feat(posthog-ai): support pi task sessions on the web

[puemos]

Problem

• A user who opens a Pi task in PostHog web sees only "Pi session logs aren't available in PostHog yet". Pi sessions work only in PostHog Desktop.
• The Pi agent server writes its own log format and takes its own commands. The web reads ACP, the format of the Claude and Codex agent server.

Changes

With the pi-acp flag on, a Pi task on /tasks and in the inbox works like any other task:

• The thread shows Pi messages, thoughts, tool cards, notices, errors and turn ends.
• A user can reply, steer with "Send now", queue a follow-up, stop the run, and run the task again.
• The composer shows every model in one list, plus the reasoning picker. It has no harness or permission mode.
• MCP approvals and Pi extension prompts (select, confirm, input, editor) use the existing approval and question cards.
• A resumed run shows the earlier runs of the session, also after a reload.
• /usage and the context ring show Pi token usage. Attachments show as files.

With the flag off, the banner stays as it is today.

Demo

Recorded against the running app, with a real Pi agent and model:

Recording	Shows
01 – start and tools	A new task, tool cards, an edit diff, the context ring
02 – follow-up, model, queue, stop	A model change, "Send now", a queued message, Stop
03 – extension prompts	All four prompt kinds, a timeout, an extension failure
04 – attachment	A file the agent reads
05 – reload and resume	The thread after a reload, and a resumed run with its history

The links open the .webm files. GitHub shows no inline player for them.

Screenshots

Tool calls, with the edit diff and a command's output open:

Every extension prompt kind, a timed-out prompt, and an extension failure:

An attachment the agent reads:

How it works

The Pi agent server now speaks ACP, so the web reads a Pi run through its existing ACP path.

Before:

flowchart LR
    P{{Pi}} --> S[Pi agent server] --> L[(Pi-format log)]
    L --> C[Desktop]
    L -.-> W[Web: banner only]
    classDef phBlue fill:#1d4aff,stroke:#1d4aff,color:#fff;
    classDef phGray fill:#e5e7eb,stroke:#c7ccd1,color:#000;
    classDef phYellow fill:#f9bd2b,stroke:#f9bd2b,color:#000;
    class P,S phBlue;
    class L phGray;
    class C,W phYellow;

Loading

After:

flowchart LR
    P{{Pi}} --> S[Pi agent server]
    S --> F{pi-acp on?}
    F -- yes --> A[(ACP log)]
    F -- no --> L[(Pi-format log)]
    A --> W[Web run surface]
    A --> C[Desktop]
    L --> C
    L --> W
    classDef phBlue fill:#1d4aff,stroke:#1d4aff,color:#fff;
    classDef phGray fill:#e5e7eb,stroke:#c7ccd1,color:#000;
    classDef phYellow fill:#f9bd2b,stroke:#f9bd2b,color:#000;
    class P,S,F phBlue;
    class A,L phGray;
    class C,W phYellow;

Loading

• The backend checks the flag for the run creator and tells the sandbox which format to write.
• With ACP, the server also accepts the ACP commands the web sends: permission_response and set_config_option.
• @posthog/agent-contracts holds the mapping between Pi events and ACP, in both directions. The Pi server, Desktop and the web all use it.
• Desktop and the web read both formats. The web shows an old Pi-format run read-only, without its prompts.
• Every Pi tool call now carries its tool name in _meta.posthog, so renderers never guess it.
• A Pi run finds its earlier runs through its task session, not resume_from_run_id.

Why in the agent server: the community Pi adapters for ACP editors do the same, for example regadas/pi-acp. Translating on the web or in Django would copy the Pi semantics into a second place.

Mechanical: regenerated OpenAPI types and MCP schemas, kea typegen, and the new setting threaded through the sandbox launch path.

Note

PostHog AI chat stays closed for Pi tasks. The backend rejects them there on purpose.

Flow details

Starting a Pi run

sequenceDiagram
    participant U as Web or Desktop
    participant D as Django
    participant T as Temporal worker
    participant S as Pi agent server
    participant P as Pi
    participant L as Run log and live stream
    U->>D: POST run for a Pi task
    D->>T: start the task workflow
    T->>T: evaluate pi-acp for the run creator
    T->>S: launch with POSTHOG_PI_CONVERSATION_FORMAT acp or pi
    S->>P: start pi in RPC mode
    S->>L: run start, as _posthog/run_started or pi_run_started
    loop every Pi event
        P-->>S: RPC event
        S->>S: translate it to a conversation event
        alt format acp
            S->>S: map it with agentConversationEventToAcpNotification
        end
        S->>L: append to the log in batches, and to the live stream at once
    end

Loading

Sending a follow-up with a new model

sequenceDiagram
    participant W as Web composer
    participant D as Django command relay
    participant S as Pi agent server
    participant P as Pi
    W->>D: set_config_option model
    D->>S: forward
    S->>P: set_model
    W->>D: user_message, as a steer when the run is busy
    D->>S: forward
    S->>P: prompt or steer
    S-->>W: ACP notifications, _posthog/turn_complete, _posthog/usage_update

Loading

Answering an MCP approval or an extension prompt

sequenceDiagram
    participant P as Pi or Pi extension
    participant S as Pi agent server
    participant L as Run log and live stream
    participant W as Web
    P->>S: approval or extension_ui_request
    S->>S: keep the open prompt and start its timeout
    S->>L: _posthog/permission_request
    L-->>W: the approval or question card replaces the composer
    alt answered in time
        W->>S: permission_response
        S->>P: decision or extension_ui_response
        S->>L: _posthog/permission_resolved
    else the timeout passes first
        S->>L: _posthog/permission_resolved, and the card closes
    end

Loading

Pi source	Entry written with the flag on
Conversation events	session/update, or _posthog/user_message, _posthog/progress, _posthog/status, _posthog/error, _posthog/turn_complete
Run start	_posthog/run_started
Turn usage	_posthog/usage_update, and a usage_update session update
MCP approval or extension prompt	_posthog/permission_request, then _posthog/permission_resolved
Extension notify and extension_error	_posthog/status with extension_notice, or _posthog/console
Other extension UI	_posthog/pi_extension_event

Every entry from a Pi extension message keeps the raw message in _meta.piExtension, so Desktop can rebuild its own extension UI.

How did you test this code?

Test rationale: the new tests guard the Pi-to-ACP mapping, the new commands, and the web reader. Each catches a regression no existing test covers.

• acp-conversation.test.ts: every Pi event survives the trip to ACP and back.
• acp-wire.test.ts: each Pi log entry and extension message maps to the right notification, and each prompt kind gets the right answer.
• pi-agent-server.test.ts: the ACP log, approvals and prompts through permission_response, prompt timeouts, and set_config_option.
• cloudPiSessionClient.test.ts (Desktop): both formats read the same.
• test_api.py, test_modal_sandbox.py: Pi run fields, the session logs, the command relay, and the sandbox format setting.
• Web: runStreamLogic, runInteractionLogic, toolResolver, composer and question tests, and RunSurfaceImpl for the flag-off banner.

End to end, on the local stack with a Docker sandbox running this branch and a real model through the LLM gateway. The recordings above come from this run. It found four bugs, all fixed here:

• The reply text was lost after a reload.
• The command relay rejected permission_response and set_config_option for Pi.
• The quill thread showed an extension notice without its text.
• A run-state write after a model change had no effect. It is removed.

Not tested:

• A Pi sandbox on Modal.
• Desktop against an ACP-format run. Unit tests cover its reader.
• An MCP approval end to end. The local stack had no MCP server.
• The flag-off banner end to end. RunSurfaceImpl.test.tsx covers it.
• agent-server.test.ts (ACP), because the sandbox blocks git commit. The ACP server change is only the shared attachment path helper.

Known gaps, not in this PR:

• Stop shows "Agent error: This operation was aborted" and leaves the running tool spinning. This predates this PR and affects Desktop too.
• Pi slash commands are not in the composer menu.
• A Pi command row reads "Terminal", because Pi bash calls carry no description.

