwright regression test asserting the 3 verbatim
todo titles render after the pill click, plus 3 aimock fixtures for the
multi-turn flow (enableAppMode -> manage_todos -> confirmation).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three layers of regression guards for the runtime reasoning-role filter and
the chained demo behavior:
1. Runtime unit test — packages/runtime/.../run-message-filtering.test.ts:
- Verifies `LangGraphAgent.run` strips `role:"reasoning"` from
`input.messages` before delegating to super.run.
- Verifies user/assistant/system/tool messages pass through in order.
- Verifies empty + missing messages arrays are tolerated.
- Verifies pre-existing forwardedProps.streamSubgraphs default + override
behavior is preserved.
- 6/6 tests pass against the runtime package's vitest config.
2. D5 harness probe — showcase/harness/.../d5-tool-rendering-reasoning-chain.ts:
- Expanded from one chained turn (flights→weather) to all three chained
pills in a single thread (stocks AAPL→MSFT, dice d20→d6, flights→weather).
- This is the canonical multi-pill regression at the harness layer:
without the runtime reasoning-role filter, the second pill would crash
before the model was called.
- Each turn asserts the per-turn delta of reasoning-block mounts (idx+1),
the minimum card count for each tool group, and unique transcript
substrings that scope to that turn.
3. Playwright e2e spec — showcase/integrations/langgraph-python/tests/e2e/
tool-rendering-reasoning-chain.spec.ts:
- Mirrors the pattern of the sibling tool-rendering-default-catchall spec
(notably its multi-pill regression at lines 162-212).
- Page-loads test verifies the 3 pills mount and no cards leak from a
prior session.
- One test per chained pill (stocks, dice, flights+weather) asserts the
full chain renders with reasoning-block + correct per-tool cards +
narration matching the aimock fixture text.
- Sequential-pills regression test clicks all 3 pills in one thread,
asserts each chain renders independently AND the reasoning-block count
increases monotonically across turns.
Agent: extend `get_stock_price` to accept optional `price_usd` and
`change_pct` arguments (mirrors the basic tool-rendering agent's signature
introduced in #4770). The aimock fixtures script the chained AAPL/MSFT
comparison by passing deterministic prices via these args; without the
wider signature, pydantic rejects the tool call and the card never mounts.
The runtime unit test is the strongest guard — it would catch any
regression on the role-filter logic without depending on the full Docker
stack. The harness probe and Playwright spec catch end-to-end regressions
in the canonical CI environment.
The tool-rendering-reasoning-chain demo previously promised chained tool
calls in its pill titles but the agent and fixtures only delivered single
tools — clicking "Weather + flights to Tokyo" produced just a WeatherCard,
"Compare two stocks" only fetched AAPL, "Find flights from SFO to JFK"
showed flights but no destination weather. Three changes close the gap.
Agent: replace the soft "call 2+ tools when relevant" system prompt with
concrete per-pill chain examples mirroring the pattern already used by the
langgraph-typescript `tool-rendering` agent (weather→flights, ticker→peer,
roll→contrast die, flights→destination weather).
Pills: drop the redundant Tokyo pill (it was the SFO/JFK chain in reverse)
and reword each remaining pill message to PRE-DISCLOSE the chain so the
model commits to the follow-up call:
- "Compare AAPL and MSFT stocks for me."
- "Roll a 20-sided die for me and compare it to a smaller one."
- "Find flights from SFO to JFK and show me the weather there."
Fixtures: 9 fixtures (3 per pill: final-content → second-leg → first-leg,
ordered by toolCallId specificity for first-match-wins). Each fixture is
scoped by a langgraph-python-UNIQUE userMessage tail ("Compare AAPL and
MSFT stocks", "compare it to a smaller one", "show me the weather there").
Those substrings appear nowhere else across the 14+ integrations sharing
showcase-aimock on Railway, so the new fixtures cannot cross-contaminate
the other reasoning-chain demos that still ship the older prompt set.
A toolName-based gate was considered and rejected because most fleet
agents register `roll_dice` and aimock's `toolName` matcher is a tool-LIST
gate, not a tool-CALL gate — it would NOT have isolated this demo.
Probe: collapse the two-turn flow (Tokyo + SFO/JFK) into one chained turn
(SFO→JFK + JFK weather) that asserts BOTH per-tool renderers
(FlightListCard + WeatherCard) mount in a single response. Same coverage
at half the wall-clock and exercises the actual chained-tool path.
Four independent showcase production bugs Alem reported, plus the
D5 multimodal harness regression they unblocked.
Shared-state-read-write: "Greet me" ("Say hi and introduce yourself.")
and "Plan a weekend" ("Suggest a weekend plan based on my interests.")
were matching the bare `hi` and `plan` catch-alls in feature-parity.json
and returning the generic showcase-assistant blurb / 5-step content plan
instead of shared-state-aware responses. Added pill-specific fixtures in
shared-state.json (mirrored into d5-all.json) so the longer userMessage
substrings win first-match-wins ahead of feature-parity.
Auth sign-out: signing out unmounted CopilotKit entirely and bounced
the user back to the SignInCard, so the demo never showcased the
runtime returning 401 — its whole point. The QA contract in
qa/auth.md spelled out the intended UX. Restored it: CopilotKit stays
mounted after the first sign-in, the AuthBanner flips to an amber
"Signed out — the agent will reject your messages" state with a
re-Sign-in button, and CopilotKit's `onError` callback drives a
`data-testid="auth-demo-error"` surface that displays the runtime's
401 the moment the user sends an unauthenticated message. Updated the
e2e spec to match (the old "SignInCard re-mounts after sign-out" test
pinned the regression).
Gen-ui-agent: the aimock fixture short-circuited the 7-step
progression spelled out in `gen_ui_agent.py`'s SYSTEM_PROMPT to a
single set_steps call with all three steps already `completed`, so
the InlineAgentStateCard rendered the final 3/3 state instantly with
no sequential pending → in_progress → completed animation.
Regenerated as a 7-leg toolCallId chain per pill (8 fixtures × 3
pills): seed leg keyed on userMessage with NO `hasToolResult` gate
(matching PR #4770's pattern — `hasToolResult: false` would block the
seed from firing on the second pill in a multi-pill session), then
six toolCallId-keyed transitions, then a final narration. Fixture
order: toolCallId legs FIRST so the most specific match wins.
Multimodal D5: the sample-attachment buttons auto-send via
`agent.addMessage + copilotkit.runAgent` (restored in PR #4761), but
the D5 harness still typed `input` + pressed Enter via the runner
after `preFill`, sending a second user message that competed with the
in-flight image upload — the v1 LangGraph runtime SSE stream got
tangled (browser DevTools showed `statusCode: pending` indefinitely)
and the assistant message never rendered. Added `skipSend?: boolean`
to ConversationTurn (distinct from `skipFill`, which still presses
Enter once the textarea has content) and switched d5-multimodal.ts to
`skipSend: true` with `responseTimeoutMs: 60_000` so the runner waits
on the assistant response without poking the chat further. Bumped the
PDF auto-prompt fixture in feature-parity.json to include the word
"document" so the existing `buildModalityAssertion("document")` check
still lands.
D5 result: 37 → 39 of 40 features passing. Only
`tool-rendering-reasoning-chain` remains and is a separate
agent/runtime bug (Tokyo Responses-API `reasoning` message survives
into the next turn's conversation history, runtime returns
`RUN_ERROR: "message role is not supported"`).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The chat-slots cell is wired to the neutral sample_agent graph (plain
ChatOpenAI, no Responses API, no reasoning config), so it never emits
AG-UI REASONING_MESSAGE_* events. The pill could never light up the
wrapped messageView.reasoningMessage slot, and its prompt didn't match
any fixture in showcase/aimock/d5-all.json — aimock-backed runs hit
"No fixture matched". Drop the pill (the QA doc and the e2e spec
already only expect "Write a sonnet" and "Tell me a joke") and leave a
note pointing reasoning demos at /demos/reasoning-default and
/demos/reasoning-custom where the dedicated reasoning_agent graph lives.
