Agentic RAG with Langchain

🌟 Project Overview

This is the implementation of using Google Gemini 1.5 Pro with langchain as framework. Allowing LLM to give an answer based on the given context (pdf). Also implemented using Langgrapph, a powerful state from Langchain, allowing developer to create custom flow or architecture, deliver with Chat Memory

✨ Features

• 📄 PDF Document Upload
• 🤖 AI-Powered Knowledge Retrieval
• 💬 Interactive Chat Interface
• 🔍 Semantic Document Search

🛠 Tech Stack

• Language Model: Google Gemini 1.5 Pro
• Framework:
  • Streamlit
  • LangChain
• Embedding: Sentence Transformers
• Vector Store: FAISS
• Programming Language: Python 3.8+

🚀 Quick Start

Prerequisites

• Python 3.8+
• Langchain
• Google Generative AI API Key

Installation

1. Clone the repository

git clone https://github.com/bayyy7/agentic_rag-v1.git
cd agentic_rag-v1.git

2. Create a virtual environment

python -m venv venv
source venv/bin/activate  # On Windows, use `venv\Scripts\activate`

3. Install dependencies

pip install -r requirements.txt

4. Configure environment variables

• Create a .env file in the project root
• Add your Google API key:

GOOGLE_GENERATIVE_AI=your_google_api_key_here

5. Create your system prompt

• create prompt folder
• create new python file system_prompt.py

def system_prompt(tool_messages):
   """
   Generate the system prompt content.
   """
   docs_content = "\n\n".join(doc.content for doc in tool_messages)
   return (
      "[YOUR PROMPT HERE]"
      f"{docs_content}\n\n"
)

Running the Application

streamlit run app.py

Custom Config

You can directly change the configuration on the config/config.py. There are several example you can change by what you want. Also be careful when changes the Embedding Model, you must known the dimension of the Embedding it self. This code below is the helper to know the size of embedding dimension.

• Using the embed_query function

vector = embeddings.embed_query("aiueo")
matrix = numpy.array(vector).astype('float32')
len(matrix)

• Using the embed_document function

vector = embeddings.embed_documents(str("aiueo"))
matrix = numpy.array(vector).astype('float32')
matrix.shape[1]