Documentation
Feather DB
Overview.
Embedded vector database + living context engine. Zero-server, file-based, powered by C++ and Rust.
Version 0.5.0
- Multimodal Pockets
- Context Graph & Chain
- Adaptive Decay
- Rust CLI Tooling
Core Features (v0.5.0)
Multimodal Pockets
Traditional vector databases force you to concatenate or separate inputs into different tables. Feather allows a single entity ID to hold text, visual, and audio vectors simultaneously within the same flat file.
db.add(id=42, vec=text_vec, modality="text")
db.add(id=42, vec=img_vec, modality="visual")
db.add(id=42, vec=img_vec, modality="visual")
Context Graph & Chain
Don't just store vectors; store their relationships natively. Use the Context Graph to link nodes. Then use Context Chain to seamlessly combine semantic vector search with N-hop BFS graph traversal.
res = db.context_chain( query=vec_q, k=5, # seed searchhops=2, # BFS expansionmodality="text")
Adaptive Decay (Living Context)
Vectors naturally decay over time based on a configurable half-life. Older information gives way to newer, more relevant context automatically, solving context overload in long-running RAG agents.
FEATHER DB
Next: Installation