README.md

metadata

title: Feather DB — Living Context Engine
emoji: 🪶
colorFrom: indigo
colorTo: purple
sdk: gradio
sdk_version: 5.9.1
app_file: app.py
pinned: true
license: mit
tags:
  - vector-database
  - embeddings
  - knowledge-graph
  - hnsw
  - context-graph
  - llm
  - rag
  - agents
  - mcp

🪶 Feather DB — Living Context Engine

Embedded vector database with sub-millisecond HNSW search, typed context graph, and adaptive decay.

What this demo shows

Tab	What it does
🔍 Semantic Search	Find nodes by meaning — with namespace/entity/product filters
🕸️ Context Chain	Vector search + BFS graph expansion — traces root causes across edges
🔬 Why Retrieved?	Score breakdown: similarity × stickiness × recency × importance
🩺 Graph Health	Tier distribution, orphan nodes, recall histogram
➕ Add Intel	Ingest a new node — immediately searchable

Install

pip install feather-db

Connect to any LLM in 5 lines

from feather_db.integrations import ClaudeConnector
conn = ClaudeConnector(db_path="my.feather", dim=3072, embedder=embed_fn)
result = conn.run_loop(client,
    messages=[{"role": "user", "content": "Why is our FD CTR dropping?"}],
    model="claude-opus-4-6")

Works with Claude, OpenAI, Gemini, Groq, Mistral, Ollama and any MCP-compatible agent (Claude Desktop, Cursor).

Links

• PyPI
• GitHub
• Integrations Guide
• Hawky.ai