anthonypdawson / vector-inspector / 22865738807

81%

Build Type

push

github

Committed by web-flow

Commit Message

Llm provider support (#25)

* feat: implement LLM provider architecture with Ollama, llama-cpp, and OpenAI-compatible support

* feat: add optional LLM runtime support with llama-cpp-python and update installation instructions

* feat: add documentation for LLM provider architecture, fake provider spec, and quickstart guide

* feat: add installation scripts for llama-cpp-python on Unix and Windows, and create documentation for LLM provider quickstart and fake provider

* feat: Implement LLM Runtime Manager and Provider Interface

- Added LLMRuntimeManager for managing LLM provider lifecycle, including provider selection, health caching, and request ID generation.
- Introduced types for LLM provider capabilities, health results, and streaming events in a new types module.
- Created a FakeLLMProvider for testing, supporting echo, streaming, and error injection modes.
- Updated tests to cover new functionality, ensuring deterministic behavior and health checks.
- Renamed and reorganized tests for clarity and improved structure.

* feat: Enhance LLM provider configuration panel with health check and model selection

* feat: Set maximum width for OpenAI key input field in LLM settings panel

* Refactor LLM provider integration and enhance settings management

- Introduced LLMRuntimeManager for improved management of LLM providers, including lazy initialization and health caching.
- Updated AppState to utilize LLMRuntimeManager, providing a clearer API for accessing LLM providers.
- Modified LLMProviderInstance tests to reflect changes in method names and behavior.
- Added comprehensive tests for new LLMRuntimeManager functionality, including provider selection precedence and health caching.
- Enhanced settings service with getters and setters for LLM provider configurations, ensuring defaults are correctly handled.
- Created UI tests for the LLM settings panel, verifying the presence of components and their functionality.
- Improved error handling and resp... (continued)

Run Details

1033 of 1219 new or added lines in 17 files covered. (84.74%)

8 existing lines in 4 files now uncovered.

12499 of 15597 relevant lines covered (80.14%)

0.8 hits per line