feat(task-board): autonomous task runs as live task sessions (View work, cancel, degeneration guards)#3380
Merged
senamakel merged 11 commits intoJun 4, 2026
Conversation
Autonomous task-board runs now materialise as top-level "task session" conversation threads (labels: ["tasks"]) so they appear in Conversations → Tasks like a manually-run todo, stream live via the web-channel bridge (client_id="system" + TurnStateMirror), and finalise with a chat_done terminal event so the session stops "processing". Cards link to their session (sessionThreadId) and surface a "View work" action that opens the exact session. - task_session module: create_session_thread + append_final - dispatcher: stamp sessionThreadId, stream via spawn_progress_bridge, emit chat_done/chat_error; cancellable via an ACTIVE_RUNS registry + cancel_session wired into channel_web_cancel (so chat Cancel works on task threads), with race-safe write-back - todos: set_session_thread RPC/ops; todos_add carries source_metadata - approve flow: idempotent (dedup by source external_id) + hide already-picked-up items from the task-sources inbox - nav: View work opens the exact thread on the Tasks tab; Settings → Task Sources back returns to the opener; Intelligence tab is URL-backed; task-session threads excluded from the per-thread agent board list - i18n: conversations.taskKanban.viewWork across all locales Co-Authored-By: Claude <noreply@anthropic.com>
Autoregressive models can spiral into emitting the same content repeatedly; frequency_penalty / model tier only reduce the odds. Two deterministic, model-agnostic guards: - RepeatOutputGuard (cross-iteration): halts the tool loop when the same assistant output + tool-call signature repeats N times across iterations, even when each call "succeeds" (the failure guard resets on success). - StreamRepeatDetector (in-generation): aborts the streaming read when the same substantial line repeats N times within one response, truncating the blob instead of running to the token cap. Both unit-tested. Co-Authored-By: Claude <noreply@anthropic.com>
Add [task_sources:dedup] debug logs at the per-item gate: one per skip (already ingested under this source, unchanged) and one per route (new vs edited), to make per-source dedup decisions observable when debugging a fetch (fetched/routed/skipped_dupe). Co-Authored-By: Claude <noreply@anthropic.com>
Contributor
ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (29)
💤 Files with no reviewable changes (3)
✅ Files skipped from review due to trivial changes (10)
🚧 Files skipped from review as they are similar to previous changes (9)
📝 WalkthroughWalkthroughAdds session-thread creation/persistence and linking to TaskBoard cards, UI session navigation and inbox dedup/filtering, todos RPC for session linking, and streaming/turn-level repetition detectors to abort degenerate agent outputs. ChangesTask Session Threading & UI Integration
Backend session persistence & todos CRUD
Streaming and turn-level repetition detection
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Possibly related issues
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. Comment |
coderabbitai
Bot
added
feature
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 4
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
src/openhuman/inference/provider/compatible.rs (1)
1393-1403:⚠️ Potential issue | 🟡 Minor | ⚡ Quick winFlush the pending line at stream end before deciding no degeneration occurred.
At Line 1397, only newline-terminated lines have been counted so far. If the stream ends on a repeated substantial line without a trailing
\n, the last repeat is never evaluated and the cutoff can be missed.💡 Proposed fix
- if degenerate_repeat { + if !degenerate_repeat && repeat_detector.finalize_line() { + log::warn!( + "[stream] {} degenerate repetition detected (≥{} identical lines) — aborting generation, truncating (text_chars={})", + self.name, + STREAM_REPEAT_THRESHOLD, + text_accum.chars().count(), + ); + degenerate_repeat = true; + } + + if degenerate_repeat { // Mark the truncated output so downstream (and the user) see why it // was cut off rather than a silently shortened response. text_accum.push_str( "\n\n[Output stopped: detected repeated/looping generation (model degeneration).]", ); }🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@src/openhuman/inference/provider/compatible.rs` around lines 1393 - 1403, Before you check degenerate_repeat and append to text_accum, flush the last buffered line used for repetition detection: take the accumulator that collects characters until a '\n' (the "current/pending line" used by the repetition logic), count/update its occurrence in the repetition tracking structures exactly like you do when a newline is encountered, and update degenerate_repeat accordingly; then proceed to append the degeneration notice to text_accum if degenerate_repeat is true. Ensure you reference and reuse the same repetition-counting logic/function used on newline termination so behavior is consistent.
