test(local-ai): serialize Ollama env mutations

[Zavianx]

Summary

• add a poison-recovering local_ai_test_guard() wrapper around the existing local-AI test mutex
• route Ollama-related test env mutations through the same local-AI mutex, including OPENHUMAN_OLLAMA_BASE_URL, OLLAMA_BIN, and local PATH lookups
• align memory/voice/local-ai tests with the domain mutex documented in src/openhuman/local_ai/README.md

Why

While validating #1589 locally, the Rust coverage path exposed a flaky race where one test could mutate or clear Ollama process-global env vars while another local-AI test was reading them. This keeps the existing test-only isolation model but applies it consistently across the affected modules. No production behavior changes.

Test plan

• cargo fmt --check
• CARGO_INCREMENTAL=0 cargo test -p openhuman openhuman::local_ai -- --nocapture
• CARGO_INCREMENTAL=0 cargo test -p openhuman openhuman::memory::store::factories::tests -- --nocapture
• sudo -n docker run ... ghcr.io/tinyhumansai/openhuman_ci:rust-1.93.0 ... cargo llvm-cov -p openhuman --lcov --output-path /tmp/lcov-core.info
• pre-push hook: format, lint, TypeScript compile, Tauri cargo check

Summary by CodeRabbit

• Tests
  • Improved test synchronization and reliability by consolidating test lock management across multiple test suites. This reduces potential test flakiness when running tests in parallel and strengthens the testing infrastructure for AI and voice processing modules.

[coderabbitai]

📝 Walkthrough

Walkthrough

This PR introduces a centralized test-only helper function local_ai_test_guard() that serializes access to process-global local-AI state across the test suite. Multiple test files are refactored to use this single guard instead of direct mutex locking, reducing duplication and preventing test flakiness from concurrent environment variable and health-status mutations.

Changes

Test synchronization via centralized local-AI guard

Layer / File(s)	Summary
Central guard helper definition src/openhuman/local_ai/mod.rs	Introduces pub(crate) fn local_ai_test_guard() that locks LOCAL_AI_TEST_MUTEX with poison recovery via into_inner(), providing a unified guard for all test synchronization.
Local-AI module test synchronization src/openhuman/local_ai/install.rs, src/openhuman/local_ai/ollama_api.rs	Updates helper functions in install.rs and ollama_api.rs test modules to delegate to the new centralized guard, consolidating environment and Ollama URL synchronization.
Service and inference test synchronization src/openhuman/local_ai/service/ollama_admin_tests.rs, src/openhuman/local_ai/service/public_infer_tests.rs, src/openhuman/local_ai/service/vision_embed.rs	Refactors 12 async test cases in ollama_admin_tests.rs and 3 test entry points in public_infer_tests.rs and vision_embed.rs to acquire the guard at test start, serializing health checks, model queries, and embedding API calls.
Cross-module test synchronization refactor src/openhuman/memory/store/factories.rs, src/openhuman/voice/postprocess.rs	Updates EnvGuard RAII logic and test entry points to use the centralized guard instead of module-local locks, including race condition prevention for the Ollama health-status latch during parallel execution; 7 transcription cleanup tests and 2 embedding tests refactored.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Possibly related PRs

• tinyhumansai/openhuman#1443: Both PRs focus on fixing test flakiness by serializing access to process-global singletons/mutable state via test-only mutex guards in their respective modules (local-AI env/ollama guards in main PR vs autocomplete global engine guards in #1443).
• tinyhumansai/openhuman#911: Both PRs update the voice postprocess test synchronization to serialize shared local AI/Ollama state using LOCAL_AI_TEST_MUTEX (direct lock vs the new local_ai_test_guard() wrapper).

Suggested reviewers

• senamakel

Poem

🐰 A scattered lock, once many and wide,
Now gathers in one where tests safely hide.
Guard local_ai threads that race through the night—
One mutex, one truth, one test gleaming bright! ✨

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'test(local-ai): serialize Ollama env mutations' accurately summarizes the main change: a refactoring to centralize test environment variable locking around Ollama-related mutations through a new test guard wrapper.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

---

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 and usage tips.}

[coderabbitai]

🧹 Nitpick comments (1)

src/openhuman/local_ai/mod.rs (1)

7-12: ⚡ Quick win

Move test guard implementation out of mod.rs to keep it export-focused.

Line 8 adds operational logic in a domain mod.rs. Prefer re-exporting from a sibling test-support module and keeping this file as declarations/re-exports only.

Proposed refactor (module stays export-focused)

 #[cfg(test)]
-pub(crate) fn local_ai_test_guard() -> std::sync::MutexGuard<'static, ()> {
-    LOCAL_AI_TEST_MUTEX
-        .lock()
-        .unwrap_or_else(|p| p.into_inner())
-}
+mod test_support;
+#[cfg(test)]
+pub(crate) use test_support::local_ai_test_guard;

As per coding guidelines: src/openhuman/*/mod.rs: Keep domain mod.rs files light and export-focused; place operational code in sibling files (e.g., ops.rs, store.rs, schedule.rs, types.rs) and re-export the public API from mod.rs.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/openhuman/local_ai/mod.rs` around lines 7 - 12, The function
local_ai_test_guard and its use of LOCAL_AI_TEST_MUTEX implement operational
test logic inside mod.rs; move that implementation into a sibling module (e.g.,
test_support.rs or ops.rs) where you define pub(crate) fn local_ai_test_guard()
-> std::sync::MutexGuard<'static, ()> and the LOCAL_AI_TEST_MUTEX there, then in
mod.rs replace the implementation with a simple re-export (pub(crate) use
test_support::local_ai_test_guard;) so mod.rs remains export-focused and the
operational code lives in the sibling module.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@src/openhuman/local_ai/mod.rs`:
- Around line 7-12: The function local_ai_test_guard and its use of
LOCAL_AI_TEST_MUTEX implement operational test logic inside mod.rs; move that
implementation into a sibling module (e.g., test_support.rs or ops.rs) where you
define pub(crate) fn local_ai_test_guard() -> std::sync::MutexGuard<'static, ()>
and the LOCAL_AI_TEST_MUTEX there, then in mod.rs replace the implementation
with a simple re-export (pub(crate) use test_support::local_ai_test_guard;) so
mod.rs remains export-focused and the operational code lives in the sibling
module.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 434a8e87-7543-47f8-8f2e-777c522a7edd

📥 Commits

Reviewing files that changed from the base of the PR and between df189c8 and 8d51c8d.

📒 Files selected for processing (8)

• src/openhuman/local_ai/install.rs
• src/openhuman/local_ai/mod.rs
• src/openhuman/local_ai/ollama_api.rs
• src/openhuman/local_ai/service/ollama_admin_tests.rs
• src/openhuman/local_ai/service/public_infer_tests.rs
• src/openhuman/local_ai/service/vision_embed.rs
• src/openhuman/memory/store/factories.rs
• src/openhuman/voice/postprocess.rs