feat(tasks): clear a finished cloud run's conversation without a sandbox

[haacked]

Problem

Follows #76457 (merged), which made /clear work on a live cloud run.

A finished run is the other half, and most cloud runs are finished when you look at them. There's no agent to send /clear to, so the composer treated it as an ordinary message and started a whole new run: boot a sandbox, resume the conversation, clear it, idle. A full run's cost to clear a conversation the next run would rebuild from the log anyway.

The cloud thread also rendered none of it. _posthog/conversation_cleared fell through foldLogToThread's catch-all, and the two clearing status frames showed up as raw Status: clearing rows.

Changes

POST runs/{id}/clear_conversation appends the typed message plus a _posthog/conversation_cleared marker to a finished run's log. The logs endpoint already serves a resume chain concatenated, so rehydration stops at the marker with no change: the next run continues the task with an empty conversation, and its checkpoints, artifacts, and visible history are untouched.

Only for a finished run. An active one has an agent that owns the clear, and a live writer on the same log object this read-modify-write append would race, so it returns 409. Same reasoning as the note at webhooks.py:245. A repeat call against a log already ending at the boundary appends nothing, and the append holds the run's row lock so a concurrent resume can't flip it active mid-write.

The web composer gates on the conversationClear capability from the latest run_started and falls back to an ordinary new run when it's absent. The thread renders the boundary and a "Clearing…" spinner mirroring compaction, and /clear is sent unwrapped, since a context block would hide the command behind it.

Note

The desktop client half lives in #82950 and merges after this deploys - the "Desktop backend coupling" check requires the split, because desktop releases aren't orchestrated with backend deploys and this endpoint must exist before a desktop build calls it.

How did you test this code?

Manually, against a local stack, with a seeded finished run:

I typed /clear in the composer on a finished run and got the /clear bubble, the "Conversation cleared" divider, and the context ring at zero, through the real endpoint. I could not test the resume half or a live cloud run locally: both need a sandbox provider and a Temporal worker that a local stack doesn't have.

Automated, run locally on this exact content: the three backend clear_conversation tests, the three web suites (runStreamLogic, runInteractionLogic, wireTypes), frontend typecheck, and a kea typegen idempotence check.

New tests, and the regression each catches:

• test_clear_conversation_records_the_boundary - the marker lands with the typed message ahead of it, tagged for client replay, and carries no sessionId. If it ever gained one, resume would target a bogus session. Also covers the repeat-call no-op: without it, a double-submit stacks duplicate dividers.
• test_clear_conversation_rejects_an_active_run (queued, in_progress) - 409 and an untouched log. Without the guard this appends into a log with a live writer.
• Fallback coverage in runInteractionLogic.test.ts - with the capability absent, the composer starts an ordinary run and never calls the endpoint. This is the silent-wrong-behavior case.
• Rollback coverage in runStreamLogic.test.ts - the capability gate follows the latest run_started, so an agent rollback mid-chain drops it instead of ratcheting true.

Automatic notifications

• Publish to changelog?

Docs update

None needed. No user-facing docs cover /clear yet.

🤖 Agent context

Autonomy: Human-driven (agent-assisted)

Claude Code (PostHog Desktop session) fixed review findings from the bot reviewers, rebased the branch after #76457 squash-merged, and split the desktop client out to #82950 when the "Desktop backend coupling" check required it - the surviving content is what charlesvien approved, minus products/desktop. Skills invoked: /stacking-prs, /writing-tests, /writing-code-comments, /writing-pr-descriptions, /resolve-conflicts, /address-pr-reviews.

