feat(agents-server-ui): stream model reasoning into the UI by kevin-dp · Pull Request #4508 · electric-sql/electric

kevin-dp · 2026-06-04T13:36:02Z

Summary

While the model is "thinking" (Anthropic extended thinking, DeepSeek-R1 reasoning, Moonshot K2, OpenAI Responses summaries) the agent response now shows the reasoning text faded above the answer, with the existing Thinking shimmer heading plus elapsed-time ticker. Once the reasoning settles it collapses to ▸ Thought for 12s — click to expand. Multiple reasoning rows per run render independently in order (one per LLM step in tool-using turns). UX intentionally mirrors Claude Code + OpenCode patterns.

Implementation (end-to-end)

Schema — reasoning row gains run_id, encrypted (Anthropic redacted-thinking opaque payload, must round-trip back to the model verbatim), and summary_title (extracted at write time). New reasoningDeltas collection mirrors textDeltas. Strictly additive.
Bridge — OutboundBridge gains onReasoningStart / onReasoningDelta / onReasoningEnd, parallel to the text path. Reasoning counter added to OutboundIdSeed.
Adapter — pi-adapter.ts routes pi-ai's thinking_start / thinking_delta / thinking_end events to the bridge. Parses a **Title**\n\n<body> heading once at write time (OpenAI Responses; no-op for Anthropic / DeepSeek / Moonshot). Defensive: handles late thinking_delta without a preceding thinking_start, and closes an open reasoning row on message_end (e.g. provider abort).
Timeline — Live reasoning: Collection<EntityTimelineReasoningItem> on EntityTimelineRunRow, content built via the same delta-join pattern as EntityTimelineTextItem.content.
UI — New <ReasoningSection> renders above items in AgentResponseLive:
- Live: faded markdown via Streamdown with ThinkingIndicator heading + summary title + elapsed-time ticker
- Settled: ▸ Thought for Ns with click-to-expand. Closure duration snapshotted from Date.now() - timestamp using the same sawStreamingRef trick from the elapsed-time PR — accurate for in-session settles, stays a bare Thought for rows already settled on first mount (no real end timestamp available client-side).
- Redacted: Anthropic safety-filter payloads render ⊘ Reasoning redacted by provider safety filters. The encrypted payload is still persisted server-side so the model gets it back on the next turn.

Reference

Patterns informed by reading OpenCode's reasoning implementation:

3-event streaming protocol (reasoning-start / reasoning-delta / reasoning-end)
ReasoningPart storage shape including encrypted for Anthropic round-trip
reasoningSummary() headline parser (5-line regex, OpenAI Responses only)
Collapsed-by-default UX with click-to-expand

Test plan

pnpm typecheck clean in agents-runtime + agents-server-ui
pnpm test outbound-bridge pi-adapter entity-timeline in agents-runtime (95 passed: 18 bridge + 21 adapter + 56 timeline)
pnpm test in agents-server-ui (66 passed)
pnpm -C packages/agents-runtime build — dist artifacts emit cleanly
Manual: prompt Anthropic Claude with extended-thinking enabled; verify streaming reasoning appears faded above the answer with elapsed ticker, then collapses to Thought for Ns on settle
Manual: multi-step tool-using turn; verify each step's reasoning renders as a separate collapsible row

Notes

Cached AgentResponse (the non-Live path used for old scrollback sections) doesn't yet surface reasoning — historical rows recorded before this PR lack the data anyway. Follow-up if we discover sessions where this matters.
The pre-existing runtime-dsl.test.ts 401 failures (and dispatch-policy-routing.test.ts 500 failures) reproduce identically on clean main and were not introduced by this PR.

🤖 Generated with Claude Code

While the model is "thinking" (Anthropic extended thinking, DeepSeek-R1 reasoning_content, Moonshot K2, OpenAI Responses summaries) the agent response now shows the reasoning text faded above the answer, with the existing `Thinking` shimmer heading + elapsed-time ticker. Once the reasoning settles, it collapses to `▸ Thought for 12s` — click to expand. Multiple reasoning rows per run render independently in order (one per LLM step in tool-using turns). End-to-end plumbing: - Schema: `reasoning` row gains `run_id`, `encrypted` (Anthropic redacted blocks must round-trip back to the model), and `summary_title` (extracted at write time). New `reasoningDeltas` collection mirrors `textDeltas` for streamed content. - Bridge: `OutboundBridge` gains `onReasoningStart` / `onReasoningDelta` / `onReasoningEnd`, parallel to text. - Adapter: `pi-adapter.ts` routes `thinking_start` / `thinking_delta` / `thinking_end` from pi-ai. Parses a `**Title**\n\n<body>` heading once at write time (OpenAI Responses; no-op for others). - Timeline: live `reasoning: Collection<EntityTimelineReasoningItem>` on `EntityTimelineRunRow`, content built via delta-join. - UI: new `<ReasoningSection>` renders above items in `AgentResponseLive`. Streamdown body, click-to-expand on settle, redacted-block placeholder for opaque Anthropic payloads.

github-actions · 2026-06-04T13:36:48Z

Electric Agents Desktop Builds

Build artifacts for commit a2d56c3.

