fix(ai): route Copilot Grok 4.5 to responses - #906
Conversation
|
Hi, thanks for taking the time to contribute to Prime Agent! Since open sourcing the project, we’ve received far more pull requests than we can responsibly review and validate. Prime Agent runs directly on users’ machines, so we need to be deliberate about which changes we accept and how they are reviewed. Rather than leave a large backlog that we cannot meaningfully work through, we’re closing the current PR queue and moving to a discussion-first contribution process. We have established new contribution guidelines to help us continue iterating on Prime Agent and better manage contributions from the community. Going forward, we won’t review unsolicited pull requests. Instead, please start with a GitHub Discussion. We’ll identify recurring bugs and feature requests, create Issues for work we want to pursue, and invite pull requests from maintainers or vouched contributors when implementation is ready. Please read the full process documented in our contribution guidelines. While we’re closing this backlog, we’re still reviewing it at a high level to identify recurring bugs, useful ideas, and important problems that we should address ourselves. Thanks again for the time you put into this! |
….8 Max route Two regen blockers surfaced by fresh upstream data: - Copilot now serves gpt-6-astra; the classification coverage invariant fails regen on the unclassified family, as designed. Classified as openai-responses: native OpenAI serves its whole catalog (including gpt-6-astra) through the Responses API here, and every OpenAI family Copilot has shipped since gpt-5 is Responses-only (upstream pi-mono #906 probe class); no Copilot family has moved from responses back to chat completions. models.dev lists it live (1.05M context, reasoning, tool-capable) so the next regen ships it on 8 provider rows. - OpenRouter renamed qwen/qwen3.8-max to qwen/qwen3.8-max-0902 while Prime Inference still serves the undated id, so fresh regens dropped the row's OpenRouter-derived thinkingLevelMap and failed the mandatory-efforts pin in prime-inference-models.test.ts (pre-existing on main, A/B-verified by the sandbox gate). Added the alias PRIME_INFERENCE_OPENROUTER_ALIASES exists for; the regenerated row keeps its exact previous map. Catalog regenerated: 42 rows differ (8 gpt-6-astra additions, the qwen rename pair, and fetch-time upstream drift: price updates including Copilot gpt-5.6-sol doubling, new ling-sante/muse rows, granite-4.1 removal). Regen byte-stable across a second run; full ai suite green against the fresh catalog. Linear: RES-1273
Summary
grok-4.5through the OpenAI Responses API.Testing
npx tsx ../../node_modules/vitest/dist/cli.js --run test/github-copilot-models.test.tsnpm run checkFixes #875
Note
Route GitHub Copilot
grok-4.5to the Responses API instead of Chat CompletionsThe
grok-4.5model under thegithub-copilotprovider does not support the Chat Completions endpoint, so it is now assignedopenai-responsesin generate-models.ts alongside the existinggpt-5andoswemodels. The generated catalog and a new test in github-copilot-models.test.ts confirm the correct routing.Macroscope summarized 76b84a2.