chore: update model catalog from bot issues - #1070
Conversation
|
Codex (@codex) review |
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: bab6228748
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "Codex (@codex) review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "Codex (@codex) address that feedback".
| "perplexity/kimi-k3": ["perplexity"], | ||
| "perplexity/glm-5.2": ["perplexity"], |
There was a problem hiding this comment.
These new perplexity/... model IDs are Perplexity Gateway Chat Completions models, whose docs call https://api.perplexity.ai/router/v1/chat/completions (see https://docs.perplexity.ai/docs/gateway/quickstart). In this proxy, a perplexity endpoint still uses EndpointProviderToBaseURL.perplexity = "https://api.perplexity.ai" and fetchOpenAI appends the normal /chat/completions path, so requests for either newly added model will be sent to https://api.perplexity.ai/chat/completions instead of the gateway route and will fail or hit the wrong API. Add a gateway-specific provider/base URL override before advertising these models as native perplexity models.
Useful? React with 👍 / 👎.
| "databricks-qwen35-122b-a10b": ["databricks"], | ||
| "databricks-qwen3-next-80b-a3b-instruct": ["databricks"], | ||
| "thinkingmachines/inkling": ["baseten"], | ||
| "thinkingmachines/inkling": ["baseten", "openrouter", "together"], |
There was a problem hiding this comment.
When adding Together for thinkingmachines/inkling, this line also adds openrouter to a multi-provider direct mapping. The sync logic in providersForExactModelName() deliberately filters openrouter unless it is the only provider, and the corresponding test asserts this for native multi-provider models; otherwise getDirectModelEndpointTypes("thinkingmachines/inkling") can select the OpenRouter aggregator even though the model has native Baseten/Together routes. This entry should omit openrouter and only add the native provider that changed.
Useful? React with 👍 / 👎.
…5895) Two fixes on this bot-issues PR: 1. Remove perplexity/kimi-k3 and perplexity/glm-5.2. These are Perplexity *Gateway* (router) models — third-party Kimi/GLM served by Perplexity — which are only reachable at https://api.perplexity.ai/router/v1/chat/completions (per the Perplexity gateway quickstart). But the `perplexity` provider points at the standard https://api.perplexity.ai in BOTH the TS proxy (EndpointProviderToBaseURL.perplexity) and lingua (providers/openai.rs: url "https://api.perplexity.ai"), which only serves Sonar. So these ids cannot be routed to the gateway endpoint and are not invocable. (Confirms the codex P1/P2.) Could not invocation-validate either way: the CI Perplexity secret is currently quota-exhausted (401 "exceeded your current quota" on every perplexity model incl. the sonar control), so this is also blocked on env. Removed from model_list.json + index.ts and added to MANUAL_SYNC_EXCLUDED_MODELS so the daily bot does not re-add them until a perplexity-router base URL/provider exists. Same pattern as the mythos (bedrock-mantle) and Vertex-MaaS exclusions. 2. thinkingmachines/inkling index.ts mapping: the PR set it to ["baseten", "openrouter", "together"], but per BT-5895 openrouter must be excluded from a native model's direct AvailableEndpointTypes when other providers exist (providersForExactModelName filters it). Corrected to ["baseten", "together"] (model_list available_providers keeps openrouter). + tests for both exclusions. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
1 participant
Automated daily batch of model catalog updates from bot issues.
Included issues
Summary
thinkingmachines/inklingthinkingmachines/inkling2
perplexity/kimi-k3perplexity/glm-5.2perplexity/kimi-k3perplexity/glm-5.22
Verified metadata
#1068: [BOT ISSUE] Together: add together to thinkingmachines/inkling available_providers (regression)
Verification notes
No LLM verification step ran; model metadata was already complete in the issue.
sync_models vs proposed update
sync_models cross-check found differences. Official provider verification was used for the applied values, and sync_models discrepancies are listed below for review.
#1069: [BOT ISSUE] Perplexity: add missing perplexity/kimi-k3 and perplexity/glm-5.2 Gateway models
Verification notes
No LLM verification step ran; model metadata was already complete in the issue.
sync_models vs proposed update
sync_models cross-check found differences. Official provider verification was used for the applied values, and sync_models discrepancies are listed below for review.