Chat Format

This handles the tedious work of reformatting prompts between different LLM providers so you're not stuck writing conversion logic when you want to switch from Claude to GPT or spin up a local Ollama model. It covers Anthropic, OpenAI, Google, Cohere, and local formats. The HNSW indexing tools are interesting because you get vector similarity search for RAG pipelines without standing up a separate database. Realistically, you'll use the chat formatting constantly if you're building anything multi-provider, and the HNSW stuff is solid for prototyping retrieval features before committing to Pinecone or whatever.