Ennoia
A Python framework for declarative document indexing, where typed schemas describe what to extract from a document and Ennoia runs the extraction, stores the structured output, and serves combined filter and vector search over it.
Ennoia lets a developer describe what to extract from each document as a typed Python schema; the framework then runs the extraction with an LLM, persists the structured output, and exposes hybrid filter-plus-vector search on top of it. The same schemas drive a Python SDK, a CLI, a REST API and an MCP tool surface.
On the project's own product-discovery benchmark, Ennoia's schema-driven retrieval leads a plain one-embedding-per-product baseline on queries that give few details, since embeddings alone have little to latch onto, and reaches parity on queries with rich detail.
It installs with pip and offers extras for hosted LLM adapters (OpenAI, Anthropic, OpenRouter), a local Ollama adapter, local embeddings, and several storage backends including filesystem, Qdrant and pgvector. The project is in alpha, with an API that is still converging.