Skip to content

Latest commit

 

History

History

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 

README.md

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