GraphRAG Llama Index MCP Server

# Deployment Handover: GraphRAG Cloud Integration ## Project Current State We have successfully decoupled the Data (S3) from the Application (ECR). The system is split into two parts: 1. Backend (Query Engine): A lightweight Docker image built for search, hosted in AWS ECR. 2. Storage (Database): DuckDB vault hosted in AWS S3, which the backend pulls on startup. --- ## Technical Details for Next Agent ### 1. Backend Infrastructure (AWS ECR) The query engine image is finalized and pushed to: - ECR URI: 575393967752.dkr.ecr.us-west-1.amazonaws.com/graphrag-query:latest - Region: us-west-1 - Startup Behavior: The image uses start_query.sh, which automatically detects S3_BUCKET_NAME and syncs the database from S3 before starting the MCP/API server. ### 2. Database Storage (AWS S3) The Knowledge Graph index is stored in the following vault: - S3 Bucket: investment-research-graphrag-tnhan - Path: backups/index-vault/investment-analysis/ --- ## Deployment Instructions ### Task A: Backend Deployment (App Runner or ECS) Vercel cannot run Docker containers directly. You must deploy the ECR image as an API first. Recommended Host: AWS App Runner (Simplest for ECR images). Required Environment Variables for Backend: | Variable | Value | | :--- | :--- | | S3_BUCKET_NAME | investment-research-graphrag-tnhan | | GRAPHRAG_DATABASE | investment-analysis | | AWS_REGION | us-west-1 | | OPENROUTER_API_KEY | (User to provide secret) | | LLM_MODEL | arcee-ai/trinity-large-preview:free | | RELATIONSHIP_PROVIDER | openrouter | Security Note: The backend service must have an IAM Role with AmazonS3ReadOnlyAccess and permissions for your specific bucket. ### Task B: Frontend Deployment (Vercel) 1. Upload the web application codebase to Vercel. 2. Environment Variables for Vercel: - NEXT_PUBLIC_API_URL: Set this to the URL generated by the AWS Backend deployment. --- ## Key Files to Review - start_query.sh: The glue logic that pulls data from S3 to the container. - mcp_server.ts: The entry point for the query API. - S3_CHEATSHEET.md: Summary of local-to-cloud sync commands. - .env.example: Template for all required variables. Status: Ready for Cloud Deployment. Repo is sanitized and ready for Git commit.