v0.1.0

Initial Release

High-performance OpenAI-compatible embedding server for Roo Code codebase indexing on Apple Silicon.

Features

• OpenAI-compatible /v1/embeddings endpoint — drop-in for Roo Code
• Qwen3 embedding models (0.6B, 4B, 8B) via MLX
• Request coalescing for concurrent batch processing
• Length-bucketed sub-batching with compiled Metal kernels
• LRU embedding cache for instant re-indexing
• Up to 5x faster than Ollama for the same models

Install

uvx roocode-code-indexer-macos