chore: update model catalog from bot issues - #1081
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
Codex (@codex) review |
![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: 440a746668
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".
| "gpt-5.4-nano": ["openai", "azure", "openrouter"], | ||
| "Qwen/Qwen3.7-Plus": ["together"], | ||
| "qwen/qwen3.6-27b": ["groq"], | ||
| "qwen/qwen3.6-27b": ["groq", "openrouter"], |
There was a problem hiding this comment.
This widens the direct endpoint mapping for a model that still has Groq as its native provider. The catalog sync code explicitly keeps OpenRouter aggregator unions out of direct index.ts endpoint types unless OpenRouter is the only provider (packages/proxy/scripts/sync_models.ts:1360-1366); otherwise getDirectModelEndpointTypes("qwen/qwen3.6-27b") starts offering OpenRouter routing/credential fallback instead of leaving this native model direct-routed to Groq. Keep openrouter in available_providers if desired, but don't add it to AvailableEndpointTypes here.
Useful? React with 👍 / 👎.
| "databricks" | ||
| ] | ||
| }, | ||
| "databricks-kimi-k3": { |
There was a problem hiding this comment.
The batch text says #1075 adds both databricks-kimi-k3 and databricks-gpt-5-4-mini, but only Kimi is introduced here; I checked the repo with rg 'databricks-gpt-5-4-mini' and there is still no model_list.json or AvailableEndpointTypes entry. In the scenario where this release is expected to expose the new Databricks GPT-5.4 Mini model, requests or selection for that ID will continue to fail as unknown even though the issue is closed.
Useful? React with 👍 / 👎.
…way Kimi Three fixes on this bot-issues PR: 1. codestral-2508 max_input_tokens 128000 -> 256000. The PR halved it, but both LiteLLM (mistral/codestral-2508 max_input_tokens 256000) and OpenRouter (mistralai/codestral-2508 context 256000) put it at 256000. Reverted the regression. 2. thinkingmachines... no: qwen/qwen3.6-27b index.ts mapping ["groq","openrouter"] -> ["groq"]. BT-5895: openrouter must be excluded from a native model's direct AvailableEndpointTypes when other providers exist. model_list available_providers keeps openrouter. (Confirms the codex P2.) 3. Remove perplexity/kimi-k2.7-code (Perplexity Gateway/router model, same class as the previously-excluded perplexity/kimi-k3 and perplexity/glm-5.2): only reachable at api.perplexity.ai/router/v1, but the perplexity provider points at the standard api.perplexity.ai (Sonar only) in both the TS proxy and lingua, so it is unroutable. Added to MANUAL_SYNC_EXCLUDED_MODELS. Could not invocation-validate anyway: the CI Perplexity secret is quota-exhausted (401 on all perplexity ids incl. the sonar control). + test. Left as-is (validated / consistent): databricks-kimi-k3 (could NOT validate -- the CI Databricks secret is invalid, 403 "Invalid access token" on ALL databricks ids incl. the gpt-5-5-pro/claude-opus-4-8 controls; env blocker, not a model defect); meta.llama4-scout/maverick bedrock activation (invoke 200 via the us. inference-profile id; the bare meta.llama4-*-v1:0 ids 400 "on-demand throughput isn't supported", identical to the existing meta.llama3-* bare entries -- a pre-existing catalog pattern, not a regression). 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
perplexity/kimi-k2.7-codeperplexity/kimi-k2.7-code2
qwen/qwen3.6-27bqwen/qwen3.6-27b2
codestral-2508codestral-25082
databricks-kimi-k3databricks-kimi-k32
anthropic.claude-3-5-sonnet-20241022-v2:0anthropic.claude-3-5-sonnet-20240620-v1:0anthropic.claude-3-5-sonnet-20241022-v2:0anthropic.claude-3-5-sonnet-20240620-v1:0google.gemma-4-31bgoogle.gemma-4-26b-a4bgoogle.gemma-4-e2bgoogle.gemma-4-31bgoogle.gemma-4-26b-a4bgoogle.gemma-4-e2b2
publishers/google/models/gemini-2.5-flashpublishers/google/models/gemini-2.5-propublishers/google/models/gemini-2.5-flashpublishers/google/models/gemini-2.5-pro2
meta.llama4-scout-17b-instruct-v1:0meta.llama4-maverick-17b-instruct-v1:0meta.llama4-scout-17b-instruct-v1:0meta.llama4-maverick-17b-instruct-v1:02
Verified metadata
#1072: [BOT ISSUE] Perplexity: add missing perplexity/kimi-k2.7-code
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.
#1073: [BOT ISSUE] Groq: fix qwen/qwen3.6-27b max_output_tokens (32768 → 16384)
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.
#1074: [BOT ISSUE] Mistral: fix codestral-2508 max_input_tokens regression (256k → 128k)
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.
#1075: [BOT ISSUE] Databricks: add missing databricks-kimi-k3 and databricks-gpt-5-4-mini
Verification notes
Verification
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.
#1076: [BOT ISSUE] Bedrock: fix Claude 3.5 Sonnet pricing (Extended Access $6/$30 since Dec 2025)
Verification notes
Verification
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.
#1077: [BOT ISSUE] Bedrock: add missing pricing for Gemma 4 models (31b, 26b-a4b, e2b)
Verification notes
Verification
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.
#1079: [BOT ISSUE] Vertex: add missing cache read pricing for Gemini 2.5 Flash and Pro
Verification notes
Verification
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.
#1080: [BOT ISSUE] Bedrock: add missing available_providers to Llama 4 Scout and Maverick base entries
Verification notes
Verification
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.