Release status

• This change is behind a feature flag and is not available to users

The pi-acp flag exists in PostHog. It is off, at 0%.

Warning

The flag switches the Pi server to ACP for every run its users start, also from Desktop. A Desktop build older than this PR shows those runs as empty. Roll out beyond internal users only after a Desktop release with this PR.

Desktop and backend in one PR

This PR carries desktop-skip-backend-check. Either half can ship first:

• An agent server without this PR ignores POSTHOG_PI_CONVERSATION_FORMAT and keeps writing the Pi format.
• A backend without this PR never sets it, so the new agent server writes the Pi format too.
• Desktop and the web read both formats, so neither depends on the backend deploy.

Docs update

None. No existing doc covers Pi tasks on the web.

🤖 Agent context

Autonomy: Human-driven (agent-assisted)

Agent: PostHog Desktop (Claude Code), Claude Opus 5.5 (claude-opus-5-5), with one session on Claude Fable 5.1 (claude-fable-5-1)

• The first rounds translated the Pi format in the web client. After review, the translation moved into the Pi agent server, with the mapping shared through @posthog/agent-contracts.
• Skills: /writing-tests, /improving-drf-endpoints, /writing-ui-components, /writing-code-comments, /writing-user-facing-copy, /writing-pr-descriptions. The local CodeRabbit pass was skipped in this cloud run.
• All story and test data is invented. The end-to-end run used a throwaway test extension that is not part of this PR.

🤖 Generated with Claude Code

---

Created with PostHog Desktop

[trunk-io]

😎 Merged successfully - details.

[github-actions]

🤖 CI report

⚠️ Trunk lane — backend Python lane

This PR is assigned to the backend Python lane. It runs backend Python tests and may merge in parallel with PRs in other lanes.

⚠️ Complexity (TypeScript) — 62 functions above the limit (max 126)

Cyclomatic complexity above the limit in changed typescript files (10 for production files, 15 for test files). Warn only: worth simplifying when you next touch these functions.

Function	Location	Complexity	Limit
foldLogFromCheckpoint	products/posthog_ai/frontend/logics/runStreamLogic.ts:1677	126	10
ingestAcpFrame	products/posthog_ai/frontend/logics/runStreamLogic.ts:4936	96	10
streamConversation	frontend/src/scenes/max/maxThreadLogic.tsx:1261	73	10
onEventImplementation	frontend/src/scenes/max/maxThreadLogic.tsx:3249	53	10
<anonymous>	frontend/src/scenes/max/maxThreadLogic.tsx:2973	47	10
askMax	frontend/src/scenes/max/maxThreadLogic.tsx:1979	45	10
sendNow	products/posthog_ai/frontend/logics/runInteractionLogic.ts:1324	45	10
handleSseEvent	products/posthog_ai/frontend/logics/runStreamLogic.ts:4172	43	10
getMcpToolPresentation	products/posthog_ai/frontend/components/tool/GenericMcpToolRenderer.tsx:34	38	10
reconcileRunLog	products/posthog_ai/frontend/logics/runStreamLogic.ts:1376	32	10
TaskRunComposer	products/posthog_ai/frontend/scenes/TaskTracker/components/TaskRunComposer.tsx:48	32	10
ComposerModelEffortPickers	products/posthog_ai/frontend/components/composer/ComposerModelEffortPickers.tsx:142	30	10
TaskDetailPage	products/posthog_ai/frontend/scenes/TaskTracker/components/TaskDetailPage.tsx:27	30	10
reconcileHistory	products/posthog_ai/frontend/logics/runStreamLogic.ts:3890	29	10
startNewRun	products/posthog_ai/frontend/logics/runInteractionLogic.ts:1529	24	10
invocationFromToolCallUpdate	products/posthog_ai/frontend/logics/runStreamLogic.ts:1494	24	10
dedupeBufferedAgainstHistory	products/posthog_ai/frontend/logics/runStreamLogic.ts:1124	23	10
streamRun	products/posthog_ai/frontend/logics/runStreamLogic.ts:4298	21	10
OptionSelector	products/posthog_ai/frontend/components/OptionSelector.tsx:35	19	10
BuiltinToolRenderer	products/posthog_ai/frontend/components/tool/builtinToolRenderers.tsx:467	19	10
<anonymous>	frontend/src/scenes/max/maxThreadLogic.tsx:2639	18	10
enhanceThreadToolCalls	frontend/src/scenes/max/maxThreadLogic.tsx:3146	18	10
MultiFieldInput	products/posthog_ai/frontend/components/QuestionField.tsx:327	18	10
ExitPlanModeRenderer	products/posthog_ai/frontend/components/tool/builtinToolRenderers.tsx:131	18	10
parsePermissionRequestFrame	products/posthog_ai/frontend/logics/runStreamLogic.ts:871	18	10
deliverPermission	products/posthog_ai/frontend/logics/runStreamLogic.ts:4569	17	10
add	products/posthog_ai/frontend/logics/runStreamLogic.ts:1240	16	10
openSseForRun	products/posthog_ai/frontend/logics/runStreamLogic.ts:4135	16	10
handleTerminalStatus	products/posthog_ai/frontend/logics/runStreamLogic.ts:4685	16	10
coversEntry	products/posthog_ai/frontend/logics/runStreamLogic.ts:1209	14	10
<anonymous>	products/posthog_ai/frontend/logics/runStreamLogic.ts:3681	14	10
sseDropped	products/posthog_ai/frontend/logics/runStreamLogic.ts:4428	14	10
TaskRunChat	products/posthog_ai/frontend/scenes/TaskTracker/components/TaskRunChat.tsx:50	14	10
<anonymous>	frontend/src/scenes/max/maxThreadLogic.tsx:3216	13	10
RunSurfaceRoot	products/posthog_ai/frontend/components/RunSurfaceImpl.tsx:91	13	10
RunSurfaceComposer	products/posthog_ai/frontend/components/RunSurfaceImpl.tsx:281	13	10
SubagentToolRenderer	products/posthog_ai/frontend/components/tool/builtinToolRenderers.tsx:280	13	10
cancelWhenReady	products/posthog_ai/frontend/logics/runCancellationLogic.ts:132	13	10
submit	products/posthog_ai/frontend/logics/runInteractionLogic.ts:923	13	10
userAttachment	products/posthog_ai/frontend/logics/runStreamLogic.ts:434	13	10
readLegacyPiEntry	products/posthog_ai/frontend/logics/runStreamLogic.ts:502	13	10
sameFoldOptions	products/posthog_ai/frontend/logics/runStreamLogic.ts:1635	13	10
bootstrapRun	products/posthog_ai/frontend/logics/runStreamLogic.ts:4040	13	10
extractAgentToolName	products/posthog_ai/frontend/utils/toolResolver.ts:40	13	10
extractMcpProxyCall	products/posthog_ai/frontend/utils/toolResolver.ts:58	13	10
resolveToolKey	products/posthog_ai/frontend/utils/toolResolver.ts:81	13	10
enqueueQueuedMessage	frontend/src/scenes/max/maxThreadLogic.tsx:1841	12	10
executePendingClientToolCall	frontend/src/scenes/max/maxThreadLogic.tsx:2408	12	10
<anonymous>	products/posthog_ai/frontend/logics/runInteractionLogic.ts:1102	12	10
mergeToolCallUpdateEntries	products/posthog_ai/frontend/logics/runStreamLogic.ts:1001	12	10
renderLiveHuman	products/posthog_ai/frontend/logics/runStreamLogic.ts:1921	12	10
routePermissionRequest	products/posthog_ai/frontend/logics/runStreamLogic.ts:4495	12	10
extractSandboxQuestionAnswer	products/posthog_ai/frontend/policy/questionUtils.ts:133	12	10
setConversation	frontend/src/scenes/max/maxThreadLogic.tsx:1718	11	10
handleKeyDown	products/posthog_ai/frontend/components/OptionSelector.tsx:63	11	10
RunSurfaceThread	products/posthog_ai/frontend/components/RunSurfaceImpl.tsx:214	11	10
parsePermissionOption	products/posthog_ai/frontend/logics/runStreamLogic.ts:836	11	10
<anonymous>	products/posthog_ai/frontend/logics/runStreamLogic.ts:1349	11	10
finalizeMessage	products/posthog_ai/frontend/logics/runStreamLogic.ts:1845	11	10
pushError	products/posthog_ai/frontend/logics/runStreamLogic.ts:1995	11	10
<anonymous>	products/posthog_ai/frontend/logics/runStreamLogic.ts:3543	11	10
<anonymous>	products/posthog_ai/frontend/logics/runStreamLogic.ts:3644	11	10

⚠️ Duplication (Python) — 5 new duplicated blocks (worst 284 tokens)

New Python code duplication introduced by this branch. Fails at 70+ tokens in app code, or 150+ tokens when both copies live in test files. Advisory while the gate proves itself: extract a shared helper instead of copying.

