agentjido / req_llm / 8fcf7e948733a98689e2ff8c2e5389ef4009deb2

49%

Build Type

push

github

Committed by web-flow

Commit Message

Refactor LLM Fixture System (#22)

* Scrub all legacy fixtures

* Move core tests to the new LLM Fixture system

* upgrade LLM fixtures for tool calling tests

* Refactor and enhance LLM fixture system

- Removed the legacy `LiveFixture` module to streamline the testing process.
- Updated `LLMFixture` to improve handling of live API calls and cached responses.
- Added new streaming test fixtures for various providers including Anthropic, Google, Groq, OpenAI, OpenRouter, and XAI to support comprehensive testing of streaming capabilities.
- Enhanced the `streaming` test suite to ensure robust validation of chunked responses and content verification.

This update improves the overall testing framework and ensures better coverage for streaming interactions across different LLM providers.

* Enhance streaming tests and timing validation

- Updated the LLMFixture to improve handling of streaming responses, ensuring proper reconstruction of raw SSE streams.
- Added new timing tests for `stream_text!` to validate immediate return behavior in live scenarios, ensuring it does not block until completion.
- Introduced new timing test fixtures for both Anthropic and OpenAI providers to support comprehensive performance validation.
- Adjusted existing test fixtures to reflect updated timestamps and response data for accurate testing.

These enhancements improve the robustness of the testing framework and ensure better coverage for streaming interactions across different LLM providers.

* Enhance LLMFixture for timing-aware streaming and update test fixtures

- Introduced timing configuration for replay in the LLMFixture, allowing for simulation of delays and acceleration during streaming tests.
- Updated the `create_timing_stream` function to handle timing metadata in streaming responses, ensuring accurate simulation of chunked data delivery.
- Adjusted existing test fixtures for Anthropic to reflect new timestamps and response data, improving the accuracy of timi... (continued)

Run Details

6 of 250 new or added lines in 6 files covered. (2.4%)

7 existing lines in 2 files now uncovered.

532 of 2527 relevant lines covered (21.05%)

446.87 hits per line