Fix/lint and ci

[EggerMarc]

Summary by CodeRabbit

• Examples & Documentation

  • Simplified human-in-the-loop approval prompts to accept a single confirmation input.
  • Updated Google Maps example coordinates.
  • Added macOS Metal-backend run instructions for the MistralRS voice example.

• Code Quality

  • Various internal refactors and formatting improvements for consistency.
  • Added Clippy lint suppressions to reduce noisy warnings across providers and adapters.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: bf8a6bbe-7792-4cf0-bb2a-8d5d431ee334

📥 Commits

Reviewing files that changed from the base of the PR and between 68bdf00 and bca5ff7.

📒 Files selected for processing (6)

• examples/claude/hitl.rs
• examples/openai/hitl.rs
• examples/router/keyword.rs
• examples/router/stream.rs
• providers/router/src/router.rs
• providers/router/src/stream_router.rs

✅ Files skipped from review due to trivial changes (1)

• examples/router/keyword.rs

---

📝 Walkthrough

Walkthrough

Mechanical cleanup across the codebase: derive Default where possible, standardize CallId fallbacks, add documented Clippy allow attributes, simplify router/stream guard patterns, tighten streaming parsing conditions, reflow formatting, and simplify example prompts. No public API changes.

Changes

Code Cleanup and Refactoring

Layer / File(s)	Summary
Default trait derivation and usage simplification core/src/types/provider_meta.rs, providers/completions/src/api/stream.rs, core/src/types/messages/parts.rs, core/src/types/messages/tool.rs, core/src/types/metadata/usage.rs	ProviderMeta and StreamState now derive Default. Tool::new and PartEnum::from_function_response use unwrap_or_default() instead of CallId::new(). Tests removed redundant ..Default::default() usage.
Clippy lint suppressions with documentation core/src/chat/stream/handle.rs, providers/claude/src/api/types/error.rs, providers/completions/src/api/types/error.rs, providers/completions/src/api/types/response.rs, providers/gemini/src/api/types/error.rs, providers/gemini/src/api/types/request.rs, providers/mistralrs/src/lib.rs, providers/responses/src/api/types/error.rs	Added #[allow(...)] attributes with explanatory comments: large_enum_variant, result_large_err, wrong_self_convention, and too_many_arguments in the noted modules.
Router circuit-breaker and error handling simplification providers/router/src/router.rs, providers/router/src/stream_router.rs	Refactored provider iteration guards and resolve_order error mapping to use map_err(ChatFailure::from_err)?, short-circuit non-retryable failures, and condensed the last-resort longest_open() fallback into single chained conditions.
Streaming event parsing and delta handling refactoring core/src/chat/mod.rs, providers/claude/src/api/stream.rs, providers/mistralrs/src/api/stream.rs	Use Option::replace for pause transitions in tool approval. Claude stream parsing and MistralRS delta/final-message checks consolidated into combined conditionals.
Formatting, struct construction, and minor simplifications core/src/chat/stream/input.rs, core/src/chat/stream/mod.rs, providers/gemini/src/api/types/request.rs, providers/mistralrs/src/api/types/request.rs, providers/mistralrs/src/api/types/response.rs, providers/openai/src/api/embedding.rs, providers/openai/src/client.rs, providers/responses/src/lib.rs, src/lib.rs	Reflowed struct literals and calls for readability; GeminiPart construction moved to struct literals with ..Default::default(); minor single-line formatting tweaks for headers, returns, and re-exports.
Example code refactoring and documentation updates examples/claude/hitl.rs, examples/gemini/google_maps.rs, examples/mistralrs/voice.rs, examples/openai/hitl.rs, examples/router/*	HITL prompts simplified to single prompt + match; Google Maps coordinates tweaked; MistralRS voice example docs include macOS Metal run snippet; router example sort logic switched to sort_by_key(Reverse(...)).
CI: ALSA dev deps .github/workflows/rust.yml	Install libasound2-dev and pkg-config in check and build jobs before lint/build steps.