First copy	Second copy	Lines	Tokens
products/tasks/backend/logic/services/agent_server_launcher.py:675	products/tasks/backend/logic/services/sandbox.py:653	41	284
products/tasks/backend/logic/services/agent_server_launcher.py:306	products/tasks/backend/logic/services/docker_sandbox.py:949	39	252
products/tasks/backend/presentation/serializers.py:3677	products/tasks/backend/presentation/serializers.py:3888	26	95
products/tasks/backend/logic/services/agent_server_launcher.py:345	products/tasks/backend/logic/services/docker_sandbox.py:992	21	86
products/tasks/backend/presentation/serializers.py:3653	products/tasks/backend/presentation/serializers.py:3867	22	82

⚠️ Duplication (TypeScript) — 1 new duplicated block (worst 172 tokens)

New TypeScript code duplication introduced by this branch. Fails at 70+ tokens in app code, or 150+ tokens when both copies live in test files. Advisory while the gate proves itself: extract a shared helper instead of copying.

First copy	Second copy	Lines	Tokens
products/posthog_ai/frontend/components/ReadonlyRunSurfaceImpl.test.tsx:22	products/posthog_ai/frontend/components/RunSurfaceImpl.test.tsx:29	11	172

⚠️ Bundle size — 🔺 +5.9 KiB (+0.0%)

Uncompressed size of every built .js bundle, compared against the base branch.

Total: 75.81 MiB · 🔺 +5.9 KiB (+0.0%)

File	Size	Δ vs base
render-query/src/render-query/render-query.js	24.38 MiB	🔺 +6.0 KiB (+0.0%)

_{Posted automatically by build-bundle-size-report · uncompressed bytes from dist-report}

✅ Eager graph — within budget

How much code each root ships on the eager path — downloaded and parsed before the surface is interactive. Measured from the esbuild output chunks (post-tree-shake, static imports only); lazy import() / React.lazy chunks are not counted.

Root	Eager (shipped)	Δ vs base	Budget
entry (logged-out pages, app bootstrap) src/index.tsx	1003.2 KiB · 23 files	no change	█████████░ 85.6% of 1.14 MiB
logged-out boot: index + App + bootApp (preloaded by every page, including /login) src/index.tsx + src/scenes/App.tsx + src/scenes/bootApp.ts	2.86 MiB · 604 files	🔺 +16 B (+0.0%)	█████████░ 94.3% of 3.03 MiB
authenticated shell (every logged-in page) src/scenes/AuthenticatedShell.tsx	7.29 MiB · 2,515 files	🔺 +4.5 KiB (+0.1%)	██████████ 95.4% of 7.64 MiB
dashboard scene src/scenes/dashboard/Dashboard.tsx	9.54 MiB · 3,599 files	🔺 +4.5 KiB (+0.0%)	█████████░ 92.6% of 10.30 MiB
today home path src/scenes/AuthenticatedShell.tsx + src/scenes/project-homepage/ProjectHomepage.tsx + src/scenes/project-homepage/today/TodayHome.tsx	7.32 MiB · 2,525 files	🔺 +4.5 KiB (+0.1%)	█████████░ 92.8% of 7.89 MiB
events scene src/scenes/activity/explore/EventsScene.tsx	9.13 MiB · 3,434 files	🔺 +4.5 KiB (+0.0%)	█████████░ 92.4% of 9.88 MiB
replay detail scene src/scenes/session-recordings/detail/SessionRecordingDetail.tsx	11.92 MiB · 4,276 files	🔺 +4.5 KiB (+0.0%)	████████░░ 79.2% of 15.05 MiB

🟢 node_modules/monaco-editor/ stays out of src/index.tsx
🟢 src/lib/components/ActivityLog/describers stays out of src/index.tsx
🟢 node_modules/@posthog/brand/dist/generated/hoggies/svg/ stays out of src/index.tsx
🟢 node_modules/@posthog/brand/dist/generated/hoggies/components/ stays out of src/index.tsx
🟢 node_modules/monaco-editor/ stays out of src/index.tsx + src/scenes/App.tsx + src/scenes/bootApp.ts
🟢 src/layout/navigation-3000/navigationLogic.tsx stays out of src/index.tsx + src/scenes/App.tsx + src/scenes/bootApp.ts
🟢 src/scenes/dashboard/dashboardLogic.tsx stays out of src/index.tsx + src/scenes/App.tsx + src/scenes/bootApp.ts
🟢 src/lib/lemon-ui/LemonMarkdown/ stays out of src/index.tsx + src/scenes/App.tsx + src/scenes/bootApp.ts
🟢 src/lib/components/RichContentEditor/ stays out of src/index.tsx + src/scenes/App.tsx + src/scenes/bootApp.ts
🟢 src/lib/components/CodeSnippet/ stays out of src/index.tsx + src/scenes/App.tsx + src/scenes/bootApp.ts
🟢 src/taxonomy/core-filter-definitions-by-group.json stays out of src/index.tsx + src/scenes/App.tsx + src/scenes/bootApp.ts
🟢 node_modules/zod/ stays out of src/index.tsx + src/scenes/App.tsx + src/scenes/bootApp.ts
🟢 src/scenes/billing/billingLogic.tsx stays out of src/index.tsx + src/scenes/App.tsx + src/scenes/bootApp.ts
🟢 src/lib/components/ProductEmptyState/productSetupPreloadLogic.ts stays out of src/index.tsx + src/scenes/App.tsx + src/scenes/bootApp.ts
🟢 node_modules/monaco-editor/ stays out of src/scenes/AuthenticatedShell.tsx
🟢 src/lib/components/ActivityLog/describers stays out of src/scenes/AuthenticatedShell.tsx
🟢 products/dashboards/frontend/widgets/previews/ stays out of src/scenes/AuthenticatedShell.tsx
🟢 src/scenes/session-recordings/player/sessionRecordingPlayerLogic.ts stays out of src/scenes/AuthenticatedShell.tsx
🟢 zod/v4/locales/de.js stays out of src/scenes/AuthenticatedShell.tsx
🟢 node_modules/@posthog/brand/dist/generated/hoggies/svg/ stays out of src/scenes/AuthenticatedShell.tsx
🟢 node_modules/@posthog/brand/dist/generated/hoggies/components/ stays out of src/scenes/AuthenticatedShell.tsx
🟢 src/scenes/session-recordings/playlist/SessionRecordingsPlaylist.tsx stays out of src/scenes/dashboard/Dashboard.tsx
🟢 src/scenes/web-analytics/tiles/WebAnalyticsTile.tsx stays out of src/scenes/dashboard/Dashboard.tsx
🟢 src/scenes/project-homepage/ai-first/AiFirstHomepage.tsx stays out of src/scenes/AuthenticatedShell.tsx + src/scenes/project-homepage/ProjectHomepage.tsx + src/scenes/project-homepage/today/TodayHome.tsx
🟢 src/scenes/project-homepage/today/TodayReportPage.tsx stays out of src/scenes/AuthenticatedShell.tsx + src/scenes/project-homepage/ProjectHomepage.tsx + src/scenes/project-homepage/today/TodayHome.tsx
🟢 src/queries/Query/Query.tsx stays out of src/scenes/AuthenticatedShell.tsx + src/scenes/project-homepage/ProjectHomepage.tsx + src/scenes/project-homepage/today/TodayHome.tsx
🟢 src/scenes/session-recordings/playlist/SessionRecordingsPlaylist.tsx stays out of src/scenes/activity/explore/EventsScene.tsx
🟢 src/scenes/web-analytics/tiles/WebAnalyticsTile.tsx stays out of src/scenes/activity/explore/EventsScene.tsx