The `[&_div[style*='flex-direction:_row']]:gap-4` arbitrary variant
(quotes inside doubly-nested brackets) is the most exotic Tailwind
syntax in this PR and lines up exactly with when the Vercel
form-filling deploy started failing. Tailwind v4's content scanner is
likely choking on the apostrophes in the nested attribute selector.
The Metric `flex-1 min-w-[120px]` and the chart `flex-1 min-w-0`
already give us even distribution inside the basic catalog's gap-less
Row; the auto-injected nested gap was nice-to-have, not load-bearing.
Drop the inline <style> override that widened the chat's `cpk:max-w-3xl`
column and the outer max-w-6xl bump. The chat keeps its normal width;
the real bug was that the basic catalog's Row/Column primitives are
bare `display: flex` divs with no gap and no min-width control on
children, so when the agent dropped multiple Metrics or charts into a
Row they collapsed to content width and looked glued together.
Four targeted fixes inside our renderers:
- Metric gains `flex-1 min-w-[120px]` so a row of KPI tiles distributes
the available width evenly inside the gap-less basic Row, instead of
shrinking to content.
- PieChart and BarChart switch from a hardcoded max-w to `flex-1
min-w-0` so two charts side-by-side each take half the card column
with Recharts' ResponsiveContainer doing the rest, instead of one
chart insisting on 640px and overflowing.
- Card's CardContent picks up a Tailwind arbitrary variant
`[&_div[style*='flex-direction:_row']]:gap-4` (plus the column
equivalent) that injects a gap into any nested basic Row/Column the
agent drops in. Underscores in the arbitrary value compile to literal
spaces, matching React's serialized inline `flex-direction: row`.
- Card itself drops the old `min-w-[260px]` floor in favour of
`min-w-0` so it cooperates if the agent ever stacks Cards horizontally
inside a Row.
The chat shell caps its scroll column at cpk:max-w-3xl (~768px), which
left A2UI-generated cards (KPI dashboards, charts, status reports)
feeling pinched on the declarative-gen-ui demo. Locally widen that
wrapper to 64rem via a scoped attribute selector on the demo and bump
the outer page wrapper from max-w-4xl to max-w-6xl so the card column
actually has room to grow.
While here, fix the InfoRow trailing-separator artifact: each row now
draws its own border-bottom with last:border-b-0 so the final row in a
Card (e.g. the Status Report demo) no longer leaves a dangling line,
regardless of whether the agent wraps the rows in a Column or drops
them directly into the Card child slot. Right-align the value with
tabular-nums for cleaner stacks. Card itself gains w-full
overflow-hidden so it stretches into the now-wider column instead of
sitting at its min-width.
## Summary
Re-lands the multimodal-attachments fix from #4584 (May 1, never merged)
onto current `main`, ported to the post-refactor file layout where
`page.tsx` was split into `legacy-converter-shim.tsx`,
`multimodal-chat.tsx`, and `file-to-data-attachment.ts`.
Auto-send was the visible regression: clicking **Try with sample image /
Try with sample PDF** only queued the attachment chip instead of sending
the canned prompt. This PR restores the full end-to-end behavior plus
five regression tests so it can't silently break again.
## What was broken and what changed
1. **Random uploads crashed with `Failed to fetch`.** aimock returned
HTTP 404 on no-match, the LangGraph SDK surfaced `NotFoundError`, the
AG-UI stream surfaced a `RUN_ERROR`, the demo crashed. → Added
`--proxy-only` + `--provider-openai https://api.openai.com` to the local
aimock command so unmatched user prompts fall through to real OpenAI
(mirrors Railway).
2. **Bundled-sample fixtures keyed on user-visible canned prompts.**
Auto-prompts are now natural and specific ("can you tell me what is in
this demo image/pdf I just attached") so they render cleanly as the user
message bubble AND can't collide with arbitrary user prompts — random
uploads phrase questions differently and fall through to the proxy.
3. **Sample buttons now auto-send via `useAgent`.** The previous
DataTransfer path queued the attachment via the chat's hidden file input
but required clicking send while the attachment was still uploading —
`CopilotChat.onSubmitInput` rejects submits during upload AND clears the
input regardless, so the canned prompt was eaten. Rewrite calls
`agent.addMessage(...)` + `copilotkit.runAgent({ agent })` directly with
the base64'd content part.
4. **PDF flattened text bled into the rendered user message.**
`_PdfFlattenMiddleware` ran in `before_model` and persisted the rewrite
to agent state. Switched to `wrap_model_call` so the PDF→text rewrite is
scoped to the model request only.
5. **Attachments doubled (and PDFs rendered as broken `<img>`).** The
`@ag-ui/langgraph` round-trip mis-tags PDFs as `image` and re-injects
the user's original modern part, doubling chips. Added
`dedupeUserMessageMedia` subscriber on `onMessagesSnapshotEvent` +
`onRunFinalized` to dedupe by `source.value` and re-key type from
mimeType. Also flipped `onRunInitialized` from REPLACE to APPEND so the
modern part stays for the UI alongside a legacy `binary` sibling for the
converter.
6. **Regression suite (`tests/e2e/multimodal.spec.ts`).** Five focused
tests, all pass against live local stack (15.4s):
- page loads with all expected affordances
- sample image: auto-sends, EXACTLY ONE `<img>`, assistant references
the logo
- sample PDF: auto-sends, EXACTLY ONE `DocumentAttachment` chip ("PDF"
label), NO `<img>`, no `[Attached document]` text bleed
- image then PDF in the same session: each message keeps its own single
chip
- PDF then image in the same session: symmetric
## Test plan
- [x] `showcase up langgraph-python` — both sample buttons auto-send;
image renders as `<img>`, PDF renders as PDF chip; random paperclip
uploads go through proxy
- [x] `BASE_URL=http://localhost:3100 CI=1 npx playwright test
multimodal.spec.ts` — **5 / 5 passing**
- [ ] Post-merge: e2e-deep cycle for langgraph-python multimodal cell
stays green
## Closes
Closes#4584.
The langgraph-python multimodal-attachments demo had a stack of bugs
that compounded each other. Fixing them required touching the local
docker-compose, the aimock fixtures, the LangChain middleware, the
client-side AG-UI shim, and the sample-attachment buttons. This
commit lands the full set together because they only make sense as
a unit — verified end-to-end against `showcase up langgraph-python`
in a headed browser. New e2e suite pins each regression.
Supersedes #4584 (the original fix from May 1 that never landed —
this is a fresh port onto the post-refactor file layout where
page.tsx is split into legacy-converter-shim.tsx, multimodal-chat.tsx,
file-to-data-attachment.ts).
What was broken and what changed:
1. Random uploads crashed with `Failed to fetch`. aimock returned
HTTP 404 on no-match, the LangGraph SDK surfaced `NotFoundError`,
the AG-UI stream surfaced a `RUN_ERROR`, the demo crashed.
Added `--proxy-only` + `--provider-openai https://api.openai.com`
to the local aimock command so unmatched user prompts fall through
to real OpenAI (mirrors the Railway aimock setup).
2. Bundled-sample fixtures keyed on user-visible canned prompts.
The auto-prompts are deliberately long, specific, and natural-
reading ("can you tell me what is in this demo image/pdf I just
attached") so they (a) render cleanly as the user message bubble,
and (b) can't collide with arbitrary user prompts — random
uploads phrase questions differently and fall through to the
proxy.