[github-actions]

🤖 CI report

⚠️ Bundle size — 🔺 +1.4 KiB (+0.0%)

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

Total: 67.86 MiB · 🔺 +1.4 KiB (+0.0%)

File	Size	Δ vs base
render-query/src/render-query/render-query.js	21.78 MiB	🔺 +1.4 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	1.27 MiB · 22 files	no change	███░░░░░░░ 28.1% of 4.51 MiB
authenticated shell (every logged-in page) src/scenes/AuthenticatedShell.tsx	8.51 MiB · 3,133 files	🔺 +131 B (+0.0%)	█████████░ 87.6% of 9.71 MiB

🟢 node_modules/monaco-editor/ stays out of src/index.tsx
🟢 src/lib/components/ActivityLog/describers stays out of src/index.tsx
🟢 [object Object] stays out of src/index.tsx
🟢 [object Object] stays out of src/index.tsx
🟢 node_modules/monaco-editor/ stays out of src/scenes/AuthenticatedShell.tsx
🟢 src/lib/components/ActivityLog/describers stays out of src/scenes/AuthenticatedShell.tsx
🟢 [object Object] stays out of src/scenes/AuthenticatedShell.tsx
🟢 [object Object] stays out of src/scenes/AuthenticatedShell.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
912 B	../node_modules/.pnpm/ieee754@1.2.1/node_modules/ieee754/index.js
789 B	src/scenes/ChunkLoadErrorBoundary.tsx
762 B	src/index.tsx

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

Size	File
304.9 KiB	../node_modules/.pnpm/posthog-js@1.417.0/node_modules/posthog-js/dist/rrweb.js
267.7 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
243.8 KiB	../node_modules/.pnpm/posthog-js@1.417.0/node_modules/posthog-js/dist/module.js
238.4 KiB	src/taxonomy/core-filter-definitions-by-group.json
154.3 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
103.8 KiB	src/lib/api.ts
94.6 KiB	../packages/quill/packages/quill/dist/index.js
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

_{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 2.22 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	2.22 MiB · 17 files	no change	████░░░░░░ 38.7% of 5.72 MiB
Deferred (lazy)	2.08 MiB · 33 files	no change	n/a — loads on demand
Loader dist/toolbar.js	1.1 KiB	no change	█░░░░░░░░░ 5.8% of 19.5 KiB

Largest eagerly-shipped chunks

Size	File
730.0 KiB	dist/toolbar/toolbar-app-LL5TI34A.css
565.3 KiB	dist/toolbar/chunk-chunk-RCUAWDGK.js
484.6 KiB	dist/toolbar/chunk-chunk-2OORENDY.js
133.6 KiB	dist/toolbar/chunk-chunk-QQ2DRXHJ.js
131.8 KiB	dist/toolbar/chunk-chunk-T5KY5WYR.js
71.2 KiB	dist/toolbar/toolbar-app-UDDFGSJT.js
69.0 KiB	dist/toolbar/chunk-chunk-27JL52RE.js
35.6 KiB	dist/toolbar/chunk-chunk-2U5IYX7Z.js
20.9 KiB	dist/toolbar/chunk-chunk-KWHQMXDE.js
12.2 KiB	dist/toolbar/chunk-chunk-PIK3PADE.js

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

✅ Dist folder size — 🔺 +35.3 KiB (+0.0%)

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

Total: 1431.09 MiB · 🔺 +35.3 KiB (+0.0%)