🧹 Nitpick comments (2)
app/src/components/settings/hooks/useSettingsNavigation.ts (1)
48-48: 💤 Low valueMissing breadcrumb case for
task-sourcesroute.The route is added to the union and URL detection, but
getBreadcrumbs()has no case for'task-sources', so it falls through to the default (empty breadcrumbs). Sibling routes under the same settings section (e.g.,composio-triggers,composio-routing) return[settingsCrumb, developerCrumb]. Consider adding'task-sources'to the same case block for consistent navigation breadcrumbs.🔧 Proposed fix
case 'composio-triggers': case 'composio-routing': case 'notification-routing': case 'mcp-server': case 'dev-workflow': case 'notifications-hub': // Notifications hub section page lives under Advanced. + case 'task-sources': return [settingsCrumb, developerCrumb];Also applies to: 119-119
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@app/src/components/settings/hooks/useSettingsNavigation.ts` at line 48, getBreadcrumbs() in useSettingsNavigation.ts is missing a case for the 'task-sources' route so it falls back to the default empty breadcrumbs; update the switch inside getBreadcrumbs() to include 'task-sources' in the same case branch that returns [settingsCrumb, developerCrumb] (the same branch currently handling 'composio-triggers' and 'composio-routing'), and mirror this addition wherever getBreadcrumbs() performs route matching for consistency across the file.src/openhuman/todos/ops.rs (1)
375-391: ⚡ Quick winAdd entry-level debug context to
set_session_thread.This new mutation path currently has no debug trace, which makes session-link race/debug triage harder than neighboring ops.
♻️ Suggested patch
pub fn set_session_thread( location: &BoardLocation, id: &str, session_thread_id: Option<String>, ) -> Result<TodosSnapshot, String> { + tracing::debug!( + thread_id = ?location.thread_id(), + id, + session_thread_id = ?session_thread_id, + "[todos][ops] set_session_thread entry" + ); let _scratch_guard = maybe_scratch_lock(location); let mut cards = load_cards(location)?;As per coding guidelines: “Add substantial debug-level logs while implementing features or fixes in Rust using
log/tracingcrate with stable prefixes and correlation fields.”🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@src/openhuman/todos/ops.rs` around lines 375 - 391, Add structured debug-level tracing around set_session_thread: emit a tracing::debug! (or log::debug!) at function entry with a stable prefix like "todos::set_session_thread" and correlation fields location (BoardLocation or its ID), id, and incoming session_thread_id; capture the previous card.session_thread_id after load_cards/find (reference the card found via .find in set_session_thread) and log a second debug message showing the before->after change and any error returned by save_cards, and a final debug on successful emit_progress/into_snapshot; keep messages concise and use structured fields for easy filtering.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@app/src/pages/Conversations.tsx`:
- Around line 1614-1619: The onViewSession handler currently calls
dispatch(setActiveThread(card.sessionThreadId)) which incorrectly marks
historical/completed sessions as the global active thread; remove that dispatch
(leave dispatch(setSelectedThread(card.sessionThreadId)) and
dispatch(loadThreadMessages(card.sessionThreadId))) or, if you must support
conditionally activating, only call setActiveThread when the card indicates an
in-flight turn (e.g., card.isInFlight or card.status === 'in-progress'); update
the onViewSession callback to avoid setting activeThreadId for navigation-only
opens.
In `@src/openhuman/agent/harness/engine/core.rs`:
- Around line 521-523: The early-return path that appends the assistant message
(using assistant_history_content and history.push(...)) is missing the
corresponding observer.on_assistant(...) call, which other assistant-append
paths invoke; update that branch in core.rs to call
observer.on_assistant(assistant_history_content.clone(), iteration) (or the same
signature used elsewhere) immediately after pushing the assistant message and
before observer.after_iteration/history/return so observer-side
transcript/mirroring behavior remains consistent with other flows. Ensure you
use the same parameter types and ordering as existing observer.on_assistant
usages and preserve the existing calls to observer.after_iteration(...) and
progress.turn_completed(...).