Largest files eagerly shipped from src/index.tsx

Size	File
126.8 KiB	../node_modules/.pnpm/react-dom@18.3.1_react@18.3.1/node_modules/react-dom/cjs/react-dom.production.min.js
24.6 KiB	../node_modules/.pnpm/buffer@6.0.3/node_modules/buffer/index.js
6.3 KiB	../node_modules/.pnpm/react@18.3.1/node_modules/react/cjs/react.production.min.js
4.5 KiB	../node_modules/.pnpm/@jspm+core@2.1.0/node_modules/@jspm/core/nodelibs/browser/process.js
3.9 KiB	../node_modules/.pnpm/scheduler@0.23.2/node_modules/scheduler/cjs/scheduler.production.min.js
1.4 KiB	../node_modules/.pnpm/base64-js@1.5.1/node_modules/base64-js/index.js
1.3 KiB	src/RootErrorBoundary.tsx
1.3 KiB	src/index.tsx
912 B	../node_modules/.pnpm/ieee754@1.2.1/node_modules/ieee754/index.js
839 B	src/scenes/ChunkLoadErrorBoundary.tsx

Largest files eagerly shipped from src/index.tsx + src/scenes/App.tsx + src/scenes/bootApp.ts

Size	File
319.8 KiB	../node_modules/.pnpm/posthog-js@1.438.8_@types+react@18.3.27_react@18.3.1/node_modules/posthog-js/dist/module.mjs
220.8 KiB	../node_modules/.pnpm/@posthog+icons@0.38.0_react-dom@18.3.1_react@18.3.1__react@18.3.1/node_modules/@posthog/icons/dist/posthog-icons.es.js
126.8 KiB	../node_modules/.pnpm/react-dom@18.3.1_react@18.3.1/node_modules/react-dom/cjs/react-dom.production.min.js
103.1 KiB	src/products.tsx
99.6 KiB	src/lib/api.ts
69.0 KiB	src/lib/lemon-ui/icons/icons.tsx
38.7 KiB	../node_modules/.pnpm/@dnd-kit+core@6.0.8_react-dom@18.3.1_react@18.3.1__react@18.3.1/node_modules/@dnd-kit/core/dist/core.esm.js
33.9 KiB	../node_modules/.pnpm/kea@4.0.0-pre.6_patch_hash=139b8d1f1304f9d9da452a9a1244c94ea679dbcb85687d8999563146879fb6f5_react@18.3.1/node_modules/kea/lib/index.cjs.js
28.6 KiB	src/scenes/scenes.ts
27.7 KiB	src/lib/constants.tsx

Largest files eagerly shipped from src/scenes/AuthenticatedShell.tsx

Size	File
319.8 KiB	../node_modules/.pnpm/posthog-js@1.438.8_@types+react@18.3.27_react@18.3.1/node_modules/posthog-js/dist/module.mjs
280.8 KiB	src/taxonomy/core-filter-definitions-by-group.json
220.8 KiB	../node_modules/.pnpm/@posthog+icons@0.38.0_react-dom@18.3.1_react@18.3.1__react@18.3.1/node_modules/@posthog/icons/dist/posthog-icons.es.js
153.7 KiB	../node_modules/.pnpm/re2js@0.4.1/node_modules/re2js/build/index.esm.js
126.8 KiB	../node_modules/.pnpm/react-dom@18.3.1_react@18.3.1/node_modules/react-dom/cjs/react-dom.production.min.js
113.0 KiB	../packages/quill/packages/quill/dist/index.js
103.1 KiB	src/products.tsx
99.6 KiB	src/lib/api.ts
93.3 KiB	../node_modules/.pnpm/prosemirror-view@1.40.1/node_modules/prosemirror-view/dist/index.js
90.6 KiB	../node_modules/.pnpm/@tiptap+core@3.20.6_@tiptap+pm@3.20.6/node_modules/@tiptap/core/dist/index.js

Largest files eagerly shipped from src/scenes/dashboard/Dashboard.tsx

Size	File
319.8 KiB	../node_modules/.pnpm/posthog-js@1.438.8_@types+react@18.3.27_react@18.3.1/node_modules/posthog-js/dist/module.mjs
280.8 KiB	src/taxonomy/core-filter-definitions-by-group.json
220.8 KiB	../node_modules/.pnpm/@posthog+icons@0.38.0_react-dom@18.3.1_react@18.3.1__react@18.3.1/node_modules/@posthog/icons/dist/posthog-icons.es.js
181.9 KiB	src/queries/validators.js
153.7 KiB	../node_modules/.pnpm/re2js@0.4.1/node_modules/re2js/build/index.esm.js
126.8 KiB	../node_modules/.pnpm/react-dom@18.3.1_react@18.3.1/node_modules/react-dom/cjs/react-dom.production.min.js
113.0 KiB	../packages/quill/packages/quill/dist/index.js
103.1 KiB	src/products.tsx
99.6 KiB	src/lib/api.ts
93.3 KiB	../node_modules/.pnpm/prosemirror-view@1.40.1/node_modules/prosemirror-view/dist/index.js

Largest files eagerly shipped from src/scenes/AuthenticatedShell.tsx + src/scenes/project-homepage/ProjectHomepage.tsx + src/scenes/project-homepage/today/TodayHome.tsx

Size	File
319.8 KiB	../node_modules/.pnpm/posthog-js@1.438.8_@types+react@18.3.27_react@18.3.1/node_modules/posthog-js/dist/module.mjs
280.8 KiB	src/taxonomy/core-filter-definitions-by-group.json
220.8 KiB	../node_modules/.pnpm/@posthog+icons@0.38.0_react-dom@18.3.1_react@18.3.1__react@18.3.1/node_modules/@posthog/icons/dist/posthog-icons.es.js
153.7 KiB	../node_modules/.pnpm/re2js@0.4.1/node_modules/re2js/build/index.esm.js
126.8 KiB	../node_modules/.pnpm/react-dom@18.3.1_react@18.3.1/node_modules/react-dom/cjs/react-dom.production.min.js
113.0 KiB	../packages/quill/packages/quill/dist/index.js
103.1 KiB	src/products.tsx
99.6 KiB	src/lib/api.ts
93.3 KiB	../node_modules/.pnpm/prosemirror-view@1.40.1/node_modules/prosemirror-view/dist/index.js
90.6 KiB	../node_modules/.pnpm/@tiptap+core@3.20.6_@tiptap+pm@3.20.6/node_modules/@tiptap/core/dist/index.js

Largest files eagerly shipped from src/scenes/activity/explore/EventsScene.tsx

Size	File
319.8 KiB	../node_modules/.pnpm/posthog-js@1.438.8_@types+react@18.3.27_react@18.3.1/node_modules/posthog-js/dist/module.mjs
280.8 KiB	src/taxonomy/core-filter-definitions-by-group.json
220.8 KiB	../node_modules/.pnpm/@posthog+icons@0.38.0_react-dom@18.3.1_react@18.3.1__react@18.3.1/node_modules/@posthog/icons/dist/posthog-icons.es.js
181.9 KiB	src/queries/validators.js
153.7 KiB	../node_modules/.pnpm/re2js@0.4.1/node_modules/re2js/build/index.esm.js
126.8 KiB	../node_modules/.pnpm/react-dom@18.3.1_react@18.3.1/node_modules/react-dom/cjs/react-dom.production.min.js
113.0 KiB	../packages/quill/packages/quill/dist/index.js
103.1 KiB	src/products.tsx
99.6 KiB	src/lib/api.ts
93.3 KiB	../node_modules/.pnpm/prosemirror-view@1.40.1/node_modules/prosemirror-view/dist/index.js