ℹ️ MCP UI apps size — 31 app(s), 16614.3 KB JS

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

App	JS	CSS
debug	599.6 KB	184.6 KB
action	457.8 KB	184.6 KB
action-list	564.4 KB	184.6 KB
cohort	456.8 KB	184.6 KB
cohort-list	563.4 KB	184.6 KB
email-template	456.6 KB	184.6 KB
error-details	472.4 KB	184.6 KB
error-issue	457.5 KB	184.6 KB
error-issue-list	564.3 KB	184.6 KB
experiment	561.5 KB	184.6 KB
experiment-list	565.2 KB	184.6 KB
experiment-results	566.7 KB	184.6 KB
feature-flag	567.2 KB	184.6 KB
feature-flag-list	570.9 KB	184.6 KB
feature-flag-testing	461.0 KB	184.6 KB
insight-actors	562.2 KB	184.6 KB
invite-email-preview	456.0 KB	184.6 KB
llm-costs	559.5 KB	184.6 KB
session-recording	458.6 KB	184.6 KB
survey	458.4 KB	184.6 KB
survey-global-stats	562.2 KB	184.6 KB
survey-list	565.1 KB	184.6 KB
survey-stats	562.2 KB	184.6 KB
trace-span	457.2 KB	184.6 KB
trace-span-list	564.3 KB	184.6 KB
workflow	457.1 KB	184.6 KB
workflow-list	563.7 KB	184.6 KB
loops-review	461.4 KB	184.6 KB
query-results	750.8 KB	184.6 KB
render-ui	828.9 KB	184.6 KB
visual-review-snapshots	461.6 KB	184.6 KB

⚠️ Backend coverage — 91.0% of changed backend lines covered — 6 uncovered

🧪 Backend test coverage

Patch coverage — changed backend lines (products + core): ██████████████████░░ 91.0% (63 / 69)

File	Patch	Uncovered changed lines
products/tasks/backend/presentation/views/api.py	83.3%	1492, 1499
products/tasks/backend/models.py	86.4%	2551–2552, 2554
products/tasks/backend/facade/api.py	90.0%	2680

🤖 Agents: add a test covering the lines above, or note why under "How did you test this code?". Machine-readable gap list: the patch-coverage artifact on this run (gh run download 31750512626 -n patch-coverage), or the coverage-data block at the end of this comment.

Per-product line coverage (touched products)

Product	Coverage	Lines
demo	███████████░░░░░░░░░ 56.3%	1,497 / 2,661
tasks	██████████████░░░░░░ 70.2%	38,020 / 54,148
signals	█████████████████░░░ 82.9%	27,774 / 33,494
cdp	█████████████████░░░ 84.2%	3,914 / 4,649
data_modeling	█████████████████░░░ 86.3%	8,909 / 10,326
actions	█████████████████░░░ 86.6%	717 / 828
notebooks	█████████████████░░░ 87.2%	8,547 / 9,806
cohorts	██████████████████░░ 87.6%	6,528 / 7,452
product_tours	██████████████████░░ 87.9%	1,303 / 1,482
data_warehouse	██████████████████░░ 88.7%	12,307 / 13,880
managed_warehouse	██████████████████░░ 88.8%	6,642 / 7,478
exports	██████████████████░░ 89.1%	7,533 / 8,455
engineering_analytics	██████████████████░░ 89.4%	6,839 / 7,649
dashboards	██████████████████░░ 89.5%	6,222 / 6,949
alerts	██████████████████░░ 90.4%	4,554 / 5,039
mcp_analytics	██████████████████░░ 90.4%	3,968 / 4,388
streamlit_apps	██████████████████░░ 90.7%	2,630 / 2,901
conversations	██████████████████░░ 90.8%	19,364 / 21,329
error_tracking	██████████████████░░ 91.2%	11,325 / 12,415
stamphog	██████████████████░░ 91.3%	4,505 / 4,936
slack_app	██████████████████░░ 92.5%	11,334 / 12,255
canvas	███████████████████░ 92.6%	2,923 / 3,155
early_access_features	███████████████████░ 92.9%	1,347 / 1,450
ai_observability	███████████████████░ 92.9%	17,244 / 18,560
web_analytics	███████████████████░ 92.9%	16,337 / 17,578
posthog_ai	███████████████████░ 93.2%	1,326 / 1,422
surveys	███████████████████░ 93.3%	5,943 / 6,373
reminders	███████████████████░ 93.4%	468 / 501
marketing_analytics	███████████████████░ 93.5%	16,405 / 17,551
approvals	███████████████████░ 93.5%	3,491 / 3,734
product_analytics	███████████████████░ 93.6%	7,124 / 7,611
review_hog	███████████████████░ 94.1%	9,319 / 9,906
endpoints	███████████████████░ 94.3%	8,814 / 9,349
workflows	███████████████████░ 94.4%	8,046 / 8,526
skills	███████████████████░ 94.7%	3,869 / 4,085
data_quality	███████████████████░ 95.3%	1,457 / 1,529
logs	███████████████████░ 95.4%	12,372 / 12,975
replay_vision	███████████████████░ 95.6%	21,019 / 21,996
experiments	███████████████████░ 95.6%	28,832 / 30,152
annotations	███████████████████░ 96.2%	732 / 761
revenue_analytics	███████████████████░ 96.3%	1,887 / 1,960
customer_analytics	███████████████████░ 96.5%	12,779 / 13,248
user_interviews	███████████████████░ 96.5%	2,638 / 2,734
feature_flags	███████████████████░ 96.6%	18,730 / 19,397
warehouse_sources	███████████████████░ 97.5%	391,616 / 401,660
data_catalog	████████████████████ 97.8%	2,746 / 2,808
pulse	████████████████████ 98.4%	2,017 / 2,049