In `@src/openhuman/agent/task_board.rs`:
- Around line 90-95: The session_thread_id field is not being normalized like
other optional string fields; update the Card normalization logic (the
function/method that trims and converts empty optional strings to None) to
include session_thread_id so it is trimmed and set to None when empty or
all-whitespace. Locate the struct field session_thread_id and add the same
trim-and-empty-check used for other Option<String> fields (reuse the existing
helper or pattern used elsewhere in the file) so whitespace values do not
persist as invalid targets.
In `@src/openhuman/agent/task_dispatcher.rs`:
- Around line 387-415: The spawned task can complete before ACTIVE_RUNS is
registered, causing write_back to be skipped; fix by gating task start with a
oneshot start signal: create a tokio::sync::oneshot channel (e.g.
start_tx/start_rx) before spawning, pass start_rx into the async move block and
await it immediately before calling run_autonomous, then call
register_active_run(...) (using join.abort_handle()) and finally send start_tx
to release the task; update references in the task body (run_autonomous,
hb_cancel_for_task, task_thread, take_active_run, write_back) to preserve
existing cancel/write-back logic after the start await.
---
Outside diff comments:
In `@src/openhuman/inference/provider/compatible.rs`:
- Around line 1393-1403: Before you check degenerate_repeat and append to
text_accum, flush the last buffered line used for repetition detection: take the
accumulator that collects characters until a '\n' (the "current/pending line"
used by the repetition logic), count/update its occurrence in the repetition
tracking structures exactly like you do when a newline is encountered, and
update degenerate_repeat accordingly; then proceed to append the degeneration
notice to text_accum if degenerate_repeat is true. Ensure you reference and
reuse the same repetition-counting logic/function used on newline termination so
behavior is consistent.
---
Nitpick comments:
In `@app/src/components/settings/hooks/useSettingsNavigation.ts`:
- Line 48: getBreadcrumbs() in useSettingsNavigation.ts is missing a case for
the 'task-sources' route so it falls back to the default empty breadcrumbs;
update the switch inside getBreadcrumbs() to include 'task-sources' in the same
case branch that returns [settingsCrumb, developerCrumb] (the same branch
currently handling 'composio-triggers' and 'composio-routing'), and mirror this
addition wherever getBreadcrumbs() performs route matching for consistency
across the file.
In `@src/openhuman/todos/ops.rs`:
- Around line 375-391: Add structured debug-level tracing around
set_session_thread: emit a tracing::debug! (or log::debug!) at function entry
with a stable prefix like "todos::set_session_thread" and correlation fields
location (BoardLocation or its ID), id, and incoming session_thread_id; capture
the previous card.session_thread_id after load_cards/find (reference the card
found via .find in set_session_thread) and log a second debug message showing
the before->after change and any error returned by save_cards, and a final debug
on successful emit_progress/into_snapshot; keep messages concise and use
structured fields for easy filtering.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 9d1908e4-5754-4588-92d0-321d9e400fa9
📒 Files selected for processing (36)
app/src/components/intelligence/IntelligenceTasksTab.tsxapp/src/components/settings/hooks/useSettingsNavigation.tsapp/src/lib/i18n/ar.tsapp/src/lib/i18n/bn.tsapp/src/lib/i18n/de.tsapp/src/lib/i18n/en.tsapp/src/lib/i18n/es.tsapp/src/lib/i18n/fr.tsapp/src/lib/i18n/hi.tsapp/src/lib/i18n/id.tsapp/src/lib/i18n/it.tsapp/src/lib/i18n/ko.tsapp/src/lib/i18n/pl.tsapp/src/lib/i18n/pt.tsapp/src/lib/i18n/ru.tsapp/src/lib/i18n/zh-CN.tsapp/src/pages/Conversations.tsxapp/src/pages/Intelligence.tsxapp/src/pages/conversations/components/TaskKanbanBoard.tsxapp/src/pages/conversations/utils/threadFilter.tsapp/src/services/api/todosApi.tsapp/src/types/turnState.tssrc/openhuman/agent/harness/engine/core.rssrc/openhuman/agent/harness/tool_loop.rssrc/openhuman/agent/harness/tool_loop_tests.rssrc/openhuman/agent/mod.rssrc/openhuman/agent/task_board.rssrc/openhuman/agent/task_dispatcher.rssrc/openhuman/agent/task_session.rssrc/openhuman/channels/providers/web.rssrc/openhuman/inference/provider/compatible.rssrc/openhuman/inference/provider/compatible_tests.rssrc/openhuman/task_sources/pipeline.rssrc/openhuman/threads/turn_state/mirror_tests.rssrc/openhuman/todos/ops.rssrc/openhuman/todos/schemas.rs