Largest files eagerly shipped from src/scenes/session-recordings/detail/SessionRecordingDetail.tsx

Size	File
319.8 KiB	../node_modules/.pnpm/posthog-js@1.438.8_@types+react@18.3.27_react@18.3.1/node_modules/posthog-js/dist/module.mjs
316.0 KiB	../node_modules/.pnpm/posthog-js@1.438.8_@types+react@18.3.27_react@18.3.1/node_modules/posthog-js/dist/rrweb.js
280.8 KiB	src/taxonomy/core-filter-definitions-by-group.json
220.8 KiB	../node_modules/.pnpm/@posthog+icons@0.38.0_react-dom@18.3.1_react@18.3.1__react@18.3.1/node_modules/@posthog/icons/dist/posthog-icons.es.js
181.9 KiB	src/queries/validators.js
153.7 KiB	../node_modules/.pnpm/re2js@0.4.1/node_modules/re2js/build/index.esm.js
126.8 KiB	../node_modules/.pnpm/react-dom@18.3.1_react@18.3.1/node_modules/react-dom/cjs/react-dom.production.min.js
113.0 KiB	../packages/quill/packages/quill/dist/index.js
103.1 KiB	src/products.tsx
99.6 KiB	src/lib/api.ts

_{Posted automatically by check-eager-graph · sizes are eager output bytes (shipped, post-tree-shake) from the esbuild metafile · part of #32479}

✅ Toolbar bundle — eager 2.23 MiB within budget

What the toolbar ships to customer pages, measured from the esbuild output (minified, post-tree-shake). The eager set is the entry plus everything statically imported from it — fetched before any feature runs; deferred chunks load lazily. The eager guardrail is 5.72 MiB. Each output file must also stay below 10 MB, where CloudFront stops compressing it. The module boundary is enforced separately by check-toolbar-graph.

Metric	Size	Δ vs base	Budget
Eager (shipped) entry + static imports	2.23 MiB · 19 files	🔺 +16 B (+0.0%)	████░░░░░░ 39.0% of 5.72 MiB
Deferred (lazy)	2.10 MiB · 40 files	no change	n/a — loads on demand
Loader dist/toolbar.js	1.2 KiB	no change	█░░░░░░░░░ 6.0% of 19.5 KiB

Largest eagerly-shipped chunks

Size	File
873.1 KiB	dist/toolbar/toolbar-app-JJ7ICAUC.css
664.0 KiB	dist/toolbar/chunk-chunk-67V4GY4R.js
257.8 KiB	dist/toolbar/chunk-chunk-TSK433WX.js
131.8 KiB	dist/toolbar/chunk-chunk-FDH2IBXT.js
129.6 KiB	dist/toolbar/chunk-chunk-RVAE6ENC.js
74.7 KiB	dist/toolbar/toolbar-app-IFUCFPTL.js
69.0 KiB	dist/toolbar/chunk-chunk-TSAL54PB.js
35.6 KiB	dist/toolbar/chunk-chunk-HYUKXQTM.js
21.7 KiB	dist/toolbar/chunk-chunk-2ZONIQGQ.js
6.8 KiB	dist/toolbar/chunk-chunk-DV7IWQNF.js

_{Posted automatically by check-toolbar-size · sizes are toolbar output bytes (shipped, post-tree-shake) from the esbuild metafile}

✅ Dist folder size — 🔺 +149.2 KiB (+0.0%)

Total size of the built frontend/dist folder (all assets), compared against the base branch.

Total: 1008.55 MiB · 🔺 +149.2 KiB (+0.0%)

ℹ️ MCP UI apps size — 32 app(s), 17274.7 KB JS

Built size of each MCP UI app (main.js + styles.css).

App	JS	CSS
debug	598.2 KB	206.2 KB
action	454.4 KB	206.2 KB
action-list	564.5 KB	206.2 KB
cohort	453.4 KB	206.2 KB
cohort-list	563.5 KB	206.2 KB
email-template	453.2 KB	206.2 KB
error-details	469.9 KB	206.2 KB
error-issue	454.8 KB	206.2 KB
error-issue-list	565.1 KB	206.2 KB
experiment	561.6 KB	206.2 KB
experiment-list	565.3 KB	206.2 KB
experiment-results	566.6 KB	206.2 KB
feature-flag	567.1 KB	206.2 KB
feature-flag-list	570.9 KB	206.2 KB
feature-flag-testing	457.6 KB	206.2 KB
inline-scan	453.9 KB	206.2 KB
insight-actors	562.7 KB	206.2 KB
llm-costs	559.6 KB	206.2 KB
session-recording	455.7 KB	206.2 KB
survey	455.0 KB	206.2 KB
survey-global-stats	562.2 KB	206.2 KB
survey-list	565.2 KB	206.2 KB
survey-stats	562.2 KB	206.2 KB
trace-span	453.8 KB	206.2 KB
trace-span-list	564.4 KB	206.2 KB
vision-observation-list	563.6 KB	206.2 KB
workflow	453.8 KB	206.2 KB
workflow-list	563.9 KB	206.2 KB
loops-review	458.1 KB	206.2 KB
query-results	816.3 KB	206.2 KB
render-ui	899.9 KB	206.2 KB
visual-review-snapshots	458.3 KB	206.2 KB

✅ Playwright — all passed

All tests passed.

View test results →

⚠️ Backend coverage — 94.0% of changed backend lines covered — 6 uncovered

🧪 Backend test coverage

Patch coverage — changed backend lines (products + core): ███████████████████░ 94.0% (105 / 111)

File	Patch	Uncovered changed lines
products/tasks/backend/temporal/process_task/activities/get_task_processing_context.py	57.1%	481–483
products/tasks/backend/presentation/serializers.py	91.7%	92, 3767
products/tasks/backend/facade/api.py	94.4%	4872

🤖 Agents: add a test only if an uncovered line exposes a realistic regression that existing tests miss. Otherwise explain why no new test is needed under "How did you test this code?". Gap list: the patch-coverage artifact on this run (gh run download 213400873286482 -n patch-coverage), or the coverage-data block at the end of this comment.

Per-product line coverage (touched products)

