Spaces:
Runtime error
Runtime error
Modular RAG
A hybrid approach to implement RAG inspired by Advance RAG. Usually implemeted with modules acting as plug and play.
Documentation
Generator:
Core component of RAG, responsible for transforming the retrieved information into natural and human sense.
Retriever:
The word "R" in RAG, serving the purpose of retrieving the top K element from knowledge base.
ReRank:
As the name suggest a model used to re-rank the relevant documents. It indexes the documents based on the similariy score between question and the retrieved documents post vector search.
Run Locally
Clone the project
git clone https://github.com/gauravprasadgp/modular-rag
Go to the project directory
cd modular-rag
Install dependencies
pip install -r requirements.txt
Run postgres locally
cd pgvector
docker compose -d up
Start the server
python main.py
API Reference
Upload file to create embedding
POST /create
| Parameter | Type | Description |
|---|---|---|
file |
file |
Required. File to upload |
Get answer from user query
POST /answer
| Parameter | Type | Description |
|---|---|---|
query |
string |
Required. user query |