chore: update model catalog from bot issues - #822
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: b0040c81da
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".
| "openai.gpt-5.5": ["bedrock"], | ||
| "openai.gpt-5.4": ["bedrock"], | ||
| "Qwen/Qwen3.7-Plus": ["together"], | ||
| "xai.grok-4.3": ["bedrock"], |
There was a problem hiding this comment.
These new Bedrock entries are backed by model_list.json specs with format: "openai", but this proxy dispatches by modelSpec.format; the OpenAI path immediately rejects Bedrock secrets with Bedrock does not support OpenAI format in fetchOpenAI. As a result, requests for the newly added Bedrock-only models such as openai.gpt-5.5, openai.gpt-5.4, and xai.grok-4.3 will fail locally before reaching AWS unless they use a Bedrock-supported format/path, such as converse, or the proxy adds an OpenAI-compatible Bedrock handler.
Useful? React with 👍 / 👎.
b0040c8 to
b678b11
Compare
Codex Review ResponseIssue: Bedrock-only models using unsupported
|
The zai-org/GLM-* ids are shared across baseten and together, but the audit attributes a newly-found shared id to a single provider (it filed GLM-5.2 as together-only in #819) and the per-provider dedup drops the other provider. Result: zai-org/GLM-5.1 was together-only despite baseten serving it, and zai-org/GLM-5.2 was missing entirely. Verified by direct provider calls (real 200 completions): - zai-org/GLM-5.2: works on BOTH baseten and together - zai-org/GLM-5.1: works on baseten (was missing from its providers) Fixes: - zai-org/GLM-5.1 available_providers -> ["baseten","together"] (model_list + index.ts; the normalize backstop only adds missing mappings, it does not widen an existing one, so index.ts was updated by hand). - Add zai-org/GLM-5.2 ["baseten","together"] (pricing 1.4/4.4 corroborated by together #819 + the fireworks glm-5p2 entry; 262k context per #819). - Also added the missing index.ts mapping for voxtral-small-2507 (another model this PR added without one) via normalize-local-models. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
xai.grok-4.3, openai.gpt-5.5, openai.gpt-5.4 are served on Amazon Bedrock's "Mantle" engine via the OpenAI Chat Completions/Responses API on the bedrock-mantle endpoint — NOT Converse (the AWS Grok 4.3 model card lists Converse as unsupported). The gateway supports the /chat/completions path on Bedrock, so set format "converse" -> "openai" for all three. Verified: the gateway now routes xai.grok-4.3 to Bedrock (AWS_DEFAULT_CREDENTIALS, format=openai); a clean 200 isn't reachable from the CI account because grok-4.3 is us-west-2-only and the CI Bedrock creds are us-east-1 (env, not a catalog defect). Also fix grok-4.3 max_output_tokens 1000000 -> 131072 (1M is the context window; the model card's max_completion_tokens default is 131072). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…rmat to converse
Both model-sync workflows have a 'Respond to Codex review with Claude Code'
step that auto-applies Codex's suggestions to the catalog. When the index.ts
Bedrock mappings for openai.gpt-5.5 / openai.gpt-5.4 / xai.grok-4.3 are present,
Codex posts a P1 ('openai format on a Bedrock model fails') because the TS
proxy's fetchOpenAI rejects Bedrock secrets, and the auto-apply step 'fixes' it
by flipping format openai->converse. That is wrong: these are Bedrock Mantle
models served only via the OpenAI-compatible bedrock-mantle endpoint, which does
not support the Converse API or InvokeModel, so converse breaks invocation at
AWS. This recurred across #834/#836/#840/#843 (all reverting the validated #822
openai fix). The LiteLLM field sync never touches format (no checkAndUpdateFormat),
so this guardrail on the codex-response prompt is the correct lever.
Adds a 'Bedrock Mantle exception' instruction to the codex-response prompt in
both sync-models.yaml and fix-missing-model-bot-issues.yaml: keep format=openai
for bedrock-only openai-format models and explain in the summary that this is an
intentional Mantle case to be handled proxy-side, not a catalog change.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
1 participant
Automated daily batch of model catalog updates from bot issues.
