TalentNode

AI-powered job search assistant that combines LLM reasoning with MCP (Model Context Protocol) tool integration to find and display job listings from multiple platforms.

Tech Stack

Layer	Technology
Backend	Node.js, Express, TypeScript, Prisma + PostgreSQL
Frontend	React, Vite, TypeScript
LLM	OpenRouter (free models) with streaming responses
MCP	Custom HTTP API servers (Remotive, Adzuna)
Auth	JWT (bcrypt hashing)
Logging	Pino (structured JSON logs)

Project Structure

TalentNode/
├── backend/                # Express API server
│   ├── src/
│   │   ├── index.ts           # App entry, middleware registration
│   │   ├── openrouter.ts       # OpenRouter client (chat + stream)
│   │   ├── mcp.ts              # HTTP MCP clients (Remotive, Adzuna)
│   │   ├── middleware/
│   │   │   ├── auth.ts          # JWT auth middleware
│   │   │   ├── optionalAuth.ts  # Optional auth for chat
│   │   │   └── requestId.ts     # UUID request tracing
│   │   ├── routes/
│   │   │   ├── chat.ts         # Streaming chat + MCP orchestration
│   │   │   ├── auth.ts         # Register/Login (JWT)
│   │   │   ├── saved-jobs.ts   # Save/unsave jobs
│   │   │   └── job-interactions.ts # Favorite/like/dislike/visited
│   │   └── utils/
│   │       └── logger.ts          # Pino structured logging
│   ├── prisma/
│   │   ├── schema.prisma     # User, SavedJob, SearchHistory, SearchResult, JobInteraction
│   │   └── .env              # Env vars (DATABASE_URL, OPENROUTER_API_KEY, JWT_SECRET)
│   └── package.json
│
├── frontend/               # React + Vite SPA
│   ├── src/
│   │   ├── App.tsx            # Main app (auth state, chat flow)
│   │   ├── api.ts             # API client (chat, auth, interactions)
│   │   ├── components/
│   │   │   ├── Login.tsx         # Login/Register form
│   │   │   ├── ChatInput.tsx     # Message input
│   │   │   ├── MessageList.tsx   # Chat messages + job cards
│   │   │   ├── JobCard.tsx       # Job card with platform badge + interaction buttons
│   │   │   └── SkeletonLoader.tsx # Glassmorphism loading state
│   │   └── App.css            # Styles (glassmorphism, badges, auth UI)
│   ├── .env                  # VITE_API_URL
│   └── package.json
│
├── mcp-servers/
│   ├── remotive/             # Remotive MCP (HTTP API)
│   │   └── src/index.ts     # Express server exposing /tools/call
│   └── adzuna/              # Adzuna MCP (HTTP API)
│       ├── .env             # ADZUNA_APP_ID, ADZUNA_APP_KEY
│       └── src/index.ts     # Express server exposing /tools/call
│
├── .gitignore              # Excludes node_modules, .env, dist, PLAN.md, etc.
├── PLAN.md                # Detailed phase-by-phase build log
└── README.md              # This file

Features

Phase 1-6: Alpha Core ✅

Project skeleton (Git, Express, React/Vite, TypeScript strict)
PostgreSQL + Prisma (User, SavedJob, SearchHistory, SearchResult, JobInteraction)
OpenRouter API client (direct fetch, streaming SSE)
Chat route with LLM + MCP tool orchestration
Glassmorphism UI with skeleton loaders
Error handling (rate limits, empty results, MCP crashes)
Streaming LLM responses (tokens stream in real-time)
Start script (start-alpha.bat)

Phase 7: Multi-Platform Job Search ✅

Remotive MCP server — free remote jobs API (no key, HTTP transport)
Adzuna MCP server — free India jobs API (needs free API key, HTTP transport)
Backend queries both platforms simultaneously via HTTP
Frontend shows platform badge (green=Remotive, yellow=Adzuna)
No duplicate rendering (LLM gives 1-2 sentence response, jobs render as cards)

Phase 8: Authentication + Data Optimization ✅

JWT Auth — Register/Login routes with bcrypt hashing
Auth middleware — protected routes (saved-jobs) + optional auth (chat)
Frontend auth — Login/Register UI, JWT in localStorage
JobInteraction table — favorite, liked, disliked, visited flags
SearchResult child table — normalized job data (replaced JSON blob)
JobCard interaction buttons — toggle favorite/like/dislike/visited
Structured logging — Pino with requestId tracing, timing metrics

