fix(grok): skip set_model when effort is unchanged or a prompt is in flight

[maria-rcks]

What Changed

applyGrokAcpModelSelection no longer treats a present reasoningEffort selection as enough reason to call session/set_model. It compares against the effort last applied on the session and only sends the RPC when the model id or effort actually changed.

sendTurn also skips the apply entirely on a steer (promptsInFlight > 0). Prep runs under the thread lock while prompt runs outside it, so a steer must not issue set_model while a prior prompt is still in flight.

Why

The composer always restates reasoningEffort once Grok capabilities are discovered. Before the reasoning-effort change, same-model turns skipped set_model entirely, so steers never hit that path. After it, every turn — including steers — called set_model.

UI Changes

None. This is a server-only apply/skip change.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes
• I included a video for animation/interaction changes

Made by Cursor Grok 4.6 via the Cursor Cloud harness while working on T3 Code.

  

Note

Skip session/set_model in GrokAdapter when effort is unchanged or a prompt is in flight

• Tracks the last applied reasoning effort via ctx.currentReasoningEffort in GrokSessionContext, set during startSession and updated after each model selection.
• sendTurn now skips session/set_model entirely on steer turns (when a prior prompt is still in flight) and only calls it when the model id or reasoningEffort actually changes.
• applyGrokAcpModelSelection accepts a currentReasoningEffort parameter and treats a present-but-unchanged effort as a no-op.
• Risk: changing the model id without an explicit reasoningEffort selection now clears ctx.currentReasoningEffort (see GrokAdapter.ts); a subsequent turn that restates the cleared effort will issue set_model again.

📊 Macroscope summarized 7db5f22. 2 files reviewed, 1 issue evaluated, 0 issues filtered, 1 comment posted

🗂️ Filtered Issues

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: bf688214-fe06-4923-8b67-36785fe4d8ba

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

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.}

[macroscopeapp]

🟡 Medium Layers/GrokAdapter.ts:788

A requested reasoning effort is cached as applied even when the initial selection has no boundModelId, so a later selection with that model and the same effort omits _meta.reasoningEffort and never applies it. Initialize currentReasoningEffort only when the start model selection was actually applied.

-            currentReasoningEffort: requestedStartReasoningEffort,
+            currentReasoningEffort: boundModelId ? requestedStartReasoningEffort : undefined,

🚀 Reply "fix it for me" or copy this AI Prompt for your agent:

In file @apps/server/src/provider/Layers/GrokAdapter.ts around line 788:

A requested reasoning effort is cached as applied even when the initial selection has no `boundModelId`, so a later selection with that model and the same effort omits `_meta.reasoningEffort` and never applies it. Initialize `currentReasoningEffort` only when the start model selection was actually applied.

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

ℹ️ No successful main baseline artifact is available yet. This run establishes the initial measurement.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	—	13.4 KiB	—	15.1 KiB	✅
Codex	Thread snapshot wire	—	6.9 KiB	—	7.3 KiB	✅
Codex	Live turn WebSocket wire	—	6.5 KiB	—	7.8 KiB	✅
Codex	Live turn WebSocket decoded	—	55.0 KiB	—	66.4 KiB	✅
Codex	Live turn messages	—	16	—	21	✅
Claude	Total thread wire	—	13.5 KiB	—	15.1 KiB	✅
Claude	Thread snapshot wire	—	6.9 KiB	—	7.3 KiB	✅
Claude	Live turn WebSocket wire	—	6.6 KiB	—	7.8 KiB	✅
Claude	Live turn WebSocket decoded	—	55.8 KiB	—	66.4 KiB	✅
Claude	Live turn messages	—	16	—	21	✅

Baseline: unavailable · PR result: 7db5f22 · Source CI: failure

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.