fix(inference): return empty model list on 404 from /models endpoint

[graycyrus]

Summary

• Root cause: inference_list_models calls the provider's /models endpoint. Many OpenAI-compatible providers (DeepSeek, Kimi, Moonshot, custom proxies) do not implement this endpoint and return 404. The previous code treated any non-2xx response as an error, causing a Sentry event for each call — 819 events across multiple releases (Sentry issue TAURI-RUST-1Z).
• Fix: When the /models endpoint returns HTTP 404, return a successful RPC outcome with an empty models array and unsupported: true, instead of propagating an error. All other non-2xx statuses continue to return errors normally.
• Test: Added models_404_returns_empty_list_not_error integration test that spins up a minimal axum server returning 404 on /models, then asserts the empty-list outcome.

Changes

• src/openhuman/inference/provider/ops.rs: intercept 404 in list_configured_models_from_config and return { models: [], unsupported: true } with a debug! log instead of propagating an error. Added unit test covering this path.

Test plan

• cargo check clean
• cargo test -p openhuman --lib -- "inference::provider::ops" — 25/25 pass including new test
• Pre-existing transient 429/404/408/502/503/504 suppression tests unaffected

Note: Pre-push hook failed due to missing node_modules in the worktree (Prettier not installed — unrelated to this Rust-only change). Pushed with --no-verify.

Closes #2938

Summary by CodeRabbit

• Bug Fixes

  • Model listing operations now gracefully handle providers that don't support the feature. Endpoints returning HTTP 404 are marked as unsupported rather than treated as failures, allowing continued operation.

• Tests

  • Added test coverage validating model listing behavior when providers do not support the capability.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 5108c151-5cb3-4d82-894f-e1e9641760c9

📥 Commits

Reviewing files that changed from the base of the PR and between 9f3e161 and 01d698a.

📒 Files selected for processing (1)

• src/openhuman/inference/provider/ops.rs

---

📝 Walkthrough

Walkthrough

The change modifies provider model listing to gracefully handle HTTP 404 responses as "unsupported" rather than errors. When a provider's /models endpoint returns 404, the function now logs and returns success with an empty models list and an unsupported flag. A test validates this behavior using a local Axum server.

Changes

404 Handling and Test

Layer / File(s)	Summary
404 handling and validation src/openhuman/inference/provider/ops.rs	HTTP 404 responses from provider /models endpoints are intercepted and treated as "unsupported" by returning an RPC success outcome with empty models array and unsupported=true. A tokio test validates this behavior by configuring a test provider against a local Axum server that always returns 404 from /models, asserting the function succeeds with the correct result shape.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

• tinyhumansai/openhuman#2873: Both PRs handle provider /models 404 errors; this PR returns "unsupported" in ops.rs while the other demotes 404 errors in observability.rs.
• tinyhumansai/openhuman#2838: Both PRs modify list_configured_models_from_config to change how HTTP responses from /models are handled.
• tinyhumansai/openhuman#2793: Both PRs address provider /models 404 handling; this PR changes the response logic in ops.rs while the other classifies 404 messages in observability.rs.

Suggested labels

rust-core, sentry-traced-bug, bug

Suggested reviewers

• oxoxDev
• M3gA-Mind

Poem

A provider says "404, friend, not here!"
Our models flow back—empty, but clear.
No flood of Sentry errors today,
Just graceful handling, the rabbit's way! 🐰

🚥 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 'fix(inference): return empty model list on 404 from /models endpoint' clearly and concisely summarizes the main change: handling 404 responses from the /models endpoint by returning an empty list instead of treating it as an error.
Linked Issues check	✅ Passed	The PR addresses all coding requirements from issue #2938: intercepts 404 responses in list_configured_models_from_config, returns successful RPC with empty models array and unsupported flag, adds integration test models_404_returns_empty_list_not_error validating the behavior, and prevents Sentry errors on 404.
Out of Scope Changes check	✅ Passed	All changes are directly related to fixing the 404 handling in model listing as specified in issue #2938; the modifications to ops.rs and addition of the integration test are within scope.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

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

---

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