Skip to content
Source
Sign inSign up
TanStack/schemas

Read the GPT-6.2 'rejects none' sentence as mandatory effort reasoning #712

Mergedtombeckenhammerged 1 commit intomain fromfix/711-openai-astra-sol-reasoningmerged by tombeckenham View on GitHub

tombeckenhamcommented

Problem

gpt-6.2-astra and gpt-6.2-sol have capabilities.reasoning: true but reasoning: null and no factSources.reasoning entry.

Root cause

parseReasoningEffort in src/server/providers/openai-model-docs.ts only recognises an effort list ("reasoning.effort supports ...") or "supports configurable reasoning effort". The GPT-6.2 pages state neither. They say only "The API rejects requests that set reasoning effort to none.", so the parser returned null and the row got no reasoning object.

Change

  • parseReasoningEffort: when the page lists no efforts but states the API rejects none effort, return { mode: 'effort', mandatory: true }. The page names the effort control and states its off value is refused, which is the mandatory: true rule on ModelReasoning. No efforts list is invented, since the page publishes none. This runs for every OpenAI model page, not a per-id case.
  • DOCS_CACHE_VERSION v42 to v43, so cached v42 page parses (reasoning null) are not served for six hours.
  • Test in native-reasoning.test.ts using both live pages added to fixtures/native-reasoning-docs.json.

factSources.reasoning follows from the existing tagDocsFacts: docs-derived, pointing at each model page.

Sources

Fixes #711

Summary by CodeRabbit

  • New Features
    • Added provider information for Kimi K3, GPT-6.2 Astra, and GPT-6.2 Sol, including supported capabilities and availability details.
  • Bug Fixes
    • GPT-6.2 reasoning settings are now identified as mandatory rather than appearing optional or unknown.
    1. tombeckenham linked this to fix #711

    2. ⚠️ No Changeset found

      Latest commit: 8e96babec130c2b74dab659765e73a2b88323ac4

      Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

      This PR includes no changesets

      When changesets are added to this PR, you'll see the packages that this PR includes changesets for and the associated semver types

      Click here to learn what changesets are, and how to add one.

      Click here if you're a maintainer who wants to add a changeset to this PR

    3. Explain Change →

      No actionable comments were generated in the recent review.

      Walkthrough

      The parser now identifies mandatory reasoning effort for OpenAI models when their documentation says requests that set effort to none are rejected. The fixtures and test cover GPT-6.2 Astra and Sol, and the docs cache version changes to v43.

      Changes: Reasoning effort inference

      Layer / File(s)Summary
      Infer and validate mandatory effort
      src/server/providers/fixtures/native-reasoning-docs.json, src/server/providers/openai-model-docs.ts, src/server/providers/native-reasoning.test.ts, src/server/providers/model-facts.ts
      The fixtures add documentation for Kimi K3 and GPT-6.2 Astra and Sol. The parser returns mandatory effort mode when documentation says effort set to none is rejected. A test checks the GPT-6.2 results, and the docs cache version changes to v43.

      Estimated code review effort: 2 (Simple) | ~10 minutes

      Change: Bug fix · Severity of issue fixed: Medium

      Merge Risk: Minimal · up to 8e96b

      The change makes GPT-6.2 Astra and Sol report mandatory reasoning effort from their documentation. The cache version bump refreshes previously cached parses. The risk of merging is low, and no blocking issue was found.