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.

[trunk-io]

   

View Full Report ↗︎ ⋅ Docs

[posthog]

🦔 ReviewHog reviewed this pull request

Found 0 must fix, 4 should fix, 3 consider.

Published 7 findings (view the review).

[posthog]

ReviewHog Alpha 🦔 If you find any issues helpful - please reply "valid", "invalid", etc., for evaluation purposes 🙏

[posthog]

ReviewHog Report

Business logic

Issues: 2 issues

Files (4)

• products/tasks/backend/models.py
• products/tasks/backend/facade/api.py
• products/tasks/backend/presentation/views/api.py
• products/tasks/mcp/tools.yaml

What were the main changes

• TaskRun.emit_conversation_cleared() writes the typed /clear message plus a _posthog/conversation_cleared marker straight into a finished run's log, so resume rebuilds an empty conversation past it
• facade.clear_task_run_conversation gates on run.is_terminal, returning not_found/not_terminal outcomes instead of racing a live agent's log writer
• New POST runs/{id}/clear_conversation/ endpoint: 200 with the updated run DTO on success, 404 if missing, 409 if the run is still active (send /clear to its agent instead)
• MCP tools.yaml registers the new operation, disabled by default like other run-mutation endpoints

Changes

Issues: 5 issues

Files (3)

• products/desktop/packages/api-client/src/posthog-client.ts
• products/desktop/packages/core/src/sessions/sessionEvents.ts
• products/desktop/packages/core/src/sessions/sessionService.ts

[greptile-apps]

Prompt To Fix All With AI

### Issue 1
products/posthog_ai/frontend/logics/runStreamLogic.ts:2087-2092
**Capability state remains stale**

When a resume chain contains an earlier run with `conversationClear` enabled and a later run whose agent does not advertise it, this sticky reducer remains true and authorizes the terminal-run clear endpoint. The UI then displays “Conversation cleared,” but the incapable agent ignores the boundary on resume and restores the full conversation.

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

_{Reviews (1): Last reviewed commit: "fix(tasks): address review findings on t..." | Re-trigger Greptile}

[github-actions]

🦔 Hogbox preview · ✅ ready

▶ 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 7b043ea · box box-44ca170e8942 · ready in 682s (push → usable) · build log · rebuilds on every push, torn down on close}

[github-actions]

React Doctor found no issues in the changed files. 🎉

_{Reviewed by React Doctor for commit db71288.}

[Copilot]

Pull request overview

Adds a dedicated “clear conversation” pathway for finished cloud task runs, avoiding the cost of spinning up a new sandbox just to execute /clear. This is implemented as a new backend endpoint that appends a /clear user turn plus a _posthog/conversation_cleared boundary marker to the run log, along with frontend + desktop rendering and capability-gated calling behavior.

