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.
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
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).