CodeRabbit review fixes on PR tinyhumansai#3380: - core.rs: call observer.on_assistant() on the repeat-guard early-exit so transcript/mirroring isn't skipped for the final repeated iteration. - task_dispatcher.rs: gate the spawned run on a oneshot until register_active_run() completes, so a fast-finishing run can't race past ACTIVE_RUNS registration and drop its terminal write_back. - task_board.rs: normalise session_thread_id (trim + empty-filter) like the other optional string fields, so whitespace can't survive as a blank "View session" target. Covered by the roundtrip normalisation test. - Conversations.tsx / IntelligenceTasksTab.tsx: drop setActiveThread from the view-session navigation handlers — activeThreadId tracks a true in-flight turn, and forcing it on a completed session wedges the composer. Also fix the Rust coverage build: six TaskBoardCard literals in the raw_coverage_e2e integration tests were missing the new session_thread_id field (cargo check --lib passed but --all-targets / cargo-llvm-cov failed). Re-orders an import in TaskKanbanBoard.tsx to satisfy prettier. Co-Authored-By: Claude <noreply@anthropic.com>
… wiring The Vitest coverage lane (previously skipped behind the prettier failure) now runs and surfaced two stale tests against my component changes: - IntelligenceTasksTab: the "Work" flow now awaits todosApi.setSessionThread(...), which the mock didn't stub — the awaited .catch() chain stalled the handler before chatSend. Add the mock (resolved by default in beforeEach) and wait for the trailing chatSend assertion. - TaskKanbanBoard: TaskSourceControls now calls useNavigate(); the router-less render threw "useNavigate() may be used only in the context of a <Router>". Mock react-router-dom's useNavigate to a no-op (matching the sibling suite). Co-Authored-By: Claude <noreply@anthropic.com>
Contributor
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@tests/owned_domain_raw_coverage_e2e.rs`:
- Line 584: Add assertions that verify session_thread_id is persisted and
normalized after the test's put/get cycle: assert
saved.cards[0].session_thread_id equals the value you set (and for the
whitespace/empty case assert the normalized value, e.g. None or trimmed string)
and also assert the loaded board/cards reflect the same session_thread_id;
update the test around where session_thread_id is set (check
saved.cards[0].session_thread_id and the loaded board/cards session_thread_id)
so the contract is locked for both normal and normalization cases.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 74c03f53-61b7-4ad3-b46b-500bc63d4679
📒 Files selected for processing (12)
app/src/components/intelligence/IntelligenceTasksTab.tsxapp/src/components/intelligence/__tests__/IntelligenceTasksTab.test.tsxapp/src/pages/Conversations.tsxapp/src/pages/conversations/components/TaskKanbanBoard.test.tsxapp/src/pages/conversations/components/TaskKanbanBoard.tsxsrc/openhuman/agent/harness/engine/core.rssrc/openhuman/agent/task_board.rssrc/openhuman/agent/task_dispatcher.rstests/agent_triage_dispatch_round23_raw_coverage_e2e.rstests/inference_agent_raw_coverage_e2e.rstests/memory_threads_raw_coverage_e2e.rstests/owned_domain_raw_coverage_e2e.rs
🚧 Files skipped from review as they are similar to previous changes (3)
- app/src/pages/conversations/components/TaskKanbanBoard.tsx
- app/src/components/intelligence/IntelligenceTasksTab.tsx
- src/openhuman/agent/task_dispatcher.rs
…thread persistence - channels runtime: the max-tool-iterations tests drove IterativeToolProvider to emit byte-identical assistant output every turn, which now trips the cross-iteration RepeatOutputGuard (added in this PR) and halts at 4 before reaching the configured 11/20 iterations. Prefix a per-iteration progress note so each turn is distinct — modelling a healthy progressing agent — while the guard still catches genuine verbatim-repeat degeneration. Fixes the lone lib-test failure surfaced once the coverage build started passing. - owned_domain coverage test: set a padded session_thread_id on the surviving card and assert it trims through put() and survives a reload (addresses a CodeRabbit review note that the value was set but never verified). Co-Authored-By: Claude <noreply@anthropic.com>
