RAG Application with LangChain, AWS Opensearch and AWS Bedrock

This repository contains a full RAG application using Terraform as IaC, LangChain as framework, AWS Bedrock as LLM and Embedding Models, AWS OpenSearch as a vector database, and deployment on AWS OpenSearch endpoint.

Main Steps

• Data Ingestion: Load data to an Opensearch Index
• Embedding and Model: Bedrock Titan
• Vector Store and Endpoint: Opensearch
• IaC: Terraform
• data: original pdf document and generated json file with embeddings

Feel free to ⭐ and clone this repo 😉

Tech Stack

Project Structure

The project has been structured with the following files:

• terraform: IaC
• tests: unittest and mock tests
• src: scripts with the app logic
• requirements.txt: project requirements
• Makefile: command for testing, linting and formating
• pyproject.toml: linting/formatting requirements

Project Set Up

The Python version used for this project is Python 3.11.

1. Clone the repo (or download it as a zip file):

   git clone https://github.com/benitomartin/aws-bedrock-opensearch-langchain.git

2. Create the virtual environment named main-env using Conda with Python version 3.10:

   conda create -n main-env python=3.11
   conda activate main-env

3. Install the requirements.txt:

   pip install -r requirements.txt
   
   or
   
   make req

4. Create infrastructure from the terraform folder. This can take up to 30 minutes

   conda install conda-forge::terraform
   terraform init
   terraform plan
   terraform apply

5. Generate embeddings from documents:

   python src/generate_embeddings.py

6. Create Index:

   python src/create_index.py

7. Ingest documents into index:

   python src/ingest_docs_with_embeddings.py

8. Test the app to get a reply:

   python src/app.py

The app contains a question. You can change it accordingly to test other scenarios.