weft-graph 1.1.1

A local-first knowledge graph for your browsing - Chrome extension + CLI

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 visually.

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

Two Ways to Use Weft

	Chrome Extension	CLI (Python)
Best for	Live tracking, visual exploration	Batch processing, scripting
Graph View	Interactive sidepanel	Terminal UI
Navigation Tracking	Automatic	Manual export
Install	Load unpacked extension	pip install weft-graph

---

Chrome Extension

Live knowledge graph that tracks your browsing in real-time.

Features

• Live Tab Tracking - Automatically captures tabs as you browse
• Navigation Edges - Tracks how you move between pages (including SPAs)
• Graph Visualization - Interactive Cytoscape.js graph with zoom/pan
• Smart Grouping - Clusters related pages by content similarity
• Keyword Extraction - Automatic keyword detection from page content
• Search - Fuzzy text, #keyword, and @domain filters
• Import/Export - Compatible with CLI JSON format

Install Extension

1. Clone or download this repo
2. Open Chrome → chrome://extensions
3. Enable "Developer mode" (top right)
4. Click "Load unpacked" → select the extension/ folder
5. Click the Weft icon or open sidepanel

Extension Usage

Views:

• Groups - Browse tabs organized by topic clusters
• Graph - Visual knowledge graph with similarity and navigation edges

Search Syntax:

• Fuzzy: distributed systems
• Keyword: #database
• Domain: @github.com

Actions:

• Click any tab to see details (URL, keywords, group)
• Click "Open Tab" to open in browser
• Use refresh button to rebuild graph after browsing
• Export/Import for backup or CLI compatibility

---

CLI Tool

Batch processing and terminal UI for exploring your knowledge graph.

Install CLI

pip install weft-graph

Or from source:

git clone https://github.com/Avi-141/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

---

How It Works

Tabs → Extract → Analyze → Dedupe → Cluster → Graph → Explore
         │          │         │         │        │
         ▼          ▼         ▼         ▼        ▼
      Browser    Keywords   SimHash   Union   Similarity
       Tabs      Content    Matching   Find    + Navigation

1. Extract - Captures tabs from browser (live or batch)
2. Analyze - Extracts keywords from page content
3. Deduplicate - Canonical URL matching + SimHash for near-duplicates
4. Cluster - Groups by similarity using Union-Find algorithm
5. Graph - Builds similarity edges + navigation edges between pages
6. Explore - Visual graph or terminal UI for discovery

Edge Types

Type	Description	Visual
Similarity	Pages with related content	Light edges
Navigation	You clicked from A to B	Bold edges

Similarity Computation

Mode	Method
Default	Jaccard similarity on keywords
With --summarize	Cosine similarity on embeddings

A domain bonus (default: 0.25) is added when tabs share the same domain.

Clustering Algorithm

Uses Union-Find with two strategies:

1. Domain Pre-grouping - Tabs from the same domain grouped together
2. Mutual KNN - Two tabs cluster only if they mutually consider each other neighbors

---

Requirements

Extension

• Chrome/Chromium browser

CLI

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

Optional: LLM Summaries (CLI)

# Install Ollama
brew install ollama

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

# Use with summarization
weft weave --summarize

---

Privacy

Weft is fully local. Your browsing data never leaves your machine:

• Extension uses IndexedDB (browser local storage)
• CLI stores data in local JSON files
• No analytics, no cloud sync, no external requests

License

MIT