- local Ollama classic RAG with directory ingestion;
- hosted OpenAI classic RAG;
- an OpenAI-compatible custom endpoint;
- agentic RAG with custom tools and agents;
- MCP tool discovery and execution.
pipeline.ingest(...), and then use a session for the chat loop. This keeps the
framework API independent from a particular terminal UX or configuration convention.
See Getting started for a complete local program.