weft-graph 1.0.0

A local-first knowledge graph for your browsing

Weft

A local-first knowledge graph for your browsing.

Weft turns your browser tabs into a searchable, clustered knowledge graph. Instead of drowning in hundreds of tabs or flat bookmark lists, Weft groups related pages, removes duplicates, and lets you explore your browsing context using a fast terminal UI.

Think: Obsidian graph view, but for the web you already opened.

Why Weft?

Tool	What it does
OneTab	Saves links in a flat list
Bookmarks	Folders you'll never organize
Weft	Builds memory from your browsing

Features

• Knowledge Graph - Pages connected by semantic similarity, not folders
• Automatic Clustering - Groups related content by meaning and domain
• Near-Duplicate Detection - SimHash identifies similar pages across sites
• Smart Search - Fuzzy text, keyword (#topic), and domain (@site) filters
• Graph Explorer - Navigate neighbors, discover related content
• Optional LLM Summaries - Local AI via Ollama for richer understanding
• Fully Local - Your data never leaves your machine

Install

pip install weft

Or from source:

git clone https://github.com/yourusername/weft.git
cd weft
pip install -e .

Quick Start

# Build knowledge graph from your browser tabs
weft weave

# Explore in terminal UI
weft explore

Commands

weft weave

Extracts tabs from browsers and weaves them into a knowledge graph.

# From all browsers (default)
weft weave

# Chrome only
weft weave --browser chrome

# Firefox only
weft weave --browser firefox

# With LLM summaries (requires Ollama)
weft weave --summarize

# Fast mode (tab titles only, no crawling)
weft weave --no-crawl

weft explore

Interactive TUI for exploring your knowledge graph.

# Default graph
weft explore

# Specific file
weft explore my_graph.json

Keyboard Shortcuts

Key	Action
[ / ]	Navigate groups
{ / }	Navigate tabs in group
o	Open tab in browser
g	Open all tabs in group
v	Switch to graph view
s	Focus search
q	Quit

Search syntax:

• Fuzzy: distributed systems
• Keyword: #performance
• Domain: @github

How It Works

Tabs → Extract → Cluster → Graph → Explore
         │          │         │
         ▼          ▼         ▼
     Keywords   SimHash   Similarity
      TF-IDF    Dedup     Matrix

1. Extract - Pulls tabs from Chrome/Firefox, optionally crawls page content
2. Cluster - Groups by semantic similarity + domain affinity
3. Deduplicate - SimHash finds near-identical content across domains
4. Graph - Builds similarity edges between related pages
5. Explore - Fast TUI for search, navigation, and discovery

Requirements

• Python 3.9+
• macOS (browser export uses AppleScript)
• Chrome and/or Firefox

Optional: LLM Summaries

# Install Ollama
brew install ollama

# Pull models
ollama pull llama3.1:8b
ollama pull nomic-embed-text

# Use with summarization
weft weave --summarize

License

MIT