Product	Coverage	Lines
platform_features	██░░░░░░░░░░░░░░░░░░ 12.1%	7 / 58
demo	███████████░░░░░░░░░ 55.8%	1,506 / 2,699
data_tools	████████████░░░░░░░░ 61.2%	90 / 147
warehouse_sources_queue	█████████████░░░░░░░ 66.8%	1,926 / 2,885
aeo	███████████████░░░░░ 76.3%	617 / 809
batch_exports	████████████████░░░░ 80.2%	21,577 / 26,893
github_commands	█████████████████░░░ 83.7%	978 / 1,168
apm	█████████████████░░░ 84.1%	1,306 / 1,553
engineering_analytics	█████████████████░░░ 85.1%	11,410 / 13,409
mcp_analytics	█████████████████░░░ 85.7%	4,849 / 5,656
ml_inference	█████████████████░░░ 86.3%	524 / 607
warehouse_suggestions	█████████████████░░░ 87.3%	2,252 / 2,580
posthog_ai	██████████████████░░ 87.7%	4,226 / 4,817
notebooks	██████████████████░░ 88.4%	15,242 / 17,235
today	██████████████████░░ 88.6%	2,866 / 3,233
mcp_registry	██████████████████░░ 89.0%	1,596 / 1,794
product_tours	██████████████████░░ 89.1%	1,337 / 1,500
dashboards	██████████████████░░ 89.2%	7,334 / 8,224
webmcp	██████████████████░░ 89.6%	240 / 268
cohorts	██████████████████░░ 89.6%	8,458 / 9,438
signals	██████████████████░░ 90.0%	69,696 / 77,462
data_modeling	██████████████████░░ 90.1%	10,856 / 12,043
ai_training	██████████████████░░ 90.4%	349 / 386
metrics	██████████████████░░ 90.6%	8,362 / 9,227
visual_review	██████████████████░░ 90.7%	9,853 / 10,867
data_warehouse	██████████████████░░ 90.8%	14,885 / 16,388
exports	██████████████████░░ 90.9%	9,772 / 10,751
canvas	██████████████████░░ 91.0%	7,805 / 8,581
autoresearch	██████████████████░░ 91.1%	10,436 / 11,453
business_knowledge	██████████████████░░ 91.2%	8,880 / 9,740
managed_warehouse	██████████████████░░ 91.2%	11,053 / 12,114
tasks	██████████████████░░ 91.5%	82,643 / 90,303
cdp	██████████████████░░ 91.6%	9,704 / 10,591
error_tracking	██████████████████░░ 91.6%	16,799 / 18,330
wizard	██████████████████░░ 91.7%	6,028 / 6,572
streamlit_apps	██████████████████░░ 91.9%	3,139 / 3,415
conversations	██████████████████░░ 92.1%	29,759 / 32,328
stamphog	███████████████████░ 92.5%	8,656 / 9,356
early_access_features	███████████████████░ 92.6%	1,339 / 1,446
managed_migrations	███████████████████░ 92.6%	1,630 / 1,760
alerts	███████████████████░ 92.9%	7,246 / 7,797
review_hog	███████████████████░ 92.9%	16,541 / 17,798
surveys	███████████████████░ 93.1%	6,626 / 7,118
web_analytics	███████████████████░ 93.1%	24,961 / 26,811
notifications	███████████████████░ 93.2%	1,152 / 1,236
approvals	███████████████████░ 93.4%	4,526 / 4,848
cross_project_dashboards	███████████████████░ 93.4%	880 / 942
slack_app	███████████████████░ 93.5%	14,855 / 15,883
context_layer	███████████████████░ 93.7%	3,409 / 3,640
marketing_analytics	███████████████████░ 93.8%	20,074 / 21,402
messaging	███████████████████░ 93.8%	4,468 / 4,762
billing_alerts	███████████████████░ 93.9%	2,090 / 2,226
mcp_store	███████████████████░ 93.9%	9,008 / 9,593
customer_analytics	███████████████████░ 93.9%	26,130 / 27,821
workflows	███████████████████░ 93.9%	17,808 / 18,957
experiments	███████████████████░ 94.1%	34,286 / 36,432
ai_observability	███████████████████░ 94.1%	26,476 / 28,126
legal_documents	███████████████████░ 94.3%	2,289 / 2,427
replay_vision	███████████████████░ 94.3%	30,115 / 31,928
logs	███████████████████░ 94.5%	15,850 / 16,770
actions	███████████████████░ 94.6%	973 / 1,029
reminders	███████████████████░ 94.8%	760 / 802
endpoints	███████████████████░ 94.8%	9,333 / 9,847
skills	███████████████████░ 94.8%	7,063 / 7,447
growth	███████████████████░ 94.8%	11,268 / 11,880
annotations	███████████████████░ 95.0%	816 / 859
tracing	███████████████████░ 95.2%	3,631 / 3,815
data_catalog	███████████████████░ 95.5%	4,420 / 4,628
access_control	███████████████████░ 95.6%	7,881 / 8,245
product_analytics	███████████████████░ 95.8%	28,751 / 30,027
alerts_platform	███████████████████░ 95.8%	5,012 / 5,234
revenue_analytics	███████████████████░ 95.9%	1,878 / 1,959
data_quality	███████████████████░ 95.9%	8,569 / 8,933
feature_flags	███████████████████░ 96.2%	26,932 / 27,997
warehouse_sources	███████████████████░ 96.3%	438,276 / 455,218
security	███████████████████░ 96.7%	1,452 / 1,501
pulse	███████████████████░ 97.4%	2,023 / 2,078
analytics_platform	████████████████████ 98.0%	2,775 / 2,833
groups	████████████████████ 98.0%	1,818 / 1,855
field_notes	████████████████████ 99.4%	172 / 173

Report-only. Patch coverage = changed backend lines covered vs origin/master. Sorted lowest first.
Known gaps: lines covered only by Temporal tests show as uncovered; core line numbers may drift if master changed the same file.

✅ Hogbox preview — ready, open the preview

▶ Open the preview

🔑 Login	test@posthog.com / 12345678 (demo data)
🧩 Running	this PR's backend and frontend, on the PostHog :master base
🔗 Link	stable across rebuilds: a re-push swaps the box underneath, the URL stays
🔒 Access	tailnet only (PostHog VPN)
🛠️ Admin	inspect & debug state in hogland
💤 Idle	sleeps after ~30 min idle (snapshot to S3, zero node cost) and wakes on your next visit in ~30s, behind a brief "waking up" screen

_{commit 7054b27 · box box-c2de2e8079e9 · ready in 1295s (push → usable) · build log · rebuilds on every push, torn down on close}

[coderabbitai]

Note

Reviews paused

It looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the reviews.auto_review.auto_pause_after_reviewed_commits setting.

Use the following commands to manage reviews:

• @coderabbitai resume to resume automatic reviews.
• @coderabbitai review to trigger a single review.

Use the checkboxes below for quick actions:

• ▶️ Resume reviews
• 🔍 Trigger review

📝 Walkthrough

The change adds Pi runtime support to task APIs, task surfaces, and stream processing. It translates Pi history and live events into stored notifications, handles Pi permission requests, and reconciles covered event IDs. Pi runs use session history and Pi-specific model and effort settings. The change also updates retry and compaction status displays, supports extension-question input fields, and adds task fixtures and tests.

Priority: ➖ Normal

Merge Risk: 🔵 Low · up to fb255

The Pi web sessions feature is behind a flag and the remaining issues are edge cases: a stale runtime marker, a late state response, attachment-footer parsing, and duplicate usage counts on history rebuilds. They are worth fixing, but the change is mergeable once the owner is aware of them.

Security architecture risk: 🟡 Moderate · up to fb255

The change adds consequential browser-based approval and task-session controls. Existing task access checks limit exposure, but approval failure handling and authorization of live configuration changes could not be verified end to end.

Retained concerns
No architecture-level concerns identified.

Security review details

Security Blast Radius

• inferred — The reviewed history change broadens readable content within an already accessible task session, not across tasks or teams. The predecessor query is bounded and requires the same active session. Commands can affect that task's sandbox and available tools; external credential and tool authority were not fully inspected.

Trust Boundaries and Controls

• observed — Browser-controlled commands pass through authenticated task access checks. The backend rejects Pi commands for non-Pi tasks, applies code-access checks to consequential Pi commands subject to existing exemptions, validates the sandbox URL, and attaches connection credentials obtained for the selected task run.
• observed — Frontend approval delivery requires a matching source run, a nonterminal current run, and a live rather than replay-only surface. Delivery keys combine run and request identity, suppress duplicate in-flight sends, and support cancellation. These client controls supplement rather than replace backend authority.

Resilience and Maintainability Implications

• observed — The backend distinguishes a terminal run whose sandbox is gone from transient sandbox readiness: the former receives a definitive run-ended response. Frontend approval delivery preserves a retry path for delivery errors and separately handles definitively ended targets. Whether Pi inner-command rejection follows this same failure contract remains unresolved.

Hardening Proposals

• proposed — Validate the deployed Pi approval acknowledgement contract and interrupted live-configuration transitions, including inner-command rejection. Confirm that unsuccessful decisions remain actionable and that live configuration changes enforce the same entitlement boundary as run creation. This is a verification proposal, not an observed vulnerability.