3. Sample buttons now auto-send via `useAgent`. The previous
DataTransfer-based path queued the attachment via the chat's
hidden file input, then required clicking send while the
attachment was still uploading — `CopilotChat.onSubmitInput`
rejects submits during upload AND clears the input regardless,
so the canned prompt was eaten. Rewrite to call
`agent.addMessage(...)` + `copilotkit.runAgent({ agent })`
directly with the base64'd content part, sidestepping the
upload race entirely.
4. PDF flattened text bled into the rendered user message.
`_PdfFlattenMiddleware` ran in `before_model` and returned
`{"messages": rewritten}`, which persisted to agent state. The
chat UI then rendered the `[Attached document]\n<pdf body>` text
part inline with the user prompt. Switched to `wrap_model_call`
so the PDF→text rewrite is scoped to the outgoing model request
only and never pollutes state.
5. Attachments doubled (and PDFs rendered as broken `<img>`). The
`@ag-ui/langgraph` round-trip translates outgoing `binary` parts
to LangChain `image_url` and incoming `image_url` back to `image`
AG-UI parts — regardless of mimeType, so PDFs came back as
`type: "image"` with `mimeType: "application/pdf"` and were
forced into `ImageAttachment`, where the load failed and the
chat showed two "Failed to load image" boxes. Plus the user's
original modern part survived alongside the round-tripped one,
doubling visible chips.
Added a `dedupeUserMessageMedia` subscriber on both
`onMessagesSnapshotEvent` and `onRunFinalized` to:
- dedupe media parts by `source.value` so the local + round-
tripped copy collapse to one chip
- re-key part `type` from `mimeType` so PDFs route to
`DocumentAttachment` (icon + filename) and images to
`ImageAttachment`.
Also flipped the `onRunInitialized` shim from REPLACE to APPEND
— keep the modern part for the UI AND emit a legacy `binary`
sibling for the converter.
6. Regression suite (`tests/e2e/multimodal.spec.ts`). Replaces the
pre-rewrite suite with five focused tests:
- page loads with all expected affordances
- sample image: auto-sends, EXACTLY ONE `<img>`, assistant
references the logo
- sample PDF: auto-sends, EXACTLY ONE `DocumentAttachment` chip
("PDF" label), NO `<img>`, no `[Attached document]` text bleed
- image then PDF in the same session: each message keeps its own
single chip, no cross-contamination
- PDF then image in the same session: symmetric
All 5 pass against the live local stack (15.4s).
The voice route's OpenAI client previously fell through to OPENAI_BASE_URL,
which docker-compose.local.yml sets to http://aimock:4010/v1. Aimock has a
catchall transcription fixture that returns "What is the weather in Tokyo?"
for every audio file, so the mic button always produced that phrase no
matter what the user actually said.
Pin baseURL to real OpenAI (overridable via OPENAI_TRANSCRIPTION_BASE_URL).
The sample-audio button stays as synchronous text injection — that's the
documented design, and what the e2e + d5 probe rely on.
Also:
- Tidy the sample button label ("Try a sample question" -> "Try a sample
audio") so the affordance matches what it does.
- Realign tests/e2e/voice.spec.ts with the shipped component (the
voice-sample-audio container testid and Sample: "..." caption it asserted
on never existed on HEAD) and add cold-start timeout headroom for the
mic-button render and the agent-flow test.
- Add "env": ".env" to langgraph.json so langgraph_cli dev picks up
OPENAI_API_KEY locally. Docker/Railway paths inject env vars directly so
this is a no-op there.
## Summary
Phase 4 validation (run today against `docs.showcase.copilotkit.ai`)
surfaced four cutover-blocking issues. This PR fixes all of them in four
focused commits.
## Commits
1. **chore(showcase): close last 2 yellow Missing snippet boxes for
cutover** — adds in-place region markers to
`langgraph-python::frontend-tools` and a sibling
`slot-overrides.snippet.tsx` teaching file for
`langgraph-python::chat-slots`. Both pages now render zero `Missing
snippet` warnings.
2. **fix(shell-docs): correct feature-viewer slug + demo-id translation
for code tab** — adds `getFeatureViewerSlug()` and
`getFeatureViewerDemoId()` helpers in `registry.ts`, with explicit
override maps for the 8 framework-name mismatches (built-in-agent,
google-adk, claude-sdk-{python,typescript}, ms-agent-{python,dotnet},
crewai-crews, llamaindex) and 5 demo-id mismatches (gen_ui_tool_based,
gen_ui_agent, shared_state_streaming, shared_state_read_write,
hitl_in_chat). When the framework or demo has no feature-viewer
equivalent, the helper returns null and the Code tab is hidden on that
page (graceful degradation; Demo tab still renders).
3. **fix(shell-docs): redirect catalog hygiene (self-loops +
framework-scoped gap)** — removes 31 self-redirect entries from
`seo-redirects.ts` (where source === destination caused infinite 301
loops on canonical URLs like /frontend-tools, /faq, /human-in-the-loop).
Reorders middleware logic so the redirect catalog is consulted before
the framework-scoped short-circuit fires, fixing 97 framework-scoped
slug-rename URLs that were soft-404ing instead of redirecting. Adds
defense-in-depth skip-when-equal guard in middleware.
4. **fix(shell-docs): resolve 53 sitemap 500s before cutover** — three
independent root causes in MDX rendering:
- 4 missing \`<Component />\` registrations in \`mdx-registry.tsx\`
(CopilotCloudConfigureCopilotKit,
SelfHostingCopilotRuntimeConfigureCopilotKit, CloudCopilotKit, Content)
— affected ~37 pages.
- 3 langgraph tutorial pages had markdown lists immediately preceding
JSX closing tags, causing remark to bail. Fix = blank line between
bullet and close tag (~9 pages).
- 5 MDX files used escaped JSX comments that Acorn cannot parse —
replaced with proper unescaped form (~7 pages).
## Verification
- Production build clean: 27 static pages generated, 0 errors
- Local probes: all 53 sitemap-500 URLs return 200; all 31 former
self-loop URLs return 200; all 85 framework-scoped non-loop catalog
entries 301 to expected destinations
- Code tab: 47/60 (framework × demo) URLs resolve to real code panels
post-fix; remaining 13 are feature-viewer per-framework deploy-coverage
gaps in the \`ag-ui-protocol/ag-ui\` repo, not this one
- Both langgraph-python pages render 0 \`Missing snippet\` boxes after
re-bundling demo-content
## Test plan
- [ ] CI passes
- [ ] Sitemap probe: every URL in \`/sitemap.xml\` returns 200
- [ ] Redirect probe: spot-check 5 framework-scoped slug renames (e.g.
\`/agno/frontend-actions\`, \`/pydantic-ai/use-agent-hook\`)
- [ ] Self-loop probe: \`/frontend-tools\` returns 200, not infinite 301
- [ ] Code tab: visit \`/built-in-agent/frontend-tools\` (tab should be
hidden), \`/langgraph-python/agentic-chat\` (tab should render code),
\`/llamaindex/agentic-chat\` (tab should render after llama-index slug
rename)
- [ ] Yellow boxes: visit \`/langgraph-python/frontend-tools\` and
\`/langgraph-python/custom-look-and-feel/slots\` — zero Missing snippet
warnings
User-facing renames so the showcase reads the way a cold visitor would
expect:
- `byoc-hashbrown` → `declarative-hashbrown` (and `byoc-json-render` →
`declarative-json-render`). The display titles already said
"Declarative UI: …"; only the URL slugs and folder paths still
leaked the internal BYOC ("Bring Your Own Components") jargon.
