fix(chat): render agent event tool results reliably#4524
Conversation
…persist them in turn-state
- tool_result socket events now carry the size-capped tool output (matching
the subagent_tool_result path) instead of a {output_chars, elapsed_ms}
metadata stub — the frontend's ChatToolResultEvent.output and the
propose_workflow proposal parser already expected the real payload.
- ToolTimelineEntry / SubagentToolCall snapshots persist a 64KiB-capped
output so the rehydrated 'View processing' timeline can show what each
tool returned after a thread switch or cold boot.
- The tinyagents event bridge's silent catch-all now trace-logs dropped
event kinds.
Claude-Session: https://claude.ai/code/session_011S48qZxBBHUbaYMZTngZ5M
…switches, harden chat event subscription - ToolTimelineEntry gains a 'result' field: set live from the tool_result socket event's output and rehydrated from the persisted snapshot's new 'output' field; ToolTimelineBlock renders it as a scrollable result block. - hydrateRuntimeFromSnapshot no longer clobbers a thread whose live socket driver is mid-turn: the global provider keeps feeding Redux while the user is on another tab, so applying a flush-boundary snapshot wiped streamed prose, tool results, and pending approval cards. Snapshots now apply only when there is no live lifecycle (cold boot / new window / interrupted). - The chat message pane now renders when live agent activity exists even before thread history loads (previously gated solely on hasVisibleMessages, blanking tool calls + streaming on thread switch). - subscribeChatEvents registers through the socketService wrapper so chat listeners queue while the socket is connecting instead of silently no-opping, and re-attach on reconnect. Claude-Session: https://claude.ai/code/session_011S48qZxBBHUbaYMZTngZ5M
…live-driver hydration guard - progress_bridge: tokio test asserting the tool_result wire event carries the real output (not the metadata stub). - turn_state mirror: parent + subagent output persistence, 64KiB cap with char-boundary truncation, empty-output elision. - chatRuntimeSlice: live-driver guard (tab-switch no-clobber, cold-boot apply) and persisted output → result mapping. - ChatRuntimeProvider: tool_result attaches output as the row result; empty output leaves it unset. ToolTimelineBlock renders the result block. - fix stale observability test: vendored tinyagents dropped ToolCompleted.started_at_ms (pre-existing lib-test compile breakage). Claude-Session: https://claude.ai/code/session_011S48qZxBBHUbaYMZTngZ5M
…by taskId With the live-driver hydration guard, live rows keep their socket-path entry ids, so the ledger merge (which keys rows as subagent:<runId>) could add a second row for a delegation already on screen. Dedupe on the subagent taskId. Claude-Session: https://claude.ai/code/session_011S48qZxBBHUbaYMZTngZ5M
📝 WalkthroughWalkthroughThis PR persists and surfaces tool call output end-to-end: Rust turn-state mirror/types now cap and store tool result text, the web progress bridge forwards real output instead of a metadata stub, and frontend runtime/UI map, hydrate, and render that result in the tool timeline. Conversations.tsx gains a live-activity flag to prevent empty-state flicker. Unrelated README anchor fixes are included. ChangesTool output persistence and display
CONTRIBUTING-BEGINNERS anchor fix
Estimated code review effort: 4 (Complex) | ~60 minutes Sequence Diagram(s)sequenceDiagram
participant Agent as Tool Execution
participant Bridge as progress_bridge.rs
participant Mirror as TurnStateMirror
participant Store as ChatRuntimeProvider/Slice
participant UI as ToolTimelineBlock
Agent->>Bridge: ToolCallCompleted(output)
Bridge->>Bridge: cap_wire_output(output)
Bridge-->>Store: tool_result event(output)
Store->>Store: derive result from event.output (if non-empty)
Store->>Store: update ToolTimelineEntry.result
Agent->>Mirror: ToolCallCompleted/SubagentToolCallCompleted(output)
Mirror->>Mirror: cap_persisted_output(output)
Mirror->>Mirror: persist output on entry/subagent call
Store->>Store: hydrateRuntimeFromSnapshot
alt live driver active (started/streaming)
Store->>Store: skip overwrite, apply taskBoard only
else no live driver
Store->>Store: map persisted output -> result, apply snapshot
end
Store-->>UI: ToolTimelineEntry with result
UI->>UI: render resultContent in expandable <pre>
Possibly related PRs
Suggested labels: Suggested reviewers: Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
Comment |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 885d885f24
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
coderabbitai
Bot
added
agent
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 08f18ccd66
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
🧹 Nitpick comments (2)
src/openhuman/channels/providers/web/progress_bridge.rs (1)
468-514: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low value
MAX_WIRE_SUBAGENT_OUTPUTname no longer matches its usage.This constant now also caps the main-agent
tool_resultoutput (previously only the subagentsubagent_tool_resultpath used it). Consider renaming to something scope-neutral (e.g.MAX_WIRE_TOOL_OUTPUT) to avoid confusing future changes that only intend to affect subagent behavior.🤖 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/channels/providers/web/progress_bridge.rs` around lines 468 - 514, The constant name is too subagent-specific even though it is now used in the main-agent tool result flow in progress_bridge.rs. Rename MAX_WIRE_SUBAGENT_OUTPUT to a scope-neutral name like MAX_WIRE_TOOL_OUTPUT, and update its uses in cap_wire_output and the WebChannelEvent tool_result/subagent_tool_result paths so the name matches the shared behavior.src/openhuman/threads/turn_state/mirror.rs (1)
27-58: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winDuplicated truncation logic vs.