🚥 Pre-merge checks | ✅ 1

✅ Passed checks (1 passed)

Check name	Status	Explanation
Description check	✅ Passed	The description is complete and self-contained. It includes the problem, user-visible changes, screenshots, flow diagrams, test rationale, automated and end-to-end validation, known gaps, release stat…

✨ Finishing Touches

📝 Generate docstrings

• Commit to this branch
• Create a new PR

🧪 Generate unit tests (beta)

• Commit to this branch
• Create a new PR

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Note

Quiet mode is enabled, so only the most important comments were posted inline. Other review comments are grouped below.

🟡 Other comments (2)

products/posthog_ai/frontend/logics/runStreamLogic.ts-2926-2933 (1)

2926-2933: 🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win

Clear piRuntime on reset.

runStreamLogic is keyed by streamKey, so a stable instance can be reused. RunSurfaceBootstrap resets the instance before bootstrap, and taskRuntime can be omitted. bootstrapRun then preserves the previous piRuntime value. Because omitted task runtime defaults to ACP, an ACP permission request can use pi/rpc instead of permission_response.

retryConnection and attachOptimisticResume do not call reset; they retain runtime state for same-task retries and resumes.

Suggested fix

         piRuntime: [
             false,
             {
                 bootstrapRun: (state, { taskRuntime }) =>
                     taskRuntime === undefined ? state : taskRuntime === TaskRuntimeEnumApi.Pi,
                 markPiRuntime: () => true,
+                reset: () => false,
             },
         ],

products/posthog_ai/frontend/logics/runStreamLogic.ts-3809-3816 (1)

3809-3816: 🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win

Normalize persisted Pi permission requests.

normalizeHistory preserves _posthog/permission_request options unchanged. An active, unresolved replay restores the pending permission card, but parsePermissionRequestFrame does not add allow_once. A persisted Pi MCP request that contains only allow_always can therefore omit the one-shot action.

Pass piRuntime to parsePermissionRequestFrame and apply the existing withPiOneShotAllow rule at that parsing boundary. The live buffered-path claim is too broad because live frames are normalized before buffering when piRuntime is already true.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository: PostHog/posthog/.coderabbit.yaml

Review profile: QUIET

Plan: Enterprise

Run ID: e1ebaf04-6f26-4fc0-90e3-80ef06f9712a

📥 Commits

Reviewing files that changed from the base of the PR and between 1c66a26 and 042de15.

📒 Files selected for processing (14)

• products/posthog_ai/frontend/components/RunSurfaceImpl.test.tsx
• products/posthog_ai/frontend/components/RunSurfaceImpl.tsx
• products/posthog_ai/frontend/components/ThreadItems.tsx
• products/posthog_ai/frontend/logics/runInteractionLogic.test.ts
• products/posthog_ai/frontend/logics/runInteractionLogic.ts
• products/posthog_ai/frontend/logics/runStreamLogic.test.ts
• products/posthog_ai/frontend/logics/runStreamLogic.ts
• products/posthog_ai/frontend/scenes/TaskTracker/TaskTracker.stories.tsx
• products/posthog_ai/frontend/scenes/TaskTracker/components/TaskDetailPage.tsx
• products/posthog_ai/frontend/scenes/TaskTracker/components/TaskRunChat.tsx
• products/posthog_ai/frontend/scenes/TaskTracker/components/TaskRunComposer.tsx
• products/posthog_ai/frontend/types/wireTypes.ts
• products/posthog_ai/frontend/utils/piWire.test.ts
• products/posthog_ai/frontend/utils/piWire.ts

Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 7 remain after this review.

[trunk-io]

   

Failed Test	Failure Summary	Logs
Scenes-App/SidePanels SidePanelNotebooks smoke-test	The test timed out while waiting for loading indicators or spinners to disappear.	Logs ↗︎
Metrics/MetricsDashboardImportModal ImportProgress smoke-test	The page timed out waiting for loading indicators to disappear; a spinner remained visible.	Logs ↗︎
Scenes-App/Project Homepage ProjectHomepage smoke-test	The test timed out while waiting for loading indicators or spinners to disappear.	Logs ↗︎
Scenes-App/OAuth/Authorize DefaultScopes smoke-test	The test timed out while waiting for loading indicators or spinners to disappear.	Logs ↗︎

... and 1 more

View Full Report ↗︎ ⋅ Docs

[coderabbitai]

Note

Quiet mode is enabled, so only the most important comments were posted inline. Other review comments are grouped below.

🟡 Other comments (1)

products/posthog_ai/frontend/utils/piWire.ts-331-331 (1)

331-331: 🩺 Stability & Availability | 🟡 Minor | ⚡ Quick win

Validate permission options before calling startsWith.

isPermissionRequestFrame checks only the top-level type. If an option is null or its kind is missing or not a string, withPiMcpOptions throws before parsePermissionOption can reject it. The exception escapes handleSseEvent and can interrupt stream processing.

Filter invalid options before this transformation.

Suggested fix

-    const options = (Array.isArray(frame.options) ? frame.options : []).map((option) =>
-        option.kind.startsWith('reject') ? { ...option, _meta: { ...option._meta, hint: PI_MCP_REJECT_HINT } } : option
-    )
+    const options = (Array.isArray(frame.options) ? frame.options : [])
+        .filter((option): option is PermissionOption => typeof option?.kind === 'string')
+        .map((option) =>
+            option.kind.startsWith('reject') ? { ...option, _meta: { ...option._meta, hint: PI_MCP_REJECT_HINT } } : option
+        )

Source: Learnings

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository: PostHog/posthog/.coderabbit.yaml

Review profile: QUIET

Plan: Enterprise

Run ID: 6125fd75-04ab-494b-83a2-916c6094c07d

📥 Commits

Reviewing files that changed from the base of the PR and between b1738e7 and c94a4ea.

📒 Files selected for processing (11)

• products/posthog_ai/frontend/components/OptionSelector.tsx
• products/posthog_ai/frontend/components/QuestionField.tsx
• products/posthog_ai/frontend/components/QuestionInput.test.tsx
• products/posthog_ai/frontend/components/QuestionInput.tsx
• products/posthog_ai/frontend/components/ThreadItems.tsx
• products/posthog_ai/frontend/logics/runStreamLogic.test.ts
• products/posthog_ai/frontend/logics/runStreamLogic.ts
• products/posthog_ai/frontend/policy/questionUtils.ts
• products/posthog_ai/frontend/types/wireTypes.ts
• products/posthog_ai/frontend/utils/piWire.test.ts
• products/posthog_ai/frontend/utils/piWire.ts

Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 6 remain after this review.

[coderabbitai]

Note

Quiet mode is enabled, so only the most important comments were posted inline. Other review comments are grouped below.

🟡 Other comments (4)

products/tasks/backend/facade/api.py-8909-8912 (1)

8909-8912: 🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win

Preserve Pi effort selections during default resolution.

run_task applies _with_ai_run_defaults to fresh runs. That helper calls apply_ai_run_defaults, which resolves effort-only selections as ACP because it does not pass the request runtime. An ACP preference can therefore add runtime_adapter and model; the reviewed branch then removes both model and the explicitly requested reasoning_effort.

Make default resolution runtime-aware. Do not apply ACP defaults to Pi requests, and preserve the requested reasoning_effort.

products/posthog_ai/frontend/logics/runStreamLogic.ts-4685-4685 (1)

4685-4685: 🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win

Keep idless turn usage stable across history rebuilds.

When a turn has no event_id, each ingestion creates a new map key. Reconnect and finalization rebuilds replay those turns, but cache.piTurnUsage survives the rebuild. The same turn therefore increases token totals again.

Rebuild the usage map from the reconciled log, or assign stable per-entry keys before replay.

products/posthog_ai/frontend/utils/piWire.ts-158-158 (1)

158-158: 🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win

Preserve user-written file lists.

A prompt ending with Attached files:\n- README.md matches this pattern without any uploaded attachment. piAttachedFiles then removes that text and creates an attachment chip instead. The replay no longer shows the original prompt.

