fix(models): correct xAI reasoning_effort support

[steebchen]

Supersedes #3438 (fixes #3436, #3403).

Problem

Six xAI mappings declare reasoning: true but omit reasoning_effort from supportedParameters, so a caller-supplied effort is dropped at the gate in packages/actions/src/prepare-request-body.ts:4077-4086. #3438 treated all six as the same oversight and declared the parameter on every one.

E2E against api.x.ai shows that premise does not hold. Three of those deployments reason unconditionally but reject the effort knob outright:

externalId	reasoning_effort	Result
grok-4	low	200 (aliases to grok-4.3 upstream)
grok-4.3	low	200
grok-4.20-beta-0309-reasoning	low / medium / high	400 does not support parameter reasoningEffort
grok-build-0.1	low / medium / high	400 same
grok-4.20-reasoning	low	400 same

Control requests with the parameter omitted return 200 for all three, so the parameter is the sole cause; streaming behaves identically. For those mappings the omission was the catalogue correctly reflecting upstream, not the bug the issues describe — declaring it turns every effort-bearing request into an upstream 400.

Change

packages/models/src/models/xai.ts — keep #3438's addition for xai/grok-4, azure-ai-foundry/grok-4-3 and azure-ai-foundry/grok-4-1-fast-reasoning; restore the shared xaiSupportedParamsNoFreqPresence list for the three rejecting mappings, each with a comment recording why the omission is load-bearing so it does not get "fixed" again:

• xai / grok-4-20-beta-0309-reasoning
• xai / grok-build-0-1
• vertex-openai / grok-4-20-reasoning — same upstream model as the first

packages/actions/src/prepare-request-body.spec.ts — split #3438's six-row table into a positive table (three mappings forward the effort) and a negative one asserting the rejecting mappings are reasoning: true, do not list reasoning_effort, and drop the value from the request body. Also narrows the mapping lookup to ProviderModelMapping; reading .supportedParameters off the raw catalogue union fails tsc with TS2339, which broke pnpm build on #3438.

Verification

Ran with a live xAI key.

• TEST_MODELS="xai/grok-4,xai/grok-4-20-beta-0309-reasoning,xai/grok-build-0-1" pnpm test:e2e — 29 passed, 2 skipped, 0 failed. On fix(models): reasoning_effort for azure grok-4.3 #3438 the same command produced 4 failures (chat-reasoning.e2e.ts > basic reasoning and chat-rs.e2e.ts > reasoning + streaming, both models, 400 vs expected 200). Those four pass on main, confirming the regression rather than a pre-existing failure.
• prepare-request-body.spec.ts + models specs: 380 passed.
• pnpm build, pnpm format, pnpm lint: clean.

Not verified

azure-ai-foundry/grok-4-1-fast-reasoning keeps the parameter but was not exercised — no Azure credentials here, and the id does not resolve on native xAI (grok-4.1-fast-reasoning → "Model not found"; our xai mapping for it is deactivated). Since every xAI reasoning model probed except grok-4.3 rejects reasoning_effort, this one is worth a /e2e run with the Azure key before merging. azure-ai-foundry/grok-4-3 is safer — native grok-4.3 accepts the parameter.

🤖 Generated with Claude Code

https://claude.ai/code/session_01RbUnEaa7vrZZkD82wSStTr

---

Generated by Claude Code

Summary by CodeRabbit

• Bug Fixes
  • Improved handling of the reasoning_effort setting across supported xAI and Azure AI Foundry models.
  • Ensured the setting is forwarded only when supported and omitted for deployments that reject it.
  • Updated Azure AI Foundry Grok 4.3 to use a default reasoning effort of medium.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 2271edaf-8f3a-4964-a10a-eb813202de69

📥 Commits

Reviewing files that changed from the base of the PR and between ed39fc9 and fc91bc9.

📒 Files selected for processing (2)

• packages/actions/src/prepare-request-body.spec.ts
• packages/models/src/models/xai.ts

---

Walkthrough

The change updates selected xAI and Azure AI Foundry model mappings to declare reasoning_effort support. Regression tests verify forwarding for supported deployments and omission for deployments whose APIs reject the parameter.

Changes

xAI reasoning effort support

Layer / File(s)	Summary
Model capability declarations packages/models/src/models/xai.ts	Grok 4, Azure Grok 4.1 Fast Reasoning, and Azure Grok 4.3 now declare reasoning_effort. Grok 4.20 reasoning variants and Grok Build 0.1 document that the parameter remains unsupported.
Request-body regression coverage packages/actions/src/prepare-request-body.spec.ts	Parameterized tests verify forwarding and omission behavior. The Azure AI Foundry Grok 4.3 expectation now requires "medium" reasoning effort.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Possibly related issues

• theopenco/llmgateway#3403 — Covers the same reasoning_effort mapping gap for the Vertex OpenAI Grok 4.20 reasoning deployment.

Possibly related PRs

• theopenco/llmgateway#3159 — Added reasoning_effort support to an xAI Grok mapping in the same model metadata file.
• theopenco/llmgateway#3125 — Added reasoning-effort declarations for Grok 4.3 mappings.
• theopenco/llmgateway#3438 — Updated Azure Grok 4.3 reasoning-effort handling and related tests.

Suggested reviewers: smakosh

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely identifies the correction to xAI reasoning_effort support.
Linked Issues check	✅ Passed	The changes add reasoning_effort support to both Azure AI Foundry mappings and add regression tests for forwarding and omission behavior [#3436].
Out of Scope Changes check	✅ Passed	The additional model-specific mappings and exclusions support the stated reasoning_effort correction and related model behavior.

✨ Finishing Touches 💡 1

📝 Generate docstrings 💡

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch claude/e2e-verify-pr-3438-lqfntw

---

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

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}