Sequence Diagram(s)

sequenceDiagram
  participant Router
  participant CircuitBreaker
  participant Provider
  Router->>CircuitBreaker: resolve_order(request)
  Router->>Provider: call provider (if available)
  Provider-->>Router: response or error
  Router->>CircuitBreaker: record_success()/record_failure()

Loading

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

• EggerMarc/chat-rs#28: Overlaps ProviderMeta Default changes and provider metadata usage.
• EggerMarc/chat-rs#43: Related edits to the streaming Input enum and stream handle API.

Poem

🐰 A cleanup hop, so neat and clean,
Defaults derived, no dupes between,
Clippy notes penned with gentle care,
Guards chained tight, parsers trimmed and spare,
Examples trimmed — the code can preen!

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 inconclusive)

Check name	Status	Explanation	Resolution
Title check	❓ Inconclusive	The title 'Fix/lint and ci' is vague and uses generic terms that don't convey meaningful information about the specific changes in this multi-faceted changeset.	Consider a more descriptive title that highlights the primary change, such as 'Refactor error handling, clippy suppressions, and CI setup' or 'Add ALSA dependencies and refactor for clippy compliance'.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
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.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/lint-and-ci

---

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]

Actionable comments posted: 6

🧹 Nitpick comments (1)

providers/mistralrs/src/lib.rs (1)

23-26: 💤 Low value

Consider function-level suppression for consistency with other providers.

While the crate-level #![allow(clippy::result_large_err)] avoids repetition, other providers (Claude, Completions, Gemini, Responses) use function-level #[allow(clippy::result_large_err)] on their error handlers. Crate-level attributes can hide future lint violations in new code that might not need the suppression.

If multiple functions truly need this suppression, the crate-level approach is reasonable. Otherwise, consider using function-level attributes for consistency and to limit the scope of the suppression.

🤖 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 `@providers/mistralrs/src/lib.rs` around lines 23 - 26, The crate currently
uses a crate-level suppression `#![allow(clippy::result_large_err)]`; change
this to function-level suppressions to match other providers by removing the
crate-level attribute and adding `#[allow(clippy::result_large_err)]` only to
the specific error-handling/conversion functions (e.g., the conversion functions
that return `ChatFailure`/`ChatError` and any other functions producing large
Err types) so the lint scope is limited and consistent with
Claude/Completions/Gemini/Responses.

🤖 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.

Inline comments:
In `@core/src/types/provider_meta.rs`:
- Around line 6-8: The ProviderMeta struct currently exposes data as
HashMap<String, Box<dyn Any + Send + Sync>> which violates the repo contract;
change the data field to use JSON values (HashMap<String, serde_json::Value>) in
the ProviderMeta definition, remove the Box<dyn Any + ...> usage, import
serde_json::Value, and update any constructors, getters, setters, impls, and
(de)serialization code that reference ProviderMeta::data to accept and return
HashMap<String, Value> so provider metadata is represented as JSON values across
the codebase.

In `@examples/claude/hitl.rs`:
- Around line 103-110: The match on prompt(...) uses raw input that may include
CR or case differences; capture the prompt result into a variable, call
.trim().to_lowercase() on it, and then match that normalized string so inputs
like "y\r" or "Y" map to "y" and produce Decision::Approve (reference the
prompt(...) call and the Decision::Approve / Decision::Reject arms).

In `@examples/openai/hitl.rs`:
- Around line 103-110: The match on the raw prompt result can misclassify inputs
like "y\r" on Windows; normalize the string from prompt() before matching by
trimming whitespace and converting to lowercase (e.g., call trim() and
to_lowercase() on the prompt() result) and then match against normalized values
so Decision::Approve and Decision::Reject branches (Decision::Approve,
Decision::Reject(...)) behave correctly across platforms.