Only strip the footer when its paths match the server-generated attachment layout and UUID-prefixed filenames. Otherwise, preserve the text.

products/posthog_ai/frontend/logics/runInteractionLogic.ts-1485-1513 (1)

1485-1513: 🩺 Stability & Availability | 🟡 Minor | ⚡ Quick win

Ignore delayed get_state results after disposal or a run change.

markRunStarted dispatches setPiSessionConfig after its request without checking that the interaction logic is still current. The active-creation flow changes the logic key when it switches to the created run, so the old logic can be disposed while its request is still pending. The late response can then dispatch against that disposed logic.

Do not reset cache.piSessionConfigRunId on failure. The documented fallback is once per run and does not require retrying the request.

Suggested fix

                 cache.piSessionConfigRunId = props.runId
+                const runId = props.runId
+                const disposables = cache.disposables
                 try {
                     const response = await tasksRunsCommandCreate(
                         String(currentProjectId),
                         props.taskId,
-                        props.runId,
+                        runId,
                         piRpcRequest({ type: 'get_state' }, uuid())
                     )
+                    if (disposables.isDisposed || props.runId !== runId) {
+                        return
+                    }
                     const config = parsePiSessionConfig(response.result)
                     if (config) {
                         actions.setPiSessionConfig(config)
                     }
                 } catch {
-                    // The picker keeps its fallback. A send with a pick still applies it.
+                    // The picker keeps its fallback. A send with a pick still applies it.
                 }

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository: PostHog/posthog/.coderabbit.yaml

Review profile: QUIET

Plan: Enterprise

Run ID: 48089def-2098-4b3d-98e7-3e70cb26863e

📥 Commits

Reviewing files that changed from the base of the PR and between c94a4ea and eff3c0b.

⛔ Files ignored due to path filters (4)

• products/logs/frontend/generated/api.ts is excluded by !**/generated/**
• products/tasks/frontend/generated/api.schemas.ts is excluded by !**/generated/**
• products/tasks/frontend/generated/api.zod.ts is excluded by !**/generated/**
• services/mcp/src/generated/tasks/api.ts is excluded by !**/generated/**

📒 Files selected for processing (20)

• frontend/src/lib/constants.tsx
• products/posthog_ai/frontend/components/RunSurfaceImpl.test.tsx
• products/posthog_ai/frontend/components/RunSurfaceImpl.tsx
• products/posthog_ai/frontend/components/composer/ComposerModelEffortPickers.test.tsx
• products/posthog_ai/frontend/components/composer/ComposerModelEffortPickers.tsx
• products/posthog_ai/frontend/logics/runInteractionLogic.test.ts
• products/posthog_ai/frontend/logics/runInteractionLogic.ts
• products/posthog_ai/frontend/logics/runStreamLogic.test.ts
• products/posthog_ai/frontend/logics/runStreamLogic.ts
• products/posthog_ai/frontend/scenes/TaskTracker/components/TaskDetailPage.tsx
• products/posthog_ai/frontend/scenes/TaskTracker/components/TaskRunComposer.tsx
• products/posthog_ai/frontend/utils/composerModels.ts
• products/posthog_ai/frontend/utils/piWire.test.ts
• products/posthog_ai/frontend/utils/piWire.ts
• products/tasks/backend/facade/api.py
• products/tasks/backend/models.py
• products/tasks/backend/presentation/serializers.py
• products/tasks/backend/presentation/views/api.py
• products/tasks/backend/tests/test_api.py
• services/mcp/src/api/generated.ts

Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 7 remain after this review.

[coderabbitai]

Caution

Some comments are outside the diff and can’t be posted inline due to GitHub limitations.

⚠️ Outside diff range comments (1)

🟠 Major · Guard the Pi state response against context replacement. · runInteractionLogic.ts:1508

products/posthog_ai/frontend/logics/runInteractionLogic.ts:1508
🩺 Stability & Availability | 🟠 Major | ⚡ Quick win

Guard the Pi state response against context replacement.

If the Kea context closes while get_state is in flight, the response still calls setPiSessionConfig on the old logic. Capture getContext() before the request and compare it before dispatching the result. Also confirm that the project, task, and run still match the request.

As per coding guidelines: “An in-flight request resolving … still needs its own guard: compare getContext() against the context the resource was set up in.”

Source: Coding guidelines

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository: PostHog/posthog/.coderabbit.yaml

Review profile: QUIET

Plan: Enterprise

Run ID: 057ac39e-4078-4271-9665-b88411dc0ae4

📥 Commits

Reviewing files that changed from the base of the PR and between eff3c0b and fb255bd.

⛔ Files ignored due to path filters (4)

• products/logs/frontend/generated/api.ts is excluded by !**/generated/**
• products/tasks/frontend/generated/api.schemas.ts is excluded by !**/generated/**
• products/tasks/frontend/generated/api.zod.ts is excluded by !**/generated/**
• services/mcp/src/generated/tasks/api.ts is excluded by !**/generated/**

📒 Files selected for processing (7)

• frontend/src/lib/constants.tsx
• products/posthog_ai/frontend/logics/runInteractionLogic.ts
• products/tasks/backend/facade/api.py
• products/tasks/backend/presentation/serializers.py
• products/tasks/backend/presentation/views/api.py
• products/tasks/backend/tests/test_api.py
• services/mcp/src/api/generated.ts

Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 6 remain after this review.

[posthog]

✅ Visual changes approved by @puemos — baseline updated in 9185f41.

View this run in PostHog

2 new.

_{Install the Visual Review Chrome extension to see visual review results at the top of your pull requests.}

[jonathanlab]

Thanks for this!

Could we ensure that there is as much code sharing happening as possible instead of duplicating the code? It would be great if we could keep the Pi logic surface as small as possible

[jonathanlab]

Owned by #team-ai-observability? Shouldnt it be your team..?

[jonathanlab]

What does singleHarness imply here? Difficult to read. Seems like we should just properly adjust modelsForRuntimeAdapter to handle the pi runtime

[jonathanlab]

A Pi task sets it as the Pi thinking level and does not accept 'ultracode
This is not a useful addition, ultracode is something we do not properly support in general AFAIK. That's just claude code's xhigh + a prompt nudge to use orchestration.

[jonathanlab]

A lot of this code seems duplicated from agent/src/pi/conversation/ and the like.

[github-actions]

React Doctor found no issues in the changed files. 🎉

_{Reviewed by React Doctor for commit 7054b27.}

[puemos]

Changes since the review

Shared code

• Before: the web had its own Pi translation (piWire.ts).
• After: @posthog/agent-contracts holds the Pi and ACP mapping. The Pi server, Desktop and the web use it.

Pi server

• Before: writes only the Pi format.
• After: writes ACP when pi-acp is on. Accepts permission_response and set_config_option.

Backend

• Before: sends no format setting.
• After: checks the flag for each run and tells the sandbox the format.

Web

• Before: translates Pi entries and sends pi/rpc.
• After: reads Pi runs through the ACP path. A small reader shows old Pi-format runs.

Desktop

• Before: reads only the Pi format.
• After: reads the two formats.

Test

• Before: unit tests only.
• After: an end-to-end test with a real model, and five videos in the PR text. The test found four bugs. This PR fixes them.

CI is green. The desktop-skip-backend-check label is on this PR, because the two halves can ship in any order.

[parameterai]

Risk: No findings

This PR makes Pi task sessions work on the web by having the Pi agent server speak ACP. Since the last review the branch was only rebased onto a newer master: the PR's own security-relevant code (ACP wire mapping, command relay allowlist, serializers, sandbox env, run stream logic) is byte-identical to the reviewed state, and the only PR-authored edits are an error_type passthrough line in the agent server and a disabled-propagation UI hardening fix. No new security-relevant behavior was introduced by the delta.

_{Sentinel reviewed 7054b27 · Review settings}

[deployment-status-posthog]

Deploy status

Environment	Status	Deployed At	Workflow
dev	✅ Deployed	2026-10-13 17:21 UTC	Run
prod-us	✅ Deployed	2026-10-13 17:38 UTC	Run
prod-eu	✅ Deployed	2026-10-13 17:38 UTC	Run