Skip to content

README.md

Latest commit

 

History

History
151 lines (109 loc) · 4.12 KB

README.md

File metadata and controls

151 lines (109 loc) · 4.12 KB

ContextCV

A full-stack “ContextCV” platform where users author their portfolio in Markdown, publish a public profile, and visitors can chat with an AI that answers strictly from the portfolio content.

Features

  • User registration and login (JWT)
  • Markdown portfolio editor with live preview
  • Public profile page rendering Markdown
  • Real-time chat with owner’s AI (answers only from portfolio context)
  • Swagger API docs
  • Pinecone vector storage for portfolio context
  • CORS configured for local dev (React 3000 → API 5000)

Tech Stack

  • Backend: Node.js, Express, Socket.io, Mongoose, JWT, Swagger, Google Generative AI, Pinecone
  • Frontend: React, Axios, socket.io-client, react-markdown, lucide-react

Backend Setup

  1. Install dependencies:
cd backend-node
npm install
  1. Environment variables: create .env in backend-node:
PORT=5000
MONGODB_URI=mongodb://localhost:27017/livingcv
JWT_SECRET=replace_with_a_strong_secret

# Google Gemini
GEMINI_API_KEY=your_gemini_api_key

# Pinecone
PINECONE_API_KEY=your_pinecone_api_key
# IMPORTANT: keep names consistent; set BOTH to the same index name
PINECONE_INDEX=my-livingcv-index
PINECONE_INDEX_NAME=my-livingcv-index
  1. Start backend:
npm start

API Overview

  • Auth
    • POST /api/auth/register
    • POST /api/auth/login → returns { token }
  • Portfolio (private)
    • GET /api/portfolio → requires Authorization: Bearer <token>
    • PUT /api/portfolio → save markdown (syncs to Pinecone), requires Bearer
  • Portfolio (public)
    • GET /api/portfolio/:userId → public view of portfolio markdown
  • Socket.io
    • Client emits: chat_with_bot { targetUserId, question }
    • Server emits: bot_stream chunks of AI answer

RAG and AI

  • The user’s portfolio markdown is embedded (Pinecone Llama v2) and stored as a single vector per user with metadata { text: <markdown> }.
  • Chat fetches the full text directly via index.fetch([userId]) to ensure accurate context.
  • Prompt enforces strict persona: first-person answers, use ONLY provided context; if missing, respond: “I haven’t included that in my portfolio yet.”

Frontend Setup

  1. Install dependencies:
cd frontend
npm install
  1. Start frontend:
npm start

Usage Flow

  • Register → Login → Edit your Markdown portfolio → Save → Click “View Public”
  • Share the public URL /profile/<userId> with anyone
  • Visitors ask questions in the chat; AI answers from your portfolio only

Sample Requests

  • Register
curl -X POST http://localhost:5000/api/auth/register ^
  -H "Content-Type: application/json" ^
  -d "{\"email\":\"dev@example.com\",\"password\":\"SuperSecret!123\"}"
  • Login
curl -X POST http://localhost:5000/api/auth/login ^
  -H "Content-Type: application/json" ^
  -d "{\"email\":\"dev@example.com\",\"password\":\"SuperSecret!123\"}"
  • Save portfolio (replace TOKEN)
curl -X PUT http://localhost:5000/api/portfolio ^
  -H "Authorization: Bearer TOKEN" ^
  -H "Content-Type: application/json" ^
  -d "{\"content\":\"# Jane Doe\\n\\n## Skills\\n- Node.js\\n- React\"}"
  • Get public portfolio
curl http://localhost:5000/api/portfolio/<userId>

Environment Notes

  • Pinecone index should match embedding dimension: 768 (llama-text-embed-v2)
  • Ensure both PINECONE_INDEX and PINECONE_INDEX_NAME reference the same index name (the code reads both in different places)
  • CORS is enabled in server.js for development

Troubleshooting

  • “No chat response”
    • Check backend logs (console) for socket events and Gemini errors
    • Confirm GEMINI_API_KEY and PINECONE_API_KEY are valid
    • Ensure portfolio has been saved at least once (so Pinecone has content)
  • “Portfolio not found”
    • The user hasn’t saved yet; save a portfolio in Dashboard first
  • Swagger shows no routes
    • Verify swagger-ui-express and swagger-jsdoc are installed
    • Check apis: ['./routes/*.js'] and JSDoc comments exist

License

MIT