progress_bridge.rs::cap_wire_output.
cap_persisted_outputhere andcap_wire_outputinsrc/openhuman/channels/providers/web/progress_bridge.rsimplement the same char-boundary truncation-with-marker algorithm (each with its ownTRUNCATION_MARKER_BUDGETconstant and cap). Consider extracting a shared helper (e.g.fn truncate_utf8_with_marker(s: &str, max_len: usize, marker_budget: usize) -> Cow<str>) into a common module so the two call sites can't silently diverge (e.g. only one handles the empty-output →Nonecase today).🤖 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/threads/turn_state/mirror.rs` around lines 27 - 58, `cap_persisted_output` in `mirror.rs` duplicates the same UTF-8 truncation-and-marker behavior already implemented by `cap_wire_output` in `progress_bridge.rs`, so refactor both call sites to use a shared helper in a common module. Extract the char-boundary truncation logic and marker budgeting into one reusable function (for example, a helper returning a `Cow<str>`), then adapt `cap_persisted_output` and `cap_wire_output` to preserve their existing empty-output behavior while relying on that shared implementation.
🤖 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.
Nitpick comments:
In `@src/openhuman/channels/providers/web/progress_bridge.rs`:
- Around line 468-514: The constant name is too subagent-specific even though it
is now used in the main-agent tool result flow in progress_bridge.rs. Rename
MAX_WIRE_SUBAGENT_OUTPUT to a scope-neutral name like MAX_WIRE_TOOL_OUTPUT, and
update its uses in cap_wire_output and the WebChannelEvent
tool_result/subagent_tool_result paths so the name matches the shared behavior.
In `@src/openhuman/threads/turn_state/mirror.rs`:
- Around line 27-58: `cap_persisted_output` in `mirror.rs` duplicates the same
UTF-8 truncation-and-marker behavior already implemented by `cap_wire_output` in
`progress_bridge.rs`, so refactor both call sites to use a shared helper in a
common module. Extract the char-boundary truncation logic and marker budgeting
into one reusable function (for example, a helper returning a `Cow<str>`), then
adapt `cap_persisted_output` and `cap_wire_output` to preserve their existing
empty-output behavior while relying on that shared implementation.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: c26f03cc-6e10-4839-b825-7d644ac10859
📒 Files selected for processing (22)
app/src/pages/Conversations.tsxapp/src/pages/conversations/components/ToolTimelineBlock.tsxapp/src/pages/conversations/components/__tests__/ToolTimelineBlock.test.tsxapp/src/providers/ChatRuntimeProvider.tsxapp/src/providers/__tests__/ChatRuntimeProvider.test.tsxapp/src/services/chatService.tsapp/src/store/chatRuntimeSlice.test.tsapp/src/store/chatRuntimeSlice.tsapp/src/types/turnState.tsdocs/README.de.mddocs/README.ja-JP.mddocs/README.ko.mddocs/README.ur-pk.mddocs/README.zh-CN.mdsrc/openhuman/channels/providers/web/progress_bridge.rssrc/openhuman/threads/turn_state/mirror.rssrc/openhuman/threads/turn_state/mirror_tests.rssrc/openhuman/threads/turn_state/store_tests.rssrc/openhuman/threads/turn_state/types.rssrc/openhuman/tinyagents/observability.rstests/memory_raw_coverage_e2e.rstests/memory_threads_raw_coverage_e2e.rs
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 3c9ac927a2
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
1 participant
Summary
Problem
Solution
Submission Checklist
diff-cover) meet the gate enforced by.github/workflows/pr-ci.yml. CI will enforce the final changed-line gate for this PR.docs/TEST-COVERAGE-MATRIX.md.Impact
Related
AI Authored PR Metadata (required for Codex/Linear PRs)
Linear Issue
Commit & Branch
Validation Run
pnpm --filter openhuman-app format:checkpnpm typecheckcargo fmt --manifest-path ../Cargo.toml --all --checkvia app format hookcargo fmt --manifest-path app/src-tauri/Cargo.toml --all --check;pnpm rust:checkValidation Blocked
command:N/Aerror:N/Aimpact:N/ABehavior Changes
Parity Contract
Duplicate / Superseded PR Handling
Note: initial
git pushran the pre-push hook through format, lint, typecheck, rust:check, and command-token lint successfully, then GitHub closed the SSH sideband after the long hook. The retry used--no-verifyonly to avoid re-running already completed local checks before uploading the same commit.Summary by CodeRabbit