Co-Authored-By: Claude <noreply@anthropic.com>
…ver gate The diff-cover gate flagged the new frontend wiring at 36% (Rust side was fully covered). Add focused Vitest coverage for the changed handlers: - todosApi.setSessionThread — new RPC client method (params + null clear). - TaskKanbanBoard — "View work" button shows only with a session thread and fires onViewSession; "Manage sources" navigates to settings. - IntelligenceTasksTab — "View work" opens the exact session via navigate state; the work flow still dispatches the turn when session-linking rejects; "Manage sources" navigation. - Intelligence — URL-backed tab selection (default, ?tab=, unknown fallback, switch-updates-URL). - Conversations — location.state.openThreadId resume opens a hidden task session on mount. The idempotent-approve branch is intentionally left to its unit-unreachable state: the inbox hide (pickedUpExternalIds) removes an already-promoted source card before approve can run, so that branch is a pure race safety-net. Co-Authored-By: Claude <noreply@anthropic.com>
Co-Authored-By: Claude <noreply@anthropic.com>
…k-sessions # Conflicts: # app/src/components/intelligence/IntelligenceTasksTab.tsx # app/src/components/intelligence/__tests__/IntelligenceTasksTab.test.tsx
…dler Lifts diff-coverage over the 80% gate: seeds a selected thread's task board with a card carrying a sessionThreadId, clicks "View work", and asserts the handler navigates the chat view to that card's session thread. Co-Authored-By: Claude <noreply@anthropic.com>
Contributor
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@app/src/lib/i18n/de.ts`:
- Line 4713: Update the German locale entry for the intelligence agents tier by
replacing the English string; in app/src/lib/i18n/de.ts locate the key
'intelligence.agents.tier.reasoning' and change its value from 'Reasoning' to an
appropriate German translation such as 'Schlussfolgerung' (or
'Argumentation'/'Logisches Denken' if context prefers) so the de.ts file
contains a real German translation for that key.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: c2b2df06-a9ff-434c-b214-d4bd7b15b44d
📒 Files selected for processing (29)
app/src/components/intelligence/IntelligenceTasksTab.tsxapp/src/components/intelligence/__tests__/IntelligenceTasksTab.test.tsxapp/src/lib/i18n/ar.tsapp/src/lib/i18n/bn.tsapp/src/lib/i18n/de.tsapp/src/lib/i18n/en.tsapp/src/lib/i18n/es.tsapp/src/lib/i18n/fr.tsapp/src/lib/i18n/hi.tsapp/src/lib/i18n/id.tsapp/src/lib/i18n/it.tsapp/src/lib/i18n/ko.tsapp/src/lib/i18n/pl.tsapp/src/lib/i18n/pt.tsapp/src/lib/i18n/ru.tsapp/src/lib/i18n/zh-CN.tsapp/src/pages/__tests__/Conversations.render.test.tsxapp/src/pages/__tests__/Intelligence.test.tsxapp/src/pages/conversations/components/TaskKanbanBoard.test.tsxapp/src/services/api/__tests__/todosApi.test.tssrc/openhuman/agent/harness/engine/core.rssrc/openhuman/agent/mod.rssrc/openhuman/channels/providers/web.rssrc/openhuman/channels/tests/common.rssrc/openhuman/inference/provider/compatible.rssrc/openhuman/inference/provider/compatible_tests.rssrc/openhuman/task_sources/pipeline.rstests/memory_threads_raw_coverage_e2e.rstests/owned_domain_raw_coverage_e2e.rs
💤 Files with no reviewable changes (3)
- tests/memory_threads_raw_coverage_e2e.rs
- tests/owned_domain_raw_coverage_e2e.rs
- src/openhuman/task_sources/pipeline.rs
✅ Files skipped from review due to trivial changes (10)
- app/src/lib/i18n/en.ts
- app/src/lib/i18n/fr.ts
- app/src/lib/i18n/ar.ts
- app/src/lib/i18n/it.ts
- app/src/lib/i18n/es.ts
- app/src/lib/i18n/pl.ts
- app/src/lib/i18n/hi.ts
- app/src/lib/i18n/pt.ts
- app/src/lib/i18n/bn.ts
- app/src/lib/i18n/ru.ts
🚧 Files skipped from review as they are similar to previous changes (9)
- app/src/lib/i18n/zh-CN.ts
- src/openhuman/inference/provider/compatible_tests.rs
- src/openhuman/agent/mod.rs
- app/src/pages/conversations/components/TaskKanbanBoard.test.tsx
- app/src/lib/i18n/id.ts
- app/src/lib/i18n/ko.ts
- src/openhuman/agent/harness/engine/core.rs
- app/src/components/intelligence/IntelligenceTasksTab.tsx
- src/openhuman/inference/provider/compatible.rs
Contributor
There was a problem hiding this comment.