Changes:

• Backend: add POST /api/projects/:id/tasks/:task_id/runs/:run_id/clear_conversation/ to record a /clear boundary only for terminal runs (409 for active runs).
• Frontend (cloud thread): render _posthog/conversation_cleared, handle “clearing”/“clearing_failed” status frames, and gate finished-run clears on the advertised conversationClear capability.
• Desktop: add an API client method + SessionService shortcut to record the boundary and paint the cleared divider locally for finished cloud runs.

Reviewed changes

Copilot reviewed 19 out of 20 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
products/tasks/mcp/tools.yaml	Registers the new clear-conversation operation in MCP tools (disabled by default).
products/tasks/frontend/generated/api.ts	Adds generated API client function for the clear-conversation endpoint.
products/tasks/backend/tests/test_api.py	Adds backend tests for boundary recording, dedupe, and active-run rejection.
products/tasks/backend/presentation/views/api.py	Adds the clear_conversation DRF action with OpenAPI annotations.
products/tasks/backend/models.py	Adds log-writing helpers to append a /clear turn + conversation boundary marker.
products/tasks/backend/facade/api.py	Adds facade method to lock/validate terminal status and append boundary to the log.
products/posthog_ai/frontend/types/wireTypes.ts	Extends notification param typings for conversation-cleared and status error fields.
products/posthog_ai/frontend/types/wireTypes.test.ts	Extends type contract tests for new notification shapes.
products/posthog_ai/frontend/types/streamTypes.ts	Adds conversation_cleared to ThreadItemType and clarifies errorMessage usage.
products/posthog_ai/frontend/logics/runStreamLogic.ts	Updates log folding to render the boundary divider and handle clearing statuses/errors; tracks conversationClearSupported.
products/posthog_ai/frontend/logics/runStreamLogic.test.ts	Adds logic tests for capability precedence and clear divider/status behavior.
products/posthog_ai/frontend/logics/runInteractionLogic.ts	Routes /clear on terminal runs to the new endpoint when capability is supported; avoids wrapping /clear in context.
products/posthog_ai/frontend/logics/runInteractionLogic.test.ts	Adds tests for terminal-run clear shortcut and capability-gated fallback behavior.
products/posthog_ai/frontend/components/ThreadView.tsx	Adds height estimate for the new conversation_cleared thread item type.
products/posthog_ai/frontend/components/ThreadRow.tsx	Renders ConversationClearedItem for the new thread item type.
products/posthog_ai/frontend/components/ThreadItems.tsx	Implements UI for “Clearing…” status and “Conversation cleared” divider.
products/desktop/packages/core/src/sessions/sessionServiceCloudClear.test.ts	Adds desktop tests for finished-run /clear boundary shortcut and fallback.
products/desktop/packages/core/src/sessions/sessionService.ts	Adds finished-run /clear shortcut to call backend boundary endpoint and paint events locally.
products/desktop/packages/core/src/sessions/sessionEvents.ts	Adds helper to create the local events sequence for a finished-run clear boundary.
products/desktop/packages/api-client/src/posthog-client.ts	Adds API client method to call the clear-conversation endpoint with typed error handling.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[trunk-io]

😎 Stack merged successfully - details.

[hosthog]

HostHog preview — posthog-desktop-web

The previews for this PR have been torn down and no longer serve.

[posthog]

👋 Visual changes detected for this PR.

Review and approve in PostHog Visual Review

If these changes are unexpected, they may be caused by a flaky test or a broken snapshot on master. Don't approve — rerun the job or wait for a fix.

[deployment-status-posthog]

Deploy status

Environment	Status	Deployed At	Workflow
dev	✅ Deployed	2026-08-13 23:56 UTC	Run
prod-us	✅ Deployed	2026-08-14 00:49 UTC	Run
prod-eu	✅ Deployed	2026-08-14 00:48 UTC	Run