feat(chat): add ACP providers to model picker - #1202
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub. |
|
Important Review skippedToo many files! This PR contains 174 files, which is 74 over the limit of 100. To get a review, reduce the PR to 100 files or fewer by splitting it into smaller PRs or changing its base branch. Upgrade to a paid plan to raise the limit. This review couldn't start because sufficient usage credits or metered capacity aren't available. Add credits or update usage-based reviews in the billing tab, then retry. ⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Team Run ID: ⛔ Files ignored due to path filters (7)
📒 Files selected for processing (174)
You can disable this status message by setting the 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. Comment |
Bugbot couldn't run - usage limit reachedBugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit. A user or team admin can review and increase usage limits in the Cursor dashboard. (requestId: serverGenReqId_c915a506-1b8d-4fd8-8dc6-d551e599c1cb) |
15f0604 to
eb2ef1d
Compare
Problem
ADE could not expose or reliably route Qwen, Kimi, Grok, and GitHub Copilot across provider status, model selection, and chat surfaces.
Cause
Provider detection, ACP runtime lifecycle, and model catalogs used separate auth and fallback assumptions. Existing pooled runtimes could also retain a prior permission mode.
Change and boundary
Verification
Kimi remains correctly sign-in-required where no authenticated local CLI is available. No API keys or vendor config files were modified.
Model and harness: GPT-5.6 / ADE Codex app-server harness.
Note
Medium Risk
Touches auth probing, Grok permission supervision, and provider enablement across chat creation and tracked terminals; mistakes could misreport sign-in state or weaken Grok approval guarantees, but changes are heavily tested and fail-safe on probe failures.
Overview
Adds four Agent Client Protocol (ACP) providers — Qwen, Kimi, Grok, and GitHub Copilot — across the CLI, RPC, mobile sync, TUI, and desktop AI stack so users can start chats, tracked CLI sessions, and model picks with the same flows as existing agents.
CLI & sync:
--provider/start_cli_sessionenums and validation now include the new names;agentRegistryregisters install/auth hints and error patterns; remote model-catalog refresh and chat-create paths treat them like OpenCode/Pi (activateRuntimefor auth-gated model lists). TUI pieces (model picker rails, themes, terminal resume, slash palette) align with sharedMODEL_PICKER_PROVIDER_ORDER.Desktop: New ACP executable resolution, disk-based credential hints in
authDetector, andprobeAcpProviderAuth(protocolsession/new, cached, forced on settings refresh). Grok gets a dedicated permission preflight that attests Claude-settings neutralization via debug logs (notgrok inspect). Qwen usersettings.jsoncan override curated model rows. Settings/status wiring exposes optional availability flags, provider connections, disabled-provider filtering, and diagnostics (doctorwhere supported).Tests: Coverage for ACP auth, executables, Grok preflight, Qwen settings; CLI serve tests set
ADE_DISABLE_RUNTIME_SERVICE_INSTALL/ADE_DISABLE_TOOLS_FETCHfor stability.Reviewed by Cursor Bugbot for commit 15f0604. Configure here.