Renamed:
/demos/byoc-hashbrown → /demos/declarative-hashbrown
/demos/byoc-json-render → /demos/declarative-json-render
/api/copilotkit-byoc-* → /api/copilotkit-declarative-*
src/app/demos/byoc-* → src/app/demos/declarative-*
qa/byoc-*.md → qa/declarative-*.md
tests/e2e/byoc-*.spec.ts → tests/e2e/declarative-*.spec.ts
Internal Python module names + langgraph graph IDs stay legacy
(`byoc_hashbrown_agent.py`, `byoc_hashbrown`) — those are not
user-facing and renaming them is a separate cross-codebase pass.
- `a2ui-fixed-schema` slug intentionally unchanged.
- Tool Rendering trio parenthetical rename (Default → Catch-all →
Custom progression reads clearly as "how much do I customize?"):
Tool Rendering (Default) — unchanged
Tool Rendering (Custom default) → Tool Rendering (Catch-all)
Tool Rendering (Specific) → Tool Rendering (Custom)
- `tool-rendering-reasoning-chain` cell renamed from
"Generative UI: Rendering multiple tools" to
"Generative UI: Tool calls + reasoning" (the demo is about combining
reasoning + tool rendering, not about quantity of tools).
- `Open Generative UI: Default` / `Open Generative UI: Custom`
descriptions expanded so a visitor understands how Open Generative UI
differs from Tool Rendering (agent composes UI from a registered
library vs. attaching a renderer to a *named* backend tool).
- Showcase index now sorts demos within each tag by `manifest.features`
order. Previously demos appeared in manifest declaration order, which
ignored the team's curated "polished flagship → simplest start →
variants" arc.
Cross-cutting registry / harness / dashboard updates that fall out of
the rename:
- `shared/feature-registry.json` adds the two new IDs alongside the
legacy `byoc-*` (so the catalog stays valid; the other 17
integrations still declare `byoc-*` in their manifests).
- `shared/constraints.yaml` adds the new IDs to the
generative-ui-approach allow-list.
- `scripts/__tests__/generate-catalog.test.ts` updates the cell-count
expectations (45 features × 18 integrations = 810; 792 after docs-
only exclusion; 45 LGP cells = 38 wired + 1 stub + 6 unshipped).
- Harness probe `d5-byoc.ts` + `d5-byoc.test.ts` now route both slug
families through `preNavigateRoute` and exercise the new branches.
- `d5-feature-mapping.ts` and `shell-dashboard/live-status.ts` mirror
the dual-ID mapping so both legacy and renamed slugs roll up under
the same `byoc` D5 featureType.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Extracts duplicated logic into shared modules so each demo file reads
as the feature it teaches, not the boilerplate around it.
Python agents:
- `_a2ui_utils.py` (new) — `sanitize_a2ui_components` and
`has_root_component`, consumed by both `a2ui_dynamic.py` and
`beautiful_chat.py` (these previously duplicated the same defensive
validator inline).
- `byoc_hashbrown_prompt.py` (new) — the 56-line system prompt extracted
from `byoc_hashbrown_agent.py` so the agent file stays focused on the
`create_agent(...)` wiring.
- `beautiful_chat.py` secondary `ChatOpenAI` now passes
`streaming=True` (matching `a2ui_dynamic.py`) so aimock's SSE-only
fixture matcher sees the call in replay mode — without it the demo
surfaced "An internal error occurred" on every load.
- `multimodal_agent.py` — top-level `from pypdf import PdfReader` (was
lazy with three layers of `# pragma: no cover` exception handling
for stages that never failed independently); kept a single log line
at the outer except so Railway logs stay triageable.
- `gen_ui_agent.py` — switched from `deepagents.create_deep_agent` (whose
planner+sub-agent middleware ate enough supersteps to trip LangGraph's
default recursion limit on this single-tool ReAct loop) to plain
`langchain.agents.create_agent`. Comment explains the math.
- `tool_rendering_agent.py` — docstring no longer claims to back the
`tool-rendering-reasoning-chain` cell (it has its own agent file).
TypeScript / TSX demos:
- `_shared/parse-json-result.ts` (new) — extracted from
`tool-rendering/parse-json-result.ts`; now consumed by three demos.
- `_shared/slot-override.ts` (new) — `makeSlotOverride<T>` centralizes
the 11 `as unknown as` casts that `chat-slots/page.tsx` previously
scattered across the slot-override block.
- `shared-state-read` — extracted `RecipeCard` component + `types.ts`,
dropped the dual-state-sync pattern that had the read-only demo
locally mutating recipe state. Page is now a thin shell that publishes
edits via `agent.setState` and reads back via `agent.state.recipe`.
Also surfaces `runAgent` rejections via `console.error` instead of
the previous silent `.catch(() => {})`.
- `headless-complete/hooks/use-auto-scroll.ts`,
`headless-complete/hooks/use-typing-indicator.ts` (new) — extracted
from `chat.tsx`. The chat file shrinks by ~40 lines. Same silent-
rejection fix on `runAgent` as shared-state-read.
- `frontend-tools-async/fake-notes-db.ts` (new) — extracted from
`page.tsx`; the demo file no longer leads with 60 lines of fake-DB
scaffolding before `useFrontendTool` (the actual feature) appears.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Real demo-time bugs in the langgraph-python integration:
- `interrupt-headless` was rendering hardcoded stale slot dates from a
`DEFAULT_SLOTS` constant instead of reading `payload.slots` that the
backend `interrupt(...)` already supplies. Sibling `gen-ui-interrupt`
does this correctly. Headless variant now matches.
- `_shared/interrupt-fallback-slots.ts` (new) — the JS fallback for
when the backend returns no `slots` array. Generates relative to
Date.now() so the picker never shows past dates. Used by both
`interrupt-headless` and `gen-ui-interrupt`. Deletes the old stale-
literal `gen-ui-interrupt/fallback-slots.ts` (dates from April that
had already decayed).
- `interrupt_agent.py` — replaced hardcoded `timezone(timedelta(-7))`
PDT with `zoneinfo.ZoneInfo("America/Los_Angeles")` so the demo
doesn't lie about offsets in winter. Also fixed a Sunday edge case
where `next_monday` collapsed to the same date as `tomorrow` (both
Python and the JS fallback) — added a `<= 1` skip-a-week guard.
- `package.json#dev` ran `uvicorn agent_server:app --port 8000` but
`src/agent_server.py` was a 3-line stub with no `app` symbol AND the
API route pointed at port 8123. Replaced with the canonical
`langgraph_cli dev --port 8123` invocation; deleted the dead stub.
- `entrypoint.sh:47` smoke-checked `src/agents/tools.py` (a file that
never existed) so every Railway boot logged a phantom ERROR. Dropped.
- `reasoning_agent.py` and `tool_rendering_reasoning_chain_agent.py`
intentionally omit `CopilotKitMiddleware` (they exercise only
reasoning-token streaming, no frontend tools / app context). Added a
one-line comment so a future maintainer doesn't cargo-cult it back in.
- `subagents._invoke_sub_agent` text-block walker had a regression
where a `{"type":"text","text":null}` payload (a known provider
quirk) would crash `"".join(parts)` with `TypeError: sequence item
N: expected str instance, NoneType found`. Restored the
`isinstance(block.get("text"), str)` guard.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
## Summary
Four LangGraph Python showcase demos updated:
- **HITL In-app**: `CopilotChat` → `CopilotPopup`
(`defaultOpen={true}`). Tickets panel fills the viewport; chat is a
popup in the corner. Approval dialog still portaled to `<body>`.