Platform	Status	Artifact
macOS Apple Silicon	Passed	DMG
macOS Intel	Passed	DMG
Windows x64	Passed	Installer
Linux x64	Passed	AppImage / deb

Workflow run

codecov · 2026-06-04T13:38:13Z

❌ 3 Tests Failed:

Tests completed	Failed	Passed	Skipped
1064	3	1061	41

View the top 2 failed test(s) by shortest run time

test/process-wake.test.ts > processWake > applies SIGINT that arrives before the handler run controller is created

Stack Traces | 0.123s run time

TypeError: Cannot read properties of undefined (reading 'toArray')
 ❯ Module.loadOutboundIdSeed src/outbound-bridge.ts:70:46
 ❯ runAgent src/context-factory.ts:554:33
 ❯ Object.handler test/process-wake.test.ts:1068:9
 ❯ Module.processWake src/process-wake.ts:2074:9
 ❯ test/process-wake.test.ts:1090:5

test/process-wake.test.ts > processWake > aborts an active run for server-handled SIGKILL without rewriting the signal

Stack Traces | 5s run time

Error: Test timed out in 5000ms.
If this is a long-running test, pass a timeout value as the last argument or configure it globally with "testTimeout".
 ❯ test/process-wake.test.ts:1184:3

View the full list of 1 ❄️ flaky test(s)

test/process-wake.test.ts > processWake > hydrates dynamic event source wake rows into the agent trigger message

Flake rate in main: 66.67% (Passed 2 times, Failed 4 times)

Stack Traces | 0.0203s run time

TypeError: Cannot read properties of undefined (reading 'toArray')
 ❯ Module.loadOutboundIdSeed src/outbound-bridge.ts:70:46
 ❯ Object.run src/context-factory.ts:608:17
 ❯ Object.handler test/process-wake.test.ts:1712:9
 ❯ Module.processWake src/process-wake.ts:2074:9
 ❯ test/process-wake.test.ts:1716:5

To view more test analytics, go to the Test Analytics Dashboard
_{📋 Got 3 mins? Take this short survey to help us improve Test Analytics.}

github-actions · 2026-06-04T13:40:19Z

Electric Agents Mobile Build

Local mobile checks ran for commit a2d56c3.

The EAS Android preview build was skipped because the mobile-eas-build label is not present.
Add the mobile-eas-build label to this PR to produce an installable preview build.

Workflow run

Previously `withProviderPayloadDefaults` short-circuited for any provider other than OpenAI / OpenAI-Codex, so picking Claude with a `reasoningEffort` higher than `auto` produced no effect — no `thinking` parameter was added to the request, so Anthropic ran in standard mode and the model emitted no `thinking_delta` events. The inbound reasoning plumbing landed in the same PR was correct but unreachable from Anthropic without this. Now: when the chosen model is Anthropic-capable for reasoning AND `reasoningEffort` is explicit (minimal/low/medium/high), inject thinking: { type: "enabled", budget_tokens: <by effort> } into the payload. Budgets follow Anthropic's docs (≥ 1024 floor): minimal=1024, low=2048, medium=8192, high=24576. `auto` stays opt-out of thinking so default sessions don't silently incur the extra reasoning tokens.

KyleAMathews

Lovely! Could you add a screenshot of the UI to the PR body?

KyleAMathews approved these changes Jun 4, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(agents-server-ui): stream model reasoning into the UI#4508

feat(agents-server-ui): stream model reasoning into the UI#4508
kevin-dp wants to merge 2 commits into
mainfrom
kevin/reasoning-content

kevin-dp commented Jun 4, 2026

Uh oh!

github-actions Bot commented Jun 4, 2026 •

edited

Loading

Uh oh!

codecov Bot commented Jun 4, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Jun 4, 2026 •

edited

Loading

Uh oh!

KyleAMathews left a comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

kevin-dp commented Jun 4, 2026

Summary

Implementation (end-to-end)

Reference

Test plan

Notes

Uh oh!

github-actions Bot commented Jun 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Electric Agents Desktop Builds

Uh oh!

codecov Bot commented Jun 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

❌ 3 Tests Failed:

Uh oh!

github-actions Bot commented Jun 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Electric Agents Mobile Build

Uh oh!

KyleAMathews left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

github-actions Bot commented Jun 4, 2026 •

edited

Loading

codecov Bot commented Jun 4, 2026 •

edited

Loading

github-actions Bot commented Jun 4, 2026 •

edited

Loading