      Pre-merge checks: Passed 3 · Inconclusive 1

      Failed Failed checks: 1 inconclusive
      Check nameStatusExplanationResolution
      Out of Scope Changes checkInconclusiveThe parser change, the cache version bump, and the test change in native-reasoning.test.ts all support #711. The fixture file src/server/providers/fixtures/native-reasoning-docs.json shows a modified …Provide the fixture diff hunks for docs.together.ai/docs/kimi-k3-quickstart.md, or confirm that the Kimi K3 change is required by a test in this PR. If it is not required, remove it from this PR.
      Passed Passed checks: 3 passed

      Check nameStatusExplanation
      Description CheckPassedCheck skipped - CodeRabbit’s high-level summary is enabled.
      Title checkPassedThe title names the main change: parseReasoningEffort in openai-model-docs.ts now reads the GPT-6.2 'rejects none' prose as mandatory effort reasoning. The raw summary confirms this parser change, the…
      Linked Issues checkPassedIssue #711 asks that gpt-6.2-astra and gpt-6.2-sol get a reasoning object or a factSources.reasoning entry. The reviewed diff adds a branch in parseReasoningEffort (src/server/providers/openai-model-d…
      Full details: Title check

      Explanation

      The title names the main change: parseReasoningEffort in openai-model-docs.ts now reads the GPT-6.2 'rejects none' prose as mandatory effort reasoning. The raw summary confirms this parser change, the v42 to v43 cache bump, and the new test for GPT-6.2 Astra and Sol. The title is specific and matches the primary intent in the PR objectives.

      Full details: Linked Issues check

      Explanation

      Issue #711 asks that gpt-6.2-astra and gpt-6.2-sol get a reasoning object or a factSources.reasoning entry. The reviewed diff adds a branch in parseReasoningEffort (src/server/providers/openai-model-docs.ts). It runs only when no effort list is parsed. It returns { mode: 'effort', mandatory: true } when the page text matches 'rejects requests that set reasoning effort to none'. The GPT-6.2 pages state this sentence, per the PR description. The test in native-reasoning.test.ts asserts this result for both model page URLs. The DOCS_CACHE_VERSION bump (v42 to v43) prevents stale parses from being served. Limits: I did not read the fixture hunks for the two GPT-6.2 page entries, so the test input was not verified. I did not verify the factSources.reasoning output of tagDocsFacts, or the live endpoint response. The code path that produces the reasoning object is established.

      Full details: Out of Scope Changes check

      Explanation

      The parser change, the cache version bump, and the test change in native-reasoning.test.ts all support #711. The fixture file src/server/providers/fixtures/native-reasoning-docs.json shows a modified line for the Together Kimi K3 quickstart page (docs.together.ai/docs/kimi-k3-quickstart.md). That entry does not relate to #711. The diff output was truncated at 16000 bytes, so I could not read the full hunk. The AI-generated summary describes Kimi K3 content in the fixture, but that summary is not source evidence. I cannot confirm whether the Kimi K3 change is unrelated or whether it is test data needed by the new or existing tests.

      Resolution

      Provide the fixture diff hunks for docs.together.ai/docs/kimi-k3-quickstart.md, or confirm that the Kimi K3 change is required by a test in this PR. If it is not required, remove it from this PR.

      Finishing Touches

      Generate unit tests (beta)

      • Commit to this branch
      • Create a new PR
      • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

      Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

      Share

      Recent review info

      Run configuration

      • Source mode: Auto
      • Repository: Unavailable
      • Branch: Unavailable
      • File: No configuration file from this repository applied
      • Commit: Unavailable
      • Layers: defaults
      • Review profile: CHILL
      • Plan: Advanced
      • Run ID: 9f730bc1-44de-4eb4-b57e-eddb442832f4
      Commits

      Reviewing files that changed from the base of the PR and between fa86c5eda5d03ed71293af2524003ff9907a4fc4 and 8e96babec130c2b74dab659765e73a2b88323ac4.

      Files selected for processing: 4

      • src/server/providers/fixtures/native-reasoning-docs.json
      • src/server/providers/model-facts.ts
      • src/server/providers/native-reasoning.test.ts
      • src/server/providers/openai-model-docs.ts

      Included review availability: This review used your included allowance. Your plan provides up to 1 included review per hour; 0 remain after this review.

    4. tombeckenham merged this3cd28b6

    5. tombeckenham closed this

Sign in to join this conversation.