This application allows users to ask natural language questions of any PDF document they upload. I've fully developed this solution to combine text generation and analysis capabilities of an LLM with a vector search of the document content. It leverages serverless services such as Amazon Bedrock for accessing foundational models, AWS Lambda to run LangChain, and Amazon DynamoDB for maintaining conversational memory.

Note This project incurs AWS costs. Refer to AWS Pricing for more details.

• Amazon Bedrock for serverless embeddings and inference.
• LangChain to orchestrate a Q&A LLM chain.
• FAISS for vector storage.
• Amazon DynamoDB for serverless conversational memory.
• AWS Lambda for serverless compute.
• Frontend built with React, TypeScript, TailwindCSS, and Vite.
• Local run support and deployment options with Vercel Hosting.
• Amazon Cognito for authentication.

1. A user uploads a PDF document through a static web frontend into an Amazon S3 bucket.
2. The upload triggers metadata extraction and document embedding, converting text to vectors for storage in S3.
3. When a user chats with a PDF document, a Lambda function retrieves relevant vector data and provides an intelligent response using an LLM.

• AWS SAM CLI
• Python 3.11 or later

1. Clone this repository:

   git clone https://github.com/mhrjdv/doc-chat.git

2. Configure Amazon Bedrock model and region parameters in backend/src/generate_response/main.py and backend/src/generate_embeddings/main.py to customize models if desired.

3. Update IAM permissions to allow model access in your preferred region:

   GenerateResponseFunction:
     Type: AWS::Serverless::Function
     Properties:
       Policies:
         - Statement:
             - Sid: "BedrockScopedAccess"
               Effect: "Allow"
               Action: "bedrock:InvokeModel"
               Resource:
                 - "arn:aws:bedrock:*::foundation-model/anthropic.claude-3-haiku"
                 - "arn:aws:bedrock:*::foundation-model/amazon.titan-embed-text-v1"

4. Build and Deploy the Application:

   cd backend
   sam build
   sam deploy --guided

5. Note the output details, which include important URLs and configuration values.

1. Create a .env.development file in the frontend directory with values from your deployment:

   VITE_REGION=us-east-1
   VITE_API_ENDPOINT=https://abcd1234.execute-api.us-east-1.amazonaws.com/dev/
   VITE_USER_POOL_ID=us-east-1_gxKtRocFs
   VITE_USER_POOL_CLIENT_ID=874ghcej99f8iuo0lgdpbrmi76k
   

2. Install frontend dependencies and start the local server:

   npm ci
   npm run dev

3. Access the application locally at http://localhost:5173.

For managed deployment using Vercel Hosting by cloning GitHub Repo.

1. Delete any secrets in AWS Secrets Manager.
2. Empty the Amazon S3 bucket created for this application.
3. Run sam delete from the backend directory to remove associated resources.

While this project demonstrates serverless document chat capabilities, please review security best practices:

• Review encryption options, especially for AWS KMS, S3, and DynamoDB.
• Adjust API Gateway access logging, S3 access logging, and apply specific IAM policies as needed.
• For advanced security needs, consider connecting AWS Lambda to a VPC using the VpcConfig setting.

This project is licensed under the MIT-0 License. See the LICENSE file for details.