Included issues
xai.grok-4.3) #805: [BOT ISSUE] Bedrock: add missing xAI Grok 4.3 model (xai.grok-4.3)Qwen/Qwen3.7-Plus) #806: [BOT ISSUE] Together: add missing Qwen3.7-Plus model (Qwen/Qwen3.7-Plus)llama3.3-70b,llama3.1-8b,llama-4-scout-17b-16e-instruct) #807: [BOT ISSUE] Cerebras: remove deprecated models (llama3.3-70b,llama3.1-8b,llama-4-scout-17b-16e-instruct)moonshotai/kimi-k2-instruct-0905,meta-llama/llama-4-maverick-17b-128e-instruct,gemma2-9b-it) #808: [BOT ISSUE] Groq: remove deprecated models (moonshotai/kimi-k2-instruct-0905,meta-llama/llama-4-maverick-17b-128e-instruct,gemma2-9b-it)zai-org/GLM-5.2model #819: [BOT ISSUE] Together: add missingzai-org/GLM-5.2modelvoxtral-small-2507model #820: [BOT ISSUE] Mistral: add missingvoxtral-small-2507modelSummary
xai.grok-4.3xai.grok-4.32
3
4
Qwen/Qwen3.7-PlusQwen/Qwen3.7-Plus2
3
llama3.3-70bllama3.1-8bllama-4-scout-17b-16e-instructllama3.3-70bllama3.1-8bllama-4-scout-17b-16e-instruct2
moonshotai/kimi-k2-instruct-0905meta-llama/llama-4-maverick-17b-128e-instructgemma2-9b-itmoonshotai/kimi-k2-instruct-0905meta-llama/llama-4-maverick-17b-128e-instructgemma2-9b-it2
openai.gpt-5.5openai.gpt-5.4openai.gpt-5.5openai.gpt-5.42
3
4
claude-mythos-5claude-mythos-52
3
4
deepseek-ai/DeepSeek-V4-Promoonshotai/Kimi-K2.6deepseek-ai/DeepSeek-V4-Promoonshotai/Kimi-K2.62
3
mistral-large-2411pixtral-large-2411open-mistral-nemo-2407mistral-large-2411pixtral-large-2411open-mistral-nemo-2407zai-org/GLM-5.2zai-org/GLM-5.22
voxtral-small-2507voxtral-small-25072
Verified metadata
#805: [BOT ISSUE] Bedrock: add missing xAI Grok 4.3 model (
xai.grok-4.3)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.
#806: [BOT ISSUE] Together: add missing Qwen3.7-Plus model (
Qwen/Qwen3.7-Plus)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.
#807: [BOT ISSUE] Cerebras: remove deprecated models (
llama3.3-70b,llama3.1-8b,llama-4-scout-17b-16e-instruct)Verification notes
Verification Checklist
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.
#808: [BOT ISSUE] Groq: remove deprecated models (
moonshotai/kimi-k2-instruct-0905,meta-llama/llama-4-maverick-17b-128e-instruct,gemma2-9b-it)Verification notes
Verification Checklist
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.
#810: [BOT ISSUE] Missing AWS Bedrock models: openai.gpt-5.5, openai.gpt-5.4
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.
#814: [BOT ISSUE] Anthropic: add missing claude-mythos-5 model
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.
#815: [BOT ISSUE] Together: stale pricing for deepseek-ai/DeepSeek-V4-Pro and missing pricing for moonshotai/Kimi-K2.6
Verification notes
Verification Checklist
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.
#816: [BOT ISSUE] Mistral: mark deprecated mistral-large-2411, pixtral-large-2411, open-mistral-nemo-2407
Verification notes
Verification Checklist
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.
#819: [BOT ISSUE] Together: add missing
zai-org/GLM-5.2modelVerification 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.
#820: [BOT ISSUE] Mistral: add missing
voxtral-small-2507modelVerification 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.