- **Shared State: Streaming**: `CopilotChat` (custom aside) →
`CopilotSidebar`. Document view fills the page; sidebar opens by
default.
- **Shared State: Read + Write**: `CopilotPopup` → `CopilotSidebar`.
Card grid breakpoint bumped from `lg:` to `xl:` so cards stack instead
of pinching when the sidebar consumes ~480px on mid-size laptops;
`overflow-y-auto` on the wrapper keeps the page scrollable.
- **Shared State: Read (Recipe)**: visual overhaul with shadcn
primitives (`Card`, `Input`, `Select`, `Textarea`, `Badge`, `Button`,
`Spinner`, `Separator`). All `data-testid` attributes and
QA-doc-asserted strings preserved verbatim ("Make Your Recipe", "AI
Recipe Assistant", "+ Add Ingredient", "+ Add Step", "Improve with AI",
default ingredients/instructions, all 7 dietary preference labels, all 5
cooking-time labels, all 3 suggestions, etc.). React state-sync logic
preserved verbatim (no behavior change).
## Why
The chat surfaces were inconsistent across demos — HITL used a full-pane
chat where the demo's premise is "approval modals appear *outside* the
chat surface", and the shared-state demos mixed `CopilotChat` and
`CopilotPopup` instead of using the prebuilt `CopilotSidebar`. The
Recipe demo was raw Tailwind while the rest of the suite uses shadcn
primitives.
## Plumbing fixes that fell out of the work
- **`clsx` + `tailwind-merge` added to `package.json`** as direct deps.
They were imported by `src/lib/utils.ts` (the `cn` helper used by all
shadcn primitives) but only transitively resolvable, which broke `next
dev --turbopack`'s stricter module resolution.
- **`globals.css` rule**: `body[data-scroll-locked] { padding-right: 0
!important }`. Radix overlays (Select, Dialog, Popover) use
`react-remove-scroll-bar`, which measures the gap between viewport and
body inner width to detect scrollbar size — and our body uses
`margin-inline-end: 480px` to make room for `<CopilotSidebar />`. The
library mis-reads that 480px as scrollbar width and "compensates" with a
matching `padding-right`, which shrinks the content area and shifts the
centered card every time a dropdown opens. We have no real scrollbar
(`body { overflow: hidden }`), so the compensation is unnecessary —
neutralize it. Comment in the diff explains the rationale.
## Test plan
- [x] `next build` clean across all 54 routes
- [x] `oxlint` reports zero new warnings (only pre-existing patterns the
diff preserved verbatim)
- [x] Manually verified `/demos/hitl-in-app`,
`/demos/shared-state-streaming`, `/demos/shared-state-read-write`,
`/demos/shared-state-read` in browser (Chrome, dev server)
- [x] Verified the Recipe page's Select dropdowns no longer shift the
card when opened (the original motivation for the `globals.css` rule)
- [ ] HITL e2e suite (`tests/e2e/hitl-in-app.spec.ts`) — selectors
preserved (`getByPlaceholder("Type a message")`, suggestion-pill
testids, approval-dialog-* testids); CopilotPopup's `defaultOpen={true}`
mounts the chat input on first paint
- [ ] Streaming and Read e2e suites are stale pre-PR (reference content
that doesn't exist in source — "AI Document Editor", "Sales Pipeline")
and remain in the same state
## CR loop
Ran `cr-loop` (Round 1, 7 agents). Reclassified one finding to bucket
(d) per user direction (npm/pnpm lockfile-convention concern is a
different PR's subject — pre-existing across all
`showcase/integrations/*` and out of scope here). Procedure 3 audit
promoted no items to bucket (a); 24 bucket (c) items confirmed as
pre-existing and routed to follow-up backlog.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Switch HITL In-app to CopilotPopup; Shared State Streaming and Read+Write
to CopilotSidebar; widen card-grid breakpoint and add overflow scroll on
Read+Write so content stays usable when the sidebar consumes ~480px of
viewport. Re-skin the read-only Recipe demo with shadcn primitives,
preserving every data-testid and QA-doc-asserted string.
Add clsx and tailwind-merge as direct deps (previously only transitively
resolvable, breaking turbopack dev) and a globals.css rule to neutralize
the Radix scroll-lock padding-right that react-remove-scroll-bar injects
when it mis-detects body's margin-inline-end (CopilotSidebar) as
scrollbar width.
QA docs updated to reference the new chat surfaces.
The tool-rendering-reasoning-chain D5 probe has been red since the
probe was added (PR #4743) — the reasoningMessage slot's
<ReasoningBlock> never mounts because the agent doesn't emit
reasoning summaries.
Compared to the working reasoning_agent.py (which powers
reasoning-default + reasoning-custom and is green), the only
meaningful difference was the reasoning config:
reasoning_agent.py (works):
reasoning={"effort": "medium", "summary": "detailed"}
tools=[]
tool_rendering_reasoning_chain_agent.py (was broken):
reasoning={"effort": "low", "summary": "auto"}
tools=[get_weather, search_flights, ...]
`summary: "auto"` lets the model decide whether to emit a reasoning
summary. With tools present the model often skips it (chain-of-thought
goes straight into the tool call without a summary). `summary:
"detailed"` forces emission on every response, which lights up the
ReasoningBlock slot. Bumping effort medium so the surface visible to
the user reads as a real chain-of-thought, matching reasoning-display.
Probe diagnostics from prod confirmed the toolCalls were firing fine
(WeatherCard mounted, assistant text landed) — only the reasoning
emission was missing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Swap every ChatOpenAI / init_chat_model call across the LGP agents
from the prior mix (gpt-4o-mini, gpt-4o, gpt-4.1, gpt-5-mini) to the
unified `gpt-5.4` model. Also updates the OPENAI_REASONING_MODEL env
default in reasoning_agent.py and tool_rendering_reasoning_chain_agent.py
so reasoning demos pick up the new model unless explicitly overridden.
Verified locally with the user's gpt-5.4 OpenAI access — all 39 LGP
demos render and respond correctly. The probe sweep on prod will
confirm the model name resolves end-to-end.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Lands the Bucket A framework fix and the fixture-correctness changes
needed to flip the remaining D4/D2 cells in the langgraph-python column
to D5. Full local D5 sweep is green (1 passed, 0 failed).
Framework: every Python `ToolMessage` constructed via `Command(update=...)`
now sets `name=` and `id=str(uuid.uuid4())`. Without these, @ag-ui/langgraph
synthesises TOOL_CALL_START events with `toolCallName: null` and
`parentMessageId: null`, which @ag-ui/client@0.0.53's Zod schema rejects;
the rejection is silently swallowed by `withAbortErrorHandling -> EMPTY`,
completing the SSE observable mid-stream so post-tool state never reaches
the consumer. Fix is applied across shared_state_streaming, shared_state_read_write,
gen_ui_agent, beautiful_chat, subagents (2 sites). Single-flag change in
a2ui_dynamic flips the secondary `_design_a2ui_surface` LLM call to
`streaming=True` so aimock's record/replay (SSE-only) sees it.
Fixtures (d5-all.json):
- toolCallId follow-ups for set_steps (3), display_flight, generate_a2ui (4),
schedule_meeting (2), generateSandboxedUi (7), and revenue chart so
multi-turn probes don't recurse into recursion-limit loops
- four hand-crafted secondary `_design_a2ui_surface` fixtures so A2UI
dynamic renders without a real LLM
- mcp-apps fixture rewritten to emit `create_view` tool call with a
minimal Excalidraw element payload; runtime middleware fetches the UI
resource and the iframe mounts
- AAPL and revenue `hasToolResult: true` follow-ups tightened to
`toolCallId` so they don't match cross-turn after prior turns' tool
results
- voice fast-path content-only fixture
- beautiful-chat-schedule-meeting first-turn fixture gains content so
the conversation runner sees an assistant message before the picker
click assertion
Probes: bumped per-card waitForSelector in d5-gen-ui-headless-complete from
15s to 60s — recharts ResponsiveContainer can be slow under 4 sequential
turns.
