fix(agent-core): gate prompt_cache_key behind isOfficialOpenAIBaseUrl for third-party endpoints

[creatiVision]

Problem

Since prompt_cache_key was introduced for session cache affinity, Kimi Code CLI sends this parameter to all OpenAI-compatible providers — not just to official OpenAI endpoints. Third-party APIs that strictly validate request parameters reject it with HTTP 400, making providers like NVIDIA NIM and Azure Foundry unusable.

• Error with models not supporting prompt cache #2166 — NVIDIA NIM: 400 Validation: Unsupported parameter(s): prompt_cache_key
• MS Foundry + Kimi Code 2.7 - 400 Unrecognized request argument supplied: prompt_cache_key #2611 — Azure Foundry: 400 Unrecognized request argument supplied: prompt_cache_key

Both issues remain open because the legacy engine (packages/agent-core) lacks the guard that agent-core-v2 already has.

Root cause

packages/agent-core/src/session/provider-manager.ts unconditionally injected prompt_cache_key into generationKwargs for every openai and openai_responses provider configuration, regardless of the actual target base URL:

// Before (openai case):
generationKwargs: { prompt_cache_key: promptCacheKey },

The agent-core-v2 engine already solved this via isOfficialOpenAIBaseUrl() — prompt_cache_key is only sent when the hostname targets api.openai.com or a regional *.api.openai.com variant. Non-Kimi vendors that support the field on other hosts opt in through a cacheKey hook.

Changes

packages/agent-core/src/session/provider-manager.ts (1 file, +34 / −5):

• Added isOfficialOpenAIBaseUrl() helper — same logic as the v2 implementation
• openai case: gated prompt_cache_key behind isOfficialOpenAIBaseUrl
• openai_responses case: same gating
• kimi case: unchanged — Kimi API always accepts prompt_cache_key
• All three cases: generationKwargs is now entirely omitted when promptCacheKey is undefined, instead of sending prompt_cache_key: undefined on the wire

Behaviour after fix

Provider config	Base URL	prompt_cache_key sent?
OpenAI	https://api.openai.com/v1 (default)	✅ Yes
Kimi	https://api.moonshot.cn/v1	✅ Yes (direct provider)
NVIDIA NIM	https://integrate.api.nvidia.com/v1	❌ No
Azure Foundry	https://{resource}.openai.azure.com/v1	❌ No
Any custom OpenAI-compatible	any other host	❌ No

Checklist

• I have read the CONTRIBUTING document.
• I have linked related issues above.
• I have added tests that prove my feature works. (Existing test suite covers this code path.)
• Ran gen-changesets skill, or this PR needs no changeset.
• Ran gen-docs skill, or this PR needs no doc update.

[changeset-bot]

🦋 Changeset detected

Latest commit: 601e041

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
@moonshot-ai/kimi-code	Patch

Not sure what this means? Click here to learn what changesets are.

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

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e430cb316b

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 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 address that feedback".

[chatgpt-codex-connector]

Add a changeset for this provider fix

This is a user-visible provider behavior change, but the commit only changes provider-manager.ts and adds no .changeset/ entry, so changeset-based release/version notes can omit the third-party endpoint fix. Add an appropriate changeset before this PR is submitted.

AGENTS.md reference: AGENTS.md:L84-L85

Useful? React with 👍 / 👎.