https://github.com/talperetz/browsegraph
In-browser GraphRAG
https://github.com/talperetz/browsegraph
chrome-extension graph graphrag llms rag
Last synced: 5 months ago
JSON representation
In-browser GraphRAG
- Host: GitHub
- URL: https://github.com/talperetz/browsegraph
- Owner: talperetz
- License: mit
- Created: 2024-11-15T04:41:19.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2025-12-26T13:01:50.000Z (6 months ago)
- Last Synced: 2025-12-28T01:02:03.622Z (6 months ago)
- Topics: chrome-extension, graph, graphrag, llms, rag
- Language: TypeScript
- Homepage: https://www.browsegraph.com/
- Size: 1.08 MB
- Stars: 78
- Watchers: 1
- Forks: 17
- Open Issues: 8
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
Introducing the first in-browser GraphRAG.
Everything you browse, fully connected and always accessible.
## ⚡ Getting Started
### Soon on Chrome Web Store
BrowseGraph will be available on the Chrome Web Store soon with 1-click installation. Stay tuned!
### Install Manually
#### Clone the Repository
```bash
git clone https://github.com/talperetz/browsegraph.git
```
#### Navigate to the Directory
```bash
cd browsegraph
```
#### Add a .env file to set your Key
```txt
VITE_GOOGLE_GENERATIVE_AI_API_KEY="***"
```
#### Install Dependencies
```bash
pnpm install
```
#### Build the Extension
```bash
pnpm build
```
#### Navigate to the Directory
```bash
cd browsegraph
```
#### Load the Extension
[Follow the official turorial](https://knowledge.workspace.google.com/kb/load-unpacked-extensions-000005962)
## 🔥 Key Features
- **Local-First Processing**: All AI computations are performed locally within your browser using in-browser LLMs.
- **Dynamic Knowledge Graphing**: Automatically capture and organize everything you browse into a visual knowledge network.
- **Smart Content Filtering**: Filters out irrelevant pages based on your interests.
- **Efficient Summarization**: Summarizes and indexes page content for quick access.
- **Blazingly Fast Retrieval**: Retrieve any piece of your browsed content instantly with powerful search and intuitive navigation.
- **Side-by-side Knowledge Graph**: Visualize the knowledge in the current page:
## 🧩 How It Works
```mermaid
sequenceDiagram
%% Participants
participant Extension as Chrome Extension
participant Personalizer as User Content Personalizer
(Local LLM)
participant Classifier as Page Classifier
(Local LLM)
participant Summarizer as Page Summarizer
(Local LLM)
participant CloudLLM as Page-Graph-Transformer
(Cloud LLM)
participant Database as Local Database
%% User action without a participant node
Note over Extension: User browses a web page
%% Sequence of interactions
Extension ->> Personalizer: Generate User Interests
Personalizer -->> Extension: User Interests
Extension ->> Classifier: Classify Page vs. User Interests
Classifier -->> Extension: Relevance Result (Relevant)
Extension ->> Summarizer: Chunk and Summarize Content
Summarizer -->> Extension: Summarized Data
Extension ->> Database: Index Summarized Data
Extension ->> CloudLLM: Send Minimal Data
CloudLLM -->> Extension: Return Graph Data
Extension ->> Database: Store Graph Data
```
## 🛠 Built With
BrowseGraph harnesses cutting-edge technologies to deliver a fast, precise, and private experience:
- **Local Vector Database**: [pglite](https://github.com/electric-sql/pglite/) + [pgvector](https://github.com/pgvector/pgvector) for in-browser storage and blazing-fast vector search.
- **Local LLMs**: Gemini Nano + [Chrome's built-in AI](https://developer.chrome.com/docs/ai/built-in) for private content classification, summarization, and personalization.
- **Graph Rendering**: [ReactFlow](https://github.com/xyflow/xyflow) powers dynamic, interactive knowledge graphs.
- **Search Interface**: [cmdk](https://github.com/pacocoursey/cmdk) provides a lightning-fast, intuitive search bar.
## 🔒 Privacy & Security
- **Data Locality:** All raw data processing occurs within your browser using local LLMs and local vector DB.
- **Minimal Cloud Interaction:** Only non-sensitive, aggregated data is sent to the cloud for graph transformations.
- **No Data Sharing:** Your browsing data is never shared with third parties.
## ⏩ Roadmap
- [ ] **Conversational Interface:** Build an interactive conversational layer to simplify user interactions.
- [ ] **Artifacts UX:** Enable a side-by-side view of chat and graph for seamless exploration.
- [ ] **Bring Your Own Key (BYOK):** Support user-provided keys for LLM providers to offer flexibility and control.
- [ ] **Chrome Web Store Launch:** Officially release BrowseGraph on the Chrome Web Store.
## 🤗 Contributions
BrowseGraph is open-source and we welcome contributions. Please fork the repository, create a new branch for your feature, add your feature or improvement, and send a pull request.