Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/cyse7125-su24-team10/pinecone-loader
Builds Docker Image for Loading Pinecone Database with Vectors from CVE data.
https://github.com/cyse7125-su24-team10/pinecone-loader
huggingface-transformers langchain llama3 pinecone
Last synced: 3 months ago
JSON representation
Builds Docker Image for Loading Pinecone Database with Vectors from CVE data.
- Host: GitHub
- URL: https://github.com/cyse7125-su24-team10/pinecone-loader
- Owner: cyse7125-su24-team10
- Created: 2024-08-14T03:13:34.000Z (5 months ago)
- Default Branch: main
- Last Pushed: 2024-09-11T18:37:12.000Z (4 months ago)
- Last Synced: 2024-10-06T20:54:33.127Z (3 months ago)
- Topics: huggingface-transformers, langchain, llama3, pinecone
- Language: Python
- Homepage:
- Size: 9.77 KB
- Stars: 0
- Watchers: 0
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Changelog: CHANGELOG.md
Awesome Lists containing this project
README
# pinecone-loader
This project downloads CVE records from a [CVE](https://github.com/CVEProject/cvelistV5/releases) URL, processes them, and stores their embeddings in a Pinecone vector database using HuggingFace's `BAAI/bge-small-en-v1.5` model. The workflow handles downloading, unzipping, processing, embedding, and cleaning up after the data has been processed.
## Requirements
- **Languages**: Python 3.9
- **Dependencies**:
- streamlit
- python-dotenv
- langchain
- langchain_community
- faiss-cpu
- langchain-groq
- pinecone-client
- langchain-pinecone
- HuggingFace embeddings model `BAAI/bge-small-en-v1.5`
- Pinecone vector store
## Workflow
1. **Downloading and Unzipping**:
- Downloads a zip file containing JSON files with CVE data, unzips the contents, and processes the JSON files.
2. **Moving Files**:
- Moves JSON files into a `data` folder for further processing.
3. **Processing CVE Data**:
- Parses JSON files, skips "REJECTED" CVEs, and processes the remaining ones to extract descriptions and metadata.
4. **Embedding with HuggingFace**:
- Uses HuggingFace’s `bge-small-en-v1.5` model to generate embeddings for storage in Pinecone.
5. **Pinecone Integration**:
- Stores embeddings in Pinecone using `langchain.vectorstores.Pinecone`.
6. **Cleanup**:
- Removes extracted files and the original zip file after processing.
## Setup Instructions
1. **Environment Setup**:
- Ensure you have Python 3.9 installed.
- Create and activate a virtual environment:
```bash
python3 -m venv venv
source venv/bin/activate
```
- Install the required dependencies:
```bash
pip install -r requirements.txt
pip install torch --index-url https://download.pytorch.org/whl/cpu
pip install sentence-transformers
```
2. **Environment Variables**:
- Create a `.env` file in the root directory with the following variables:
```bash
PINECONE_API_KEY=your_pinecone_api_key
PINECONE_CLOUD=aws
PINECONE_REGION=us-east-1
URL=https://your_download_link.com/deltaCves.zip
```
3. **Running the Application**:
- To execute the script, run:
```bash
python pinecone_db.py
```
## Docker Instructions
1. **Docker Build**:
- Build the Docker image using the following command:
```bash
docker build -t cve-processor:latest .
```
2. **Running the Docker Container**:
- Run the container with the necessary environment variables:
```bash
docker run --env-file .env cve-processor:latest
```
## Key Components
### `pinecone_db.py`
This script handles:
- Downloading and unzipping CVE data.
- Moving JSON files to the `data` directory.
- Processing and embedding CVE data using HuggingFace models.
- Storing embeddings in Pinecone and cleaning up resources.
### `Dockerfile`
The `Dockerfile` sets up the environment, installs dependencies, and defines the entry point to run the `pinecone_db.py` script inside a containerized environment.
## License
This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.