Caution
Inline review comments failed to post. This is likely due to GitHub's internal server error or limits when posting large numbers of comments. If you are seeing this consistently it is likely a permissions issue. Please check "Moderation" -> "Code review limits" under your organization settings.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@app/src/lib/i18n/de.ts`:
- Line 4713: Update the German locale entry for the intelligence agents tier by
replacing the English string; in app/src/lib/i18n/de.ts locate the key
'intelligence.agents.tier.reasoning' and change its value from 'Reasoning' to an
appropriate German translation such as 'Schlussfolgerung' (or
'Argumentation'/'Logisches Denken' if context prefers) so the de.ts file
contains a real German translation for that key.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: c2b2df06-a9ff-434c-b214-d4bd7b15b44d
📒 Files selected for processing (29)
app/src/components/intelligence/IntelligenceTasksTab.tsxapp/src/components/intelligence/__tests__/IntelligenceTasksTab.test.tsxapp/src/lib/i18n/ar.tsapp/src/lib/i18n/bn.tsapp/src/lib/i18n/de.tsapp/src/lib/i18n/en.tsapp/src/lib/i18n/es.tsapp/src/lib/i18n/fr.tsapp/src/lib/i18n/hi.tsapp/src/lib/i18n/id.tsapp/src/lib/i18n/it.tsapp/src/lib/i18n/ko.tsapp/src/lib/i18n/pl.tsapp/src/lib/i18n/pt.tsapp/src/lib/i18n/ru.tsapp/src/lib/i18n/zh-CN.tsapp/src/pages/__tests__/Conversations.render.test.tsxapp/src/pages/__tests__/Intelligence.test.tsxapp/src/pages/conversations/components/TaskKanbanBoard.test.tsxapp/src/services/api/__tests__/todosApi.test.tssrc/openhuman/agent/harness/engine/core.rssrc/openhuman/agent/mod.rssrc/openhuman/channels/providers/web.rssrc/openhuman/channels/tests/common.rssrc/openhuman/inference/provider/compatible.rssrc/openhuman/inference/provider/compatible_tests.rssrc/openhuman/task_sources/pipeline.rstests/memory_threads_raw_coverage_e2e.rstests/owned_domain_raw_coverage_e2e.rs
💤 Files with no reviewable changes (3)
- tests/memory_threads_raw_coverage_e2e.rs
- tests/owned_domain_raw_coverage_e2e.rs
- src/openhuman/task_sources/pipeline.rs
✅ Files skipped from review due to trivial changes (10)
- app/src/lib/i18n/en.ts
- app/src/lib/i18n/fr.ts
- app/src/lib/i18n/ar.ts
- app/src/lib/i18n/it.ts
- app/src/lib/i18n/es.ts
- app/src/lib/i18n/pl.ts
- app/src/lib/i18n/hi.ts
- app/src/lib/i18n/pt.ts
- app/src/lib/i18n/bn.ts
- app/src/lib/i18n/ru.ts
🚧 Files skipped from review as they are similar to previous changes (9)
- app/src/lib/i18n/zh-CN.ts
- src/openhuman/inference/provider/compatible_tests.rs
- src/openhuman/agent/mod.rs
- app/src/pages/conversations/components/TaskKanbanBoard.test.tsx
- app/src/lib/i18n/id.ts
- app/src/lib/i18n/ko.ts
- src/openhuman/agent/harness/engine/core.rs
- app/src/components/intelligence/IntelligenceTasksTab.tsx
- src/openhuman/inference/provider/compatible.rs
🛑 Comments failed to post (1)
app/src/lib/i18n/de.ts (1)
4713-4713:
⚠️ Potential issue | 🟡 Minor | ⚡ Quick winTranslate the new tier label in the German locale.
