my-rag-system

RAG System with FAISS Vector Database

This repository contains a Retrieval-Augmented Generation (RAG) system with:

• FAISS vector database for efficient similarity search
• Document embeddings using thenlper/gte-small
• Compatible with HuggingFaceH4/zephyr-7b-beta for generation

Usage with API

from huggingface_hub import hf_hub_download
from langchain.vectorstores import FAISS
from langchain_community.embeddings import HuggingFaceEmbeddings
import json
import os

# Download the vector database and config
vector_db_path = hf_hub_download(repo_id="tush1507/my-rag-system", repo_type="model", filename="vector_database/index.faiss")
vector_db_dir = os.path.dirname(vector_db_path)
config_path = hf_hub_download(repo_id="tush1507/my-rag-system", repo_type="model", filename="config/embedding_config.json")

# Load the embedding configuration
with open(config_path, 'r') as f:
    embedding_config = json.load(f)

# Create the embedding model
embedding_model = HuggingFaceEmbeddings(
    model_name=embedding_config["model_name"],
    encode_kwargs={"normalize_embeddings": embedding_config.get("normalize_embeddings", True)}
)

# Load the vector database
knowledge_base = FAISS.load_local(vector_db_dir, embedding_model)

# Search for similar documents
query = "Your question here"
docs = knowledge_base.similarity_search(query, k=5)

# Now process with your preferred LLM
# [Your LLM code here]