Add generic OpenAI-compatible DeepSeek V4 routing - #106
Conversation
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![github-code-quality[bot]](https://avatars.githubusercontent.com/u/9919?s=60&v=4){kind=small}
![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: f5fa2102af
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
@codex address that feedback Repair only the live exact-head findings on f5fa210: (1) P1 first tool call falsely detected as repeat; preserve repeat detection and add regression coverage. (2) Honor registry model environment overrides on the generic openai-compatible dispatcher path. (3) Redact the configured credential before any provider-error truncation so partial opaque keys cannot leak. (4) Do not surface DeepSeek Pro through the unrestricted DeepSeek catalog/presets; retain the separate ws-gated deepseek-pro provider. (5) Remove the unused _call_responses import. Keep DeepSeek as configuration over the generic adapter and do not reintroduce a provider-specific adapter. Run the focused provider contracts and broader non-database gate, then request exact-head review. Do not merge. |
|
@codex review Full exact-head review of 9a3a8b7. Recheck all prior P1/P2 findings after the local repairs: first-tool repeat safety, registry model env override, credential redaction-before-truncation, resolved-model tier gating, and obsolete transport import removal. Preserve the generic openai-compatible architecture; do not reintroduce a DeepSeek-specific adapter. Both exact-head workflows are green. Do not merge. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 9a3a8b71db
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
@codex address that feedback Repair only the current exact-head P1: make |
|
To use Codex here, create an environment for this repo. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 5491892ae8
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
Exact-head repair update ( |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: d0d732994c
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
Exact-head follow-up ( |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 0c6978d602
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
Exact-head repairs are now at |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 2ab72d0e26
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
Final repair candidate is exact head |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 2a981aef65
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
@codex review Exact-head review requested for |
|
@codex review Please review exact head
Verification: all 11 repaired remote blobs exactly match the locally tested tree; 101 focused provider/package tests pass; the executable contract runner passes 26 pytest nodes, 97 contracts, and 75 checks; 27 registry/context/approval tests pass; compileall and whitespace checks pass; wheel build and source-isolated wheel imports pass. The executable review check includes the new dispatch-approval contract and the live registered |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 450706574b
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
@codex review Please review exact head
All 14 repaired remote files exactly match the locally tested bytes. Verification: 103 provider/package tests pass; the executable repair check passes 28 selected nodes and audits 101 contracts / 75 checks; 29 registry/context/approval tests pass; compileall, whitespace, annotation idempotence, and msdmd regeneration idempotence pass; source-isolated wheel imports pass. The modified service/provider/test modules remain within the 400-code-line doctrine; the preexisting oversized chat route is three code lines smaller than the preceding head. |
|
Codex Review: Didn't find any major issues. Chef's kiss. Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
1 participant
Summary
openai-compatibleadapterbase_url,api_key_env, model alias, Responses/Chat family, reasoning map, tool profile, and pricinga0/runtime and FastAPI provider dispatcher through the shared registry; explicitA0_PROVIDERchoices fail closedenv_keytoapi_key_env, fix primary-model attribution before preset lookup, and make optional DB memory lookup genuinely non-blockingDEEPSEEK_API_KEYsecret and add a provider-contract CI gatea0/adapters/openai_adapter.py; no DeepSeek-specific adapter or retired model identifier is introducedDeepSeek transport/model facts were checked against the current official docs:
File plan
python/services/providers/*,python/services/inference.pya0/provider_registry.py,a0/adapters/*,a0/router.pypython/config/providers.json,python/config/pricing.json.github/workflows/deploy.yml,cloudbuild.yaml,DEPLOYMENT.mda0p-deepseek-api-keytests/*openai_compatible*,python/tests/test_openai_compatible_contracts.pyREADME.md,CLAUDE.md,docs/ARCHITECTURE.md,suggest.mda0_msdmd.ts, ratio seals, package sources,.gitignoreVerification
77 passed— focused provider, standalone, package, and msdmd witness suite181 passed— all non-database tests (live-server and real-Postgres artifact tests excluded)npm run checknode --import tsx script/build.ts; thenpm run buildwrapper was blocked before project execution by the runner's local IPC policycheck_openai_compatible_registry_wiringpassed; the repository-wide contract runner still has pre-existing environment/fixture errors outside this changehmmm
skill-lib/ai-launcheris not present in this repository or canonicalskill-lib; after merge, its deployment environment still needs this branch/release plusDEEPSEEK_API_KEY.a0python/package was not given a third registry copy. The documented active Termux path is rootrun.sh -> a0/; decide separately whether to retire or independently packagea0python/.