Shell-dojo: hide CLI Start Command from the dojo navigation via
`HIDDEN_DOJO_FEATURE_IDS`. Registry/manifests untouched so
harness/parity/dashboard still see it.
/langgraph-python/frontend-tools referenced regions
frontend-tool-registration and frontend-tool-handler which were missing
markers on the production demo. Added in-place @region markers on the
existing useFrontendTool block in
showcase/integrations/langgraph-python/src/app/demos/frontend-tools/page.tsx
since the demo is a clean teaching example.
/langgraph-python/custom-look-and-feel/slots referenced regions
register-welcome-slot, register-assistant-message-slot, and
register-disclaimer-slot. The chat-slots production demo registers ~12
slot overrides at once with `as unknown as typeof X` casts that obscure
the per-pattern teaching shape, so added a sibling
slot-overrides.snippet.tsx file (mirrors the llamaindex chat-slots
sibling pattern) with the three minimal teaching regions.
Verified by re-bundling demo-content.json, running shell-docs production
build, and curling both pages on a local server: zero Missing snippet
boxes, and the region code (change_background, CustomWelcomeScreen,
CustomDisclaimer, CustomAssistantMessage) renders on the pages.
Three classes of regression are now pinned:
1. Secondary-LLM tool name doesn't collide with the A2UI middleware's
default intercept list (`render_a2ui`). New
`src/agents/test_a2ui_internal_tools.py` parametrises over
`beautiful_chat._design_a2ui_surface`, `a2ui_dynamic._design_a2ui_surface`,
and `a2ui_fixed.display_flight` and asserts none match the
middleware's `a2uiToolNames` default. Catches accidental rename
reverts that would re-enable the bypass.
2. `generate_a2ui` force-pins the canonical `catalog_id` even when the
secondary LLM hallucinates a wrong one. The new test stubs
`ChatOpenAI` with a fake response carrying a bogus catalogId and
asserts the surface op carries the module's `CUSTOM_CATALOG_ID`.
3. `generate_a2ui` short-circuits with a clean error string when the
LLM emits a root component without a `component` field — never
feeds the renderer the partial tree that surfaced as the "Cannot
create component root without a type" infinite-loop.
7 unit tests, all green locally (`pytest src/agents/test_a2ui_internal_tools.py`).
E2E tests on the same fixes now also assert:
- No `A2UI render error: Catalog not found` banner on the page after
Beautiful Chat → Sales Dashboard, Declarative Gen UI → BarChart, and
A2UI Fixed Schema → Find SFO → JFK round-trips.
- No `Cannot create component … without a type` banner on the same
three pills.
- Exactly ONE flight card on A2UI Fixed Schema (was 6+ on deploy
pre-fix from the `display_flight` loop) — `Flight Details` count
pinned to 1, `Book flight` count pinned to 1.
- At most one ResponsiveContainer on the BarChart pill (loops would
stack multiple).
- At most two ResponsiveContainers on Beautiful Chat → Sales Dashboard
(one pie + one bar = single dashboard render).
Two follow-up fixes layered on the previous internal-tool rename:
(1) `a2ui_fixed.py` — fixed-schema demo infinite loop on deploy. The
`display_flight` tool returns the raw `a2ui.render(...)` JSON descriptor
as its tool result. gpt-4o-mini reads that opaque blob, can't tell the
flight was rendered, and re-calls `display_flight` indefinitely (visible
on the deployed showcase as 6+ duplicate flight cards stacked under
repeated assistant text). Local was just lucky.
Hardened the docstring + system prompt to spell out: the JSON return
value is the surface descriptor, the card is already rendered, do NOT
call again, reply with one short confirmation and stop.
(2) Rename `render_a2ui` → `_design_a2ui_surface` in shared and
langgraph-python parity copies of `tools/generate_a2ui.py` (+
`tools/__init__.py` re-export `RENDER_A2UI_TOOL_SCHEMA` →
`DESIGN_A2UI_SURFACE_TOOL_SCHEMA`), and in `showcase/shared/typescript/
tools/generate-a2ui.ts`. These shared helpers were the source-of-truth
for the secondary-LLM tool name across integrations; renaming here keeps
parity with the langgraph-python agents already renamed in
`beautiful_chat.py` / `a2ui_dynamic.py`. Other framework integrations
keep their own `render_a2ui` for now (separate parity sweep).
(3) `showcase/aimock/feature-parity.json` — added a sibling fixture
matching `toolName: "_design_a2ui_surface"` for the beautiful-chat Sales
Dashboard pill so the langgraph-python e2e suite still hits a
deterministic mock on Railway. The original `render_a2ui` fixture is
kept above it so other integrations whose secondary LLM still requests
`render_a2ui` continue to match.
(4) Comment update in `beautiful-chat.spec.ts` to name the new internal
tool.
The A2UI middleware (`@ag-ui/a2ui-middleware`) defaults `a2uiToolNames` to
`["render_a2ui"]` and synthesises ACTIVITY_SNAPSHOT events from the
streaming tool-call args of any matching call — using the LLM's RAW
catalogId and components verbatim, before the Python tool body has a
chance to validate or normalise.
Both `beautiful_chat.py` and `a2ui_dynamic.py` use a `generate_a2ui` tool
that internally invokes a secondary LLM bound to a structured-output
`render_a2ui` helper. Because the helper's name matched the middleware's
default intercept list, the secondary LLM's hallucinated catalogId
(`declarative-gen-ui-catalog` leaking into the beautiful-chat dashboard)
and malformed root components (no `component` field on KPI dashboard)
were emitted to the frontend bypass, surfacing as:
- "A2UI render error: Catalog not found: declarative-gen-ui-catalog" on
the beautiful-chat Sales Dashboard pill
- "A2UI render error: Cannot create component root without a type"
infinite-loop on the declarative-gen-ui KPI Dashboard pill
The earlier force-pin of `catalog_id` and the defensive component sweep
in `generate_a2ui` were correct but ran too late — they execute on the
tool-result, after the middleware has already fired surface events from
the streaming args.
Fix: rename the internal helper to `_design_a2ui_surface` (and update
`tool_choice`, prompt header, and module docstring) so it falls outside
the middleware's intercept list. The explicit `a2ui.render(...)` ops the
outer `generate_a2ui` returns are then the only path to the frontend,
and our Python validation layer is authoritative.
Verified locally: Beautiful Chat → Sales Dashboard pill renders Total
Revenue / New Customers / etc. with no errors.
## Summary
- Beautiful Chat **Search Flights** pill: restored
`_build_flight_components` so the agent emits a flat literal-children
FlightCard tree (the structural-children template form via
`flight_schema.json` + `update_data_model` doesn't expand correctly
through GenericBinder for our custom catalog — it was working until the
recent flagship-cell pass swapped it back). FlightCards render again
instead of falling through to the default tool card.
- Beautiful Chat **Sales Dashboard** pill: hardened `generate_a2ui` so
the secondary LLM gets explicit catalog-id + component-shape rules and
the resulting `catalog_id` is force-pinned to the registered catalog.
Kills the "Catalog not found: declarative-gen-ui-catalog" sibling-demo
hallucination.
- **Declarative Gen UI** (`a2ui_dynamic.py`): same `generate_a2ui`
hardening, plus a defensive sweep that drops malformed components and
bails clean when the LLM omits a typed root. Kills the "Cannot create
component root without a type" infinite-loop renderer error reported on
the KPI Dashboard pill.