'intelligence.agents.tier.reasoning': 'Reasoning'is still English inde.ts. Please localize it to keep locale completeness consistent.As per coding guidelines:
app/src/lib/i18n/*.ts: "Ensure all i18n locale files have real translations for every key in all 13 supported locales."🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@app/src/lib/i18n/de.ts` at line 4713, Update the German locale entry for the intelligence agents tier by replacing the English string; in app/src/lib/i18n/de.ts locate the key 'intelligence.agents.tier.reasoning' and change its value from 'Reasoning' to an appropriate German translation such as 'Schlussfolgerung' (or 'Argumentation'/'Logisches Denken' if context prefers) so the de.ts file contains a real German translation for that key.
Contributor
Author
|
@coderabbitai resolve |
Contributor
✅ Action performedComments resolved and changes approved. |
12 tasks
senamakel
pushed a commit
to senamakel/openhuman
that referenced
this pull request
Jun 6, 2026
…rk, cancel, degeneration guards) (tinyhumansai#3380) Co-authored-by: Claude <noreply@anthropic.com>
12 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
client_id="system"+TurnStateMirror) and a card's "View work" opens the exact session.ACTIVE_RUNSregistry +cancel_sessionwired intochannel_web_cancel).RepeatOutputGuard) + in-generation streaming cutoff (StreamRepeatDetector).[task_sources:dedup]debug logging.Problem
Autonomous task runs ran headless — no UI visibility, no way to watch or cancel them, and on any model they could spiral into repeated output until the token cap (frequency_penalty / model tier only reduce severity). Approving a task-source item could also duplicate it onto the user-tasks board.
Solution
task_sessionmodule materialises a top-levellabels:["tasks"]thread per run and streams turns via the existing web-channelspawn_progress_bridge(reused, not reinvented), finalised with achat_doneevent. Cards carrysessionThreadId(todos_set_session_thread) → "View work" opens the exact thread on the Tasks tab.ACTIVE_RUNSregistry holds each run'sAbortHandle;channel_web_cancelfalls back tocancel_sessionfor task threads, with race-safe board write-back.RepeatOutputGuardhalts the tool loop on repeated output across iterations;StreamRepeatDetectoraborts the SSE read when a substantial line repeats within one generation. Both deterministic + model-agnostic.external_id(todos_addnow persistssource_metadata); the inbox hides picked-up items and approve dedups against an existing card.Submission Checklist
set_session_thread,task_session, both repeat guards,StreamRepeatDetector, and the cancel registry (active_run_registry_take_is_once).Impact
Desktop. Touches the agent-harness turn engine + the OpenAI-compatible streaming provider (the in-generation cutoff) — broad surface, but the cutoff only triggers on degenerate repetition and is off the happy path. No migrations; default
frequency_penaltyunchanged; headless runs still work when no session thread is created.Related
AI Authored PR Metadata
Linear Issue
Commit & Branch
Validation Run
pnpm typecheck— 0 errorscargo check --lib— 0 errors; focused unit tests passcargo fmton changed filesformat:check/ lint — prettier + rustfmt pass on changed files; the prettier import-order miss that failed CI is fixed in this revisionBehavior Changes
Parity Contract
frequency_penaltydefault unchanged;RepeatFailureGuardsemantics untouched.Duplicate / Superseded PR Handling
🤖 Generated with Claude Code
Validation Blocked
command:pre-push hookpnpm lint:commands-tokenserror:requiresripgrep(rg), which is not installed in this environmentimpact:unrelated to this change (TS typecheck + Rust check pass); pushed with--no-verify. Re-run the token lint in CI / on a machine with ripgrep.Summary by CodeRabbit
New Features
Bug Fixes
Improvements