In `@providers/gemini/src/api/types/request.rs`:
- Around line 190-213: The current PartEnum::Tool branch calls tool.to_tuple()
and unconditionally serializes a GeminiFunctionCall into assistant_parts, which
can leak unresolved Tool states; change this to use tool.try_to_tuple() (or
explicitly check ToolStatus and early-return/skip for Pending/Approved/Running)
and only build GeminiPart/GeminiFunctionCall and function_parts when
try_to_tuple() yields Some((fc, maybe_fr)) (i.e., the tool is in a resolved
state like Completed/Rejected/Failed); update code paths that reference
assistant_parts, function_parts, GeminiPart, and GeminiFunctionCall to only run
when the tool is resolved.

In `@providers/router/src/router.rs`:
- Around line 57-61: The loop over providers currently records every provider
error into last_failure and continues, which violates the contract; change the
error handling inside the loop that calls provider.complete() (and the
surrounding block that checks self.circuit_breaker and order iteration) so that:
for each provider call, if it succeeds return immediately; if it fails and the
returned ChatError::is_retryable() is true, set last_failure to that error and
continue to the next provider; if it fails and is_retryable() is false
(non-retryable errors such as Provider, InvalidResponse, MaxStepsExceeded,
Callback, Other), return that error immediately without updating last_failure;
ensure this logic replaces the current unconditional last_failure assignment in
the loop (affecting the blocks around where last_failure is set and where
iteration advances).

In `@providers/router/src/stream_router.rs`:
- Around line 46-50: Both complete and stream currently stash every provider
failure and continue, causing non-retryable (fatal) provider errors to be
retried or masked; update the failure handling in the retry/fallback loops
inside complete and stream so that when an attempt returns an error you check
error.is_retryable() (or equivalent) and if it returns false you immediately
return that error instead of pushing it into the failures list and continuing;
remove/stash only retryable errors and ensure the same change is applied to all
fallback loops handling provider responses (the loops around order iteration and
circuit_breaker checks) so non-retryable errors short-circuit the method.

---

Nitpick comments:
In `@providers/mistralrs/src/lib.rs`:
- Around line 23-26: The crate currently uses a crate-level suppression
`#![allow(clippy::result_large_err)]`; change this to function-level
suppressions to match other providers by removing the crate-level attribute and
adding `#[allow(clippy::result_large_err)]` only to the specific
error-handling/conversion functions (e.g., the conversion functions that return
`ChatFailure`/`ChatError` and any other functions producing large Err types) so
the lint scope is limited and consistent with
Claude/Completions/Gemini/Responses.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: f73c26bc-8057-4ebe-98eb-d78789f45d11

📥 Commits

Reviewing files that changed from the base of the PR and between 11e8822 and d110020.

📒 Files selected for processing (30)

• core/src/chat/mod.rs
• core/src/chat/stream/handle.rs
• core/src/chat/stream/input.rs
• core/src/chat/stream/mod.rs
• core/src/types/messages/parts.rs
• core/src/types/messages/tool.rs
• core/src/types/metadata/usage.rs
• core/src/types/provider_meta.rs
• examples/claude/hitl.rs
• examples/gemini/google_maps.rs
• examples/mistralrs/voice.rs
• examples/openai/hitl.rs
• providers/claude/src/api/stream.rs
• providers/claude/src/api/types/error.rs
• providers/completions/src/api/stream.rs
• providers/completions/src/api/types/error.rs
• providers/completions/src/api/types/response.rs
• providers/gemini/src/api/types/error.rs
• providers/gemini/src/api/types/request.rs
• providers/mistralrs/src/api/stream.rs
• providers/mistralrs/src/api/types/request.rs
• providers/mistralrs/src/api/types/response.rs
• providers/mistralrs/src/lib.rs
• providers/openai/src/api/embedding.rs
• providers/openai/src/client.rs
• providers/responses/src/api/types/error.rs
• providers/responses/src/lib.rs
• providers/router/src/router.rs
• providers/router/src/stream_router.rs
• src/lib.rs

💤 Files with no reviewable changes (1)

• core/src/types/metadata/usage.rs