Two files touched (`src/agents/beautiful_chat.py`,
`src/agents/a2ui_dynamic.py`); no frontend / runtime / package changes.
Existing aimock fixtures in `feature-parity.json` already use the
canonical catalogId and well-formed components, so e2e specs should pass
without fixture changes.
## Test plan
- [x] Verified locally: `langgraph_cli dev` + `next dev`, clicked
**Search Flights** on `/demos/beautiful-chat` — Delta DL 405 ($329) and
United UA 120 ($289) FlightCards rendered live with airline, route,
times, status, Select.
- [x] Verified locally: clicked **Show a KPI dashboard** on
`/demos/declarative-gen-ui` — REVENUE / SIGNUPS / CHURN metric tree
rendered cleanly, no "Cannot create component root without a type"
error.
- [ ] CI / `nx run @copilotkit/showcase-langgraph-python:e2e`
(beautiful-chat spec already authored against the literal-children
form).
Three regressions, two files. Without these the Search Flights pill falls
through to the default tool card instead of rendering FlightCards, the
Sales Dashboard pill errors with "Catalog not found:
declarative-gen-ui-catalog", and the Declarative Gen UI cell loops on
"Cannot create component root without a type."
beautiful_chat.py — search_flights:
- Restore _build_flight_components and have search_flights emit the flat
literal-children component tree it produces, instead of the
flight_schema.json structural-children template (Row.children =
{ componentId, path: "/flights" }) plus update_data_model. The
GenericBinder doesn't reliably expand the structural form for our
custom FlightCard catalog — sibling demos avoid the form for the same
reason. The literal form has been the working shape since it was
introduced; the recent flagship-cell pass swapped it back to the
schema/data-model form and broke fixed-schema rendering.
- Make Flight TypedDict permissive (total=False) and drop the required
id / statusIcon fields. langchain rejected calls when the LLM (or
aimock fixture) omitted those, surfacing the validation string as the
tool result and never producing a surface.
beautiful_chat.py + a2ui_dynamic.py — generate_a2ui:
- Prepend a hard-requirements header to the secondary LLM's prompt
pinning the canonical catalogId and the component-shape contract
(every entry — including root — must carry both `id` AND `component`).
With injectA2UITool: false the runtime context alone leaves the LLM
enough room to hallucinate sibling-demo catalog IDs (e.g.
declarative-gen-ui-catalog leaking into the beautiful-chat dashboard)
and to omit `component` on the root entry (the source of the
"Cannot create component root without a type" infinite loop).
- Force catalog_id to the module-level CUSTOM_CATALOG_ID after the tool
call so any residual hallucination still routes to the registered
frontend catalog.
- Drop malformed component entries before constructing the operations
list, and bail with a clean error string if no valid root survives —
fail-soft instead of looping the renderer.
The renderer-side definitions.ts/renderers.tsx changes from the same
flagship pass are kept as-is; their Row/Column children union
(string-array OR { componentId, path }) is forward-compatible and
doesn't hurt the literal-children path.
The voice demo's <CopilotKit> didn't pass enableInspector, so
shouldShowDevConsole(undefined) defaulted to isLocalhost(), which
auto-mounts <cpk-web-inspector> on any local Docker host
(localhost:3100 in showcase compose). The inspector overlay
intercepts pointer events on top of the voice sample-audio button,
so dev/D5 probe runs can't click it through Playwright.
Production isn't localhost, so the inspector never mounts there —
voice is D5 in prod and D4 locally for this reason alone. Set
enableInspector={false} explicitly so the demo behaves the same in
both environments.
Probe result: d5:langgraph-python/voice flips green.
The headless-complete refactor replaced the old 5-chip set
(Weather/AAPL/Highlight/Sketch/Largest continent) with a new 4-chip
set wired through `useConfigureSuggestions`/SuggestionBar:
Weather → "What's the weather in Tokyo?" → WeatherCard
Stock price → "What's the price of AAPL right now?" → StockCard
Highlight a note → "Highlight this note for me: 'ship the demo on Friday'."
→ HighlightNote
Revenue chart → "Show me a chart of revenue over the last six months."
→ ChartCard
Probe rewrite:
- Drop the wrapper-scoped chip selector (`[data-testid=
"headless-suggestions"] >> text=...`) — the new SuggestionBar
doesn't have that container testid. Click via plain `button >>
text="<chip title>"` instead.
- Per-turn assert the matching tool card testid mounts
(headless-weather-card, headless-stock-card, headless-highlight-card,
headless-revenue-chart) plus a distinguishing text token. Driven
off a TURN_EXPECTATIONS table so adding a chip is one entry.
- Drop the readMessagesText helper that scoped to a missing
`[data-testid="headless-complete-messages"]` container; collect
text by concatenating all `headless-message-assistant` bubbles.
Fixture rewrite:
- Drop the Excalidraw + Largest-continent entries (no longer in the
chip set). Weather is already covered by feature-parity.json.
- Add a Revenue chart fixture for `get_revenue_chart` with inline
data so ChartCard renders without backend.
- Stock + Highlight fixtures preserved (still match the new chip
messages via substring).
Also restore `data-message-role` attributes on the headless-complete
UserBubble + AssistantBubble so the runner's settle plateau cascade
can resolve them — matches the runner's documented headless-template
contract.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The headless-simple demo was refactored to a deliberately minimal
"two hooks, one shadcn shell" template — text-in/text-out only, no
gen-UI. The D5 type literal `gen-ui-headless` no longer described
what the probe tests, and the old probe (Profile-card useComponent
+ continent fallback) was asserting against UI that no longer exists.
Three coordinated changes:
1. Rename `gen-ui-headless` D5FeatureType to `headless-simple` so the
slug matches the demo. Updated d5-registry, REGISTRY_TO_D5,
CATALOG_TO_D5_KEY (dashboard), and dependent tests/comments.
`headless-complete` keeps its existing literal because that demo
still drives the full gen-UI surface.
2. Replace d5-gen-ui-headless.{ts,test.ts,fixture} with
d5-headless-simple.ts + headless-simple.json fixture. New probe
clicks the "Say hello in one short sentence." chip and asserts the
`[data-testid="headless-message-assistant"]` bubble mounts with
non-empty content.
3. Restore `data-message-role` attributes on the headless-simple
UserBubble + AssistantBubble. The runner's chat-input cascade
documents these as the headless-template contract; the refactor
dropped them, breaking the runner's settle plateau detection.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The chat-slots refactor replaced the explicit `data-testid="custom-assistant-message"`
attribute with a SlotMarker wrapper component. The probe still asserted
the old testid which no longer exists, so the cell stuck at D4.
Add `data-slot-label={label}` to SlotMarker's outer span — idiomatic
data attribute that mirrors the existing `label` prop and gives the
probe a stable contract without restoring the legacy testid pattern.
The marker was already passing label="MessageView.AssistantMessage";
this just surfaces it in the DOM.
Update d5-chat-slots probe + test to assert
`[data-slot-label="MessageView.AssistantMessage"]`.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two related changes targeting the fc=97 'Target page closed' D5 failure
on `beautiful-chat-toggle-theme`:
1. Probe (`_beautiful-chat-shared.ts:assertToggleTheme`) — replace the
manual 200ms `page.evaluate` poll over 30s with Playwright's
`waitForFunction`. The native polling is event-driven inside the
browser context, disconnects cleanly on page-close, and avoids
~150 round-trip evaluates per probe. The catch branch now checks
`page.isClosed()` first and surfaces a diagnostic message naming
the renderer-crash case explicitly, so operators don't have to
guess what 'Target page closed' meant.
