fix: Rate limiter returns retry after None instead of a duration

[nickpismenkov]

When the nearai_chat provider hits its rate limit, the error message
shows:
"LLM error: Provider nearai_chat rate limited, retry after None"
The "None" value indicates the retry delay is not being extracted or set correctly.
This also means the agent has no backoff duration to work with and may retry
immediately (causing more rate limit hits) or not retry at all

[gemini-code-assist]

Summary of Changes

Hello, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request addresses a critical issue where rate-limited LLM and embedding providers would return an undefined retry duration, leading to immediate retries or no retries at all. By introducing a default 60-second fallback for the Retry-After header, the system can now gracefully handle rate limits even when the server response is incomplete or malformed, improving resilience and preventing excessive API calls.

Highlights

• Rate Limit Retry-After Fallback: Implemented a 60-second default fallback duration for Retry-After headers across Anthropic, NearAi Chat, OpenAI Embeddings, and NearAi Embeddings providers. This ensures that if the Retry-After header is missing or unparseable, the system will still have a defined delay instead of None.
• Comprehensive Unit Tests: Added extensive unit tests for Retry-After header parsing in all affected providers, covering valid delay-seconds, RFC2822 date formats (for NearAi Chat), missing headers, invalid formats, zero-second delays, and large delay values to ensure robust behavior.

Changelog

• src/llm/anthropic_oauth.rs
  • Implemented a 60-second fallback duration for Retry-After header parsing in AnthropicOAuthProvider to prevent None values.
  • Added new unit tests to verify correct parsing of Retry-After headers and the 60-second fallback mechanism.
• src/llm/nearai_chat.rs
  • Introduced a 60-second default retry duration for NearAiChatProvider when the Retry-After header is not present or cannot be parsed.
  • Expanded unit tests for Retry-After header parsing to cover delay-seconds, RFC2822 dates, and various fallback scenarios.
• src/workspace/embeddings.rs
  • Configured OpenAiEmbeddings and NearAiEmbeddings to use a 60-second default for retry_after when the header is missing or invalid.
  • Added unit tests to confirm the correct handling and fallback of Retry-After headers for embedding providers.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for GitHub and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request addresses a bug where rate limit errors would not include a retry-after duration, causing issues with backoff and retry logic. The fix involves adding a 60-second fallback duration when the Retry-After header is missing or unparseable. This change has been consistently applied across the anthropic_oauth, nearai_chat, and embeddings providers.

You've also added comprehensive regression tests for each provider to validate the new fallback behavior, which is excellent. My review includes a few suggestions to improve the maintainability of these new tests by refactoring the test helpers to avoid duplicating production logic, aligning with best practices for robust and maintainable test suites.

[zmanian]

Review: fix rate limiter "retry after None" display bug

Simple, correct fix. Adds .or(Some(Duration::from_secs(60))) as a fallback when the Retry-After header is missing or unparseable. All 4 HTTP Retry-After parsing sites are covered:

• anthropic_oauth.rs
• nearai_chat.rs
• embeddings.rs (OpenAiEmbeddings + NearAiEmbeddings)

I verified there are no other HTTP Retry-After parsing sites in the codebase -- other retry_after: None occurrences are internally constructed errors with their own fallback handling.

CI fully green.

Positives:

• Complete coverage of all affected sites
• 60s fallback is a reasonable default
• Good test coverage for each module

Minor notes:

• The test helpers (e.g. parse_retry_after_anthropic_for_test) reimplement parsing logic rather than testing actual production code. If someone later changes the production parsing chain but not the test helper, tests could pass while production is broken. Consider integration-level tests with mock HTTP responses in a follow-up.
• The retry.rs test (commit 3) constructs a RateLimited with Some(60s) and asserts it's Some(60s) -- this doesn't exercise any code path. It's harmless but not providing real regression protection.
• Option<Duration> is now effectively always-Some from these call sites. Minor type-level inaccuracy but not worth changing.

LGTM.