fix(server): Grok status check uses grok models, not a full ACP session

[troybrave]

Fixes #7746.

What changed

Grok provider health no longer starts grok agent stdio + session/new.

It now:

1. Runs grok --version (unchanged).
2. Runs grok models to get available models and login state.
3. Marks the provider ready when the CLI is installed and logged in.
4. Marks warning (not error) if models listing fails/times out, or if Grok is installed but not logged in.

Why

The old probe created a full ACP session under a 15s deadline. session/new boots every configured MCP server for the workspace cwd. On a real workspace that misses 15s, T3 writes status: "error" to ~/.t3/caches/grok.json, and the UI says Grok is broken even while Grok chats already work.

grok models lists models and login without starting MCP servers. Direct ACP handshake on this machine is ~2s; the timeout was the MCP-backed session, not a missing CLI.

Tests

pnpm exec vp test run src/provider/Layers/GrokProvider.test.ts — 10/10 passed.

Related

• Complements fix(server): a provider probe timeout no longer marks the provider broken #7232 (timeout should not persist error). This PR removes the ACP session from the probe entirely.
• Also sets auth.status from grok models (#7505).

---

Note

Medium Risk
Changes Grok provider status, auth, and model discovery. A brittle CLI-output parser could mis-report login or models, but chats still work when listing fails.

Overview
Stops the Grok health probe from starting grok agent stdio / session/new (which booted workspace MCP servers and often timed out as error). After grok --version, it now runs grok models and parses login plus model ids.

Installed + logged-in CLIs are ready. Missing login, or a failed/timed-out grok models, is a warning with fallback models so chats are not treated as broken. Auth is set from that output (authenticated / unauthenticated / unknown).

^{Reviewed by Cursor Bugbot for commit 60f089c. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Replace ACP session with grok models CLI for Grok provider status check

• checkGrokProviderStatus now probes models and auth by running grok models instead of starting a full ACP session, parsing the output via the new parseGrokModelsCliOutput utility
• Adds grokAuthFromModelsCli to map login state into ServerProviderAuth, and displayNameFromGrokModelSlug to derive human-friendly model names from slugs
• Generalizes the version CLI runner into runGrokCliCommand supporting arbitrary subcommands
• Behavioral Change: grok models failures and timeouts now return status warning (previously error); when logged out, status is warning with guidance to run grok login; empty model listings fall back to built-in models with status ready

^{Macroscope summarized 60f089c.}

[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: d98ecc7b-f860-4eba-b73c-e68fd2ac4880

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]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This PR fundamentally changes Grok provider status checks from ACP session-based to CLI-based discovery, and changes multiple error states to warnings. These runtime behavior changes and the new auth detection logic warrant human review.

^{You can add or adjust custom eligibility rules. Learn more.}