GraphQL CRUD + LLM Query Generator

Overview

This project demonstrates a GraphQL CRUD API backend by MongoDB and a lightweight LLM service that translates natural language requests into executable GraphQL operations. A small React frontend sends a user prompt to the LLM service, which returns a GraphQL query and variables; the LLM service then executes the query against the GraphQL API and returns formatted results to the UI.

Stack

• Backend API: Apollo Server (Node.js)
• Database: MongoDB via Mongoose
• LLM Service: Express + Axios
• Frontend: React (Vite)

Repository Layout

• backend/
  • index.js: Starts GraphQL server (Apollo)
  • llm.js: LLM bridge server (POST /chat)
• frontend/llm-frontend/: React UI (Vite)

Quick Start

1. Install dependencies

cd backend && npm install
cd ../frontend/llm-frontend && npm install

2. Start services (3 terminals)

# Terminal 1: GraphQL API and LLM (port 4000)
cd backend
npm run start:all

# Terminal 2: Frontend (port 5173)
cd frontend/llm-frontend
npm run dev

Configuration

• LLM local endpoint used by backend/llm.js (default): http://localhost:12434/engines/llama.cpp/v1/chat/completions
• GraphQL API: http://localhost:4000/graphql
• LLM bridge: http://localhost:3001/chat

You can change ports/hosts in backend/llm.js and frontend/llm-frontend/src/App.tsx if needed.

Notes

• The API expects Mongo-style filters as a string. If you prefer typed filter inputs, extend the schema with input types and adjust resolvers accordingly.
• The UI is optimized for reading queries, variables, and results with clear sectioning and styling.