fix(cache): preserve client cache_control for Xiaomi MiMo (#3088)

[diegosouzapw]

Closes #3088

Problem

Xiaomi MiMo via cc-switch + OmniRoute showed zero cache hits, while calling Xiaomi directly cached normally.

Root cause

CACHING_PROVIDERS in open-sse/utils/cacheControlPolicy.ts did not include xiaomi-mimo. So providerSupportsCaching("xiaomi-mimo") returned false → shouldPreserveCacheControl() returned false → filterToOpenAIFormat() stripped every cache_control block before forwarding. Xiaomi's API therefore never saw the cache breakpoints and treated each request as a miss.

Fix

Add xiaomi-mimo to the prompt-caching provider allowlist (1 line). Now Claude Code's cache_control hints are preserved for single-model routing.

Tests

Added regression assertions to tests/unit/cache-control-policy.test.ts:

• providerSupportsCaching("xiaomi-mimo") → true (+ case-insensitive)
• shouldPreserveCacheControl({ userAgent: claude-cli, isCombo: false, targetProvider: "xiaomi-mimo" }) → true

node --import tsx/esm --test tests/unit/cache-control-policy.test.ts → 37 pass / 0 fail. ESLint clean.

[gemini-code-assist]

Code Review

This pull request adds the 'xiaomi-mimo' provider to the prompt-caching allowlist in open-sse/utils/cacheControlPolicy.ts to ensure client-side cache_control breakpoints are preserved and not stripped. It also includes corresponding unit tests in tests/unit/cache-control-policy.test.ts and updates the changelog. I have no feedback to provide as the changes are well-implemented and fully tested.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 576c9c5224

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".

[chatgpt-codex-connector]

Preserve cache_control in the OpenAI translator too

When Xiaomi MiMo is targeted through the normal OpenAI-format path, this allowlist entry only makes shouldPreserveCacheControl() return true; translateRequest() still unconditionally calls filterToOpenAIFormat() for targetFormat === FORMATS.OPENAI, and that helper still removes cache_control from every content block. In the Claude Code/cc-switch → xiaomi-mimo scenario described here, the upstream payload therefore still loses the cache breakpoints, so the regression remains unfixed unless the OpenAI-format filter also honors preserveCacheControl.

Useful? React with 👍 / 👎.

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

This is a clean, focused bug fix PR. It adds xiaomi-mimo to the CACHING_PROVIDERS set in open-sse/utils/cacheControlPolicy.ts and includes appropriate test coverage.

Files Reviewed (3 files)

• open-sse/utils/cacheControlPolicy.ts - Changes to CACHING_PROVIDERS constant
• tests/unit/cache-control-policy.test.ts - Added test for xiaomi-mimo caching support
• CHANGELOG.md - Updated with fix entry

Verification

• Provider ID xiaomi-mimo matches existing registration in src/shared/constants/providers.ts
• Test additions follow existing patterns and cover both providerSupportsCaching() and shouldPreserveCacheControl()
• Comments reference issue [BUG] Zero cache hits with Xiaomi MiMo via cc-switch and OmniRoute #3088 for traceability

No code quality issues, no security concerns, and no edge cases identified. The fix is minimal and well-targeted.

---

_{Reviewed by laguna-m.1-20260312:free · 526,955 tokens}