https://github.com/nexckycort/rag-poc
https://github.com/nexckycort/rag-poc
fastapi markdown pinecone python rag typescript vector-database
Last synced: 3 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/nexckycort/rag-poc
- Owner: nexckycort
- Created: 2025-04-03T23:37:42.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2025-04-04T00:03:17.000Z (about 1 year ago)
- Last Synced: 2025-04-09T18:05:48.358Z (about 1 year ago)
- Topics: fastapi, markdown, pinecone, python, rag, typescript, vector-database
- Language: TypeScript
- Homepage:
- Size: 17.6 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# RAG Proof of Concept
This project is a Proof of Concept (PoC) for a RAG (Retrieval-Augmented Generation) system that combines a vector database with a language model to answer questions based on stored information.
## Technologies Used
- **FastAPI** for the embedding server
- **SentenceTransformers** with `all-MiniLM-L6-v2` for generating embeddings
- **Pinecone** as the vector database
- **TypeScript** for query processing
## Installation and Setup
### 1. Clone the Repository
```sh
git clone https://github.com/nexckycort/rag-poc.git
cd rag-poc
```
### 2. Install Monorepo Dependencies
```sh
bun install
```
### 3. Set Up the Embedding Server
```sh
cd packages/model-server
bun create:venv # Create virtual environment
bun activate:venv # Activate virtual environment
bun pip:install # Install dependencies
bun dev # Start server
```
### 4. Run Query Logic
```sh
cd packages/api-server
bun dev # Run query logic
```
## Usage
1. **Add documents**: Files are processed and stored as embeddings in Pinecone.
2. **Ask a question**: An embedding is generated from the user’s query, and relevant text is retrieved.
3. **Generate answer**: A language model generates a response based on the retrieved context.
## API Endpoints
### Embedding Server (FastAPI)
- `POST /embed` → Generates text embeddings
- `POST /ask` → Queries the model
## License
This project is licensed under the MIT License.