2. Demo (`use-generative-ui-examples.tsx`) — drop `[theme, setTheme]`
from the `useFrontendTool` deps array. The handler reads `document`
directly and the setter is stable across renders, so the deps
array forced a re-registration after every theme flip. That race
could collide with an in-flight tool result and surface as a
renderer error during multi-turn sequences. Removing deps keeps
the tool registration stable for the duration of the conversation.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
_invoke_sub_agent's last-resort branch silently returned '' or a
Python repr like \"[{'type': 'text', ...}]\" for block-list content,
which the UI would render as a blank/garbled card. Return a stable
SUB_AGENT_EMPTY_SENTINEL ('<sub-agent produced no output>') instead
so the d5-subagents probe can match it against its boilerplate-marker
list and fail the genuine-pass test loudly when a sub-agent produces
no usable output.
The get_stock_price tool returned randint-based price/change every
call, so e2e specs asserting $338.37 / -2.96% would only pass when
aimock short-circuited the entire call. The Python tool body still
runs server-side under aimock — only the LLM call is mocked.
Mirror the deterministic-`value` pattern on roll_d20: accept optional
price_usd / change_pct arguments and echo them back when present.
Defaults to random mock data when the args are omitted, preserving
the legacy live-LLM behaviour.
The blitz removed the explicit useDefaultRenderTool() call expecting a
framework-level fallback to handle zero-hook registrations, but the
integration uses the published @copilotkit/react-core@1.56.5 which does
not yet ship that fallback. Without the call, useRenderToolCall has no
'*' renderer and tool calls render invisibly — the user only sees the
agent's final text summary instead of the OOTB tool card.
Restore the explicit useDefaultRenderTool() invocation. The framework
fallback (committed in this PR but inert until react-core publishes a
release that includes it) becomes a no-op once that ships.
Resolved d5-all.json conflict by appending B6's headless-simple/complete fixtures (Say hello, joke, fun fact, Highlight, chart) before the tool-rendering and open-gen-ui entries already on integration.
Adds production-code testids needed by the Phase-2B genuine D5 probes:
- frontend-tools/background.tsx: data-testid='frontend-tools-background'
+ data-background-value mirror so the probe can read the live gradient
off the DOM without computing styles.
- declarative-gen-ui/a2ui/renderers.tsx: data-testid for Card,
StatusBadge, Metric, PieChart, BarChart so per-pill probes can assert
the expected catalog component painted.
- a2ui-fixed-schema/a2ui/renderers.tsx: data-testid='a2ui-fixed-card'
on the Card override so the fixed-schema component-tree mount is
observable without keyword-matching the transcript.
Each cell's Layer 1 spec is already green; this is purely additive — no
spec rewrites.
Add stable testids and rendering surfaces for the three tool-rendering
cells so the e2e suite can distinguish each cell's strategy:
- tool-rendering: register useRenderTool for get_stock_price and
roll_d20; new StockCard / D20Card components with testids
stock-card / d20-card / stock-price / stock-change / d20-value.
Rename FlightListCard testid flight-list-card -> flights-card.
- tool-rendering-default-catchall: drop the custom shadcn
useDefaultRenderTool registration so the cell is truly zero
custom-render-hooks. The framework's built-in
DefaultToolCallRenderer now paints every tool call, with stable
data-testid='copilot-tool-render' wrapper plus data-tool-name,
data-args, and data-result attributes for inspection without
expanding the card.
- tool-rendering-custom-catchall: rename the wildcard renderer's
testids from custom-catchall-* to custom-wildcard-* so the cell
is distinguishable from the (now-OOTB) default-catchall demo.
- packages/react-core: when no per-tool / wildcard renderer is
registered, useRenderToolCall now falls back to the built-in
DefaultToolCallRenderer instead of returning null.
Replace the random roll_dice tool with roll_d20(value), which echoes
the LLM-supplied value back as the result. Aimock fixtures script the
five sequential calls returning [7, 14, 3, 19, 20] so the e2e suite can
assert exact values rather than rolling until 20 lands.
Update SYSTEM_PROMPT to allow multi-tool chaining when the user
explicitly asks for it (Chain tools pill emits 3 tool calls in one
turn).
Drop the 5 cross-origin contentFrame() / page.on('console', ...) skipped
assertions across the two specs — sandbox=allow-scripts only blocks host
introspection of the iframe DOM, and console-spying on the host page
catches no inner-iframe logs. Replace with iframe-presence assertions:
each pill click must produce iframe[sandbox*='allow-scripts'] with a
non-empty srcdoc (or src) attribute. That is the load-bearing signal
that the open-generative-ui pipeline mounted SOMETHING.
Rewrite each suggestion message string as a short verbatim label that
doubles as a deterministic aimock fixture key (paired with the new
fixtures in showcase/aimock/d5-all.json). Drop pill-title parentheticals
per the cosmetic note in lgp-test-genuine-pass.md so titles read as
natural human prompts; keep the message field aligned with the fixture
key.
Final test counts: 5 minimal (page-load + 4 pill-iframe), 4 advanced
(page-load + 3 pill-iframe). All .skip() removed. The sandbox-function
round-trip (evaluateExpression / notifyHost) is intentionally not
asserted here — that requires a same-origin sandbox option or a
host-side spy on the runtime's sandbox-function-call event, both
deferred to a follow-up.
- subagent-activity-card: emit data-testid="subagent-card-<role>" on
each card wrapper (researcher | writer | critic), data-testid=
"subagent-result" on the Result content div, and data-testid=
"subagent-status" on the status pill. The previous testids
(subagent-activity-card, subagent-activity-result) collapsed across
roles, so the e2e suite couldn't count or content-assert per role.
- delegation-log: render a fixed row of 3 always-visible role
indicators (data-testid="subagent-indicator-<role>") so the page
exposes a stable hook for the load-state assertion regardless of
whether the supervisor has delegated yet.
- Annotate AgentState.delegations with operator.add reducer so concurrent
sub-agent emissions in one supervisor step accumulate instead of
raising INVALID_CONCURRENT_GRAPH_UPDATE (HTTP 400 on the Summarize pill).
- Update _delegation_update to return only the new entry (the reducer
concatenates) instead of echoing the full prior list, which would
duplicate entries each step under operator.add.
- Cap supervisor -> critique_agent loop at _MAX_CRITIQUE_ITERATIONS
(default 1). Re-entrant critique calls short-circuit with a finish-now
ToolMessage and do not append a second delegation, so the UI shows
exactly one critic card per supervisor run.
- _invoke_sub_agent now walks messages newest-first and returns the
first non-empty AIMessage content (handles list-of-content-blocks
shape too). Prevents the previous failure mode where a final empty
AIMessage made the card Result blank or echoed the showcase-assistant
intro.
- Strengthen supervisor system prompt: each sub-agent must be called
exactly once, with no further calls after critique returns.
Adds stable data-testid hooks on the hand-rolled headless chat surface
shared by headless-simple and headless-complete so e2e specs can target
the headless surface (and not the default CopilotChat surface) by
selector.
Shared:
- 'headless-message-assistant' on the custom assistant bubble
- 'headless-message-user' on the custom user bubble
- 'headless-composer' on the composer container
headless-complete only:
- 'headless-weather-card' on the WeatherCard rendered via useRenderTool
- 'headless-stock-card' on the StockCard
- 'headless-highlight-card' on the HighlightNote rendered via useComponent
- 'headless-revenue-chart' on the ChartCard
If the headless surface ever silently regresses to the default
CopilotChat surface, the headless-specific testids are absent and the
spec fails. Each tool-card testid is scoped per-component so a regression
in a single render hook fails only that test.
Tyler Slaton