Database Schema

model User {
  id        String   @id @default(cuid())
  email     String   @unique
  password  String   // bcrypt hash
  name      String?
  createdAt DateTime @default(now())
  updatedAt DateTime @updatedAt
  savedJobs       SavedJob[]
  searchHistory   SearchHistory[]
  jobInteractions JobInteraction[]
}

model SavedJob {
  id          String   @id @default(cuid())
  userId      String
  jobData     Json
  source      String   // "remotive" | "adzuna"
  externalId  String
  title       String
  company     String
  location    String?
  savedAt     DateTime @default(now())
  user         User @relation(fields: [userId], references: [id], onDelete: Cascade)
  jobInteractions JobInteraction[]
  @@unique([userId, externalId, source])
}

model SearchHistory {
  id        String   @id @default(cuid())
  userId    String?
  query     String
  filters   Json?
  createdAt DateTime @default(now())
  user         User?        @relation(fields: [userId], references: [id], onDelete: SetNull)
  searchResults SearchResult[]
}

model SearchResult {
  id          String   @id @default(cuid())
  searchId    String
  jobId       String
  source      String
  title       String
  company     String
  location    String?
  url         String?
  postedAt    String?
  searchHistory SearchHistory @relation(fields: [searchId], references: [id], onDelete: Cascade)
}

model JobInteraction {
  id          String   @id @default(cuid())
  userId      String
  jobId       String
  source      String
  type        String   // "favorite" | "liked" | "disliked" | "visited"
  createdAt   DateTime @default(now())
  user         User @relation(fields: [userId], references: [id], onDelete: Cascade)
  @@unique([userId, jobId, source, type])
}

Setup

Prerequisites

Node.js 18+
PostgreSQL running locally
OpenRouter API key (free tier works)
Adzuna API key (free at developer.adzuna.com)

Backend

cd backend
npm install
cp .env.template .env   # then add: DATABASE_URL, OPENROUTER_API_KEY, JWT_SECRET
npx prisma db push
npm run dev          # starts at http://localhost:5000

Required backend/.env:

PORT=5000
DATABASE_URL="postgresql://postgres:password@localhost:5432/talentnode"
OPENROUTER_API_KEY="sk-or-v1-..."
OPENROUTER_BASE_URL="https://openrouter.ai/api/v1"
JWT_SECRET="your-random-secret-string"
REMOTIVE_URL="http://localhost:3002"   # local MCP
ADZUNA_URL="http://localhost:3001"      # local MCP

Frontend

cd frontend
npm install
cp .env.template .env   # then set VITE_API_URL
npm run dev                 # starts at http://localhost:5173

Required frontend/.env:

VITE_API_URL=http://localhost:5000

MCP Servers (for local dev)

# Remotive (port 3002)
cd mcp-servers/remotive && npm install && npm run build && node dist/index.js

# Adzuna (port 3001)
cd mcp-servers/adzuna
cp .env.template .env   # add ADZUNA_APP_ID, ADZUNA_APP_KEY
npm install && npm run build && node dist/index.js

Get Adzuna credentials:

Sign up at https://developer.adzuna.com/
Your App ID: 40790798
Your API Key: 8929ffaf073136b703120eb59055b5c2

API Endpoints

Auth

Method	Endpoint	Description	Auth
POST	`/api/auth/register`	Register new user	None
POST	`/api/auth/login`	Login, returns JWT	None

Chat (streaming SSE)

Method	Endpoint	Description	Auth
POST	`/api/chat`	Streaming chat + job search	Optional

Request:

{
  "messages": [
    { "role": "user", "content": "find me react developer jobs" }
  ]
}

Response (SSE stream):

data: {"token": "I found"}
data: {"token": " several jobs"}
...
data: {"done": true, "jobs": [...]}

Saved Jobs

Method	Endpoint	Description	Auth
GET	`/api/saved-jobs`	List saved jobs	Required
POST	`/api/saved-jobs`	Save a job	Required
DELETE	`/api/saved-jobs/:id`	Unsave a job	Required

Job Interactions

Method	Endpoint	Description	Auth
GET	`/api/job-interactions?jobId=...`	Get interactions for a job	Required
POST	`/api/job-interactions`	Toggle favorite/like/disliked/visited	Required

Request Flow

User → Frontend (React)
         ↓
     POST /api/chat (with JWT if logged in)
         ↓
     Backend (Express + optionalAuth)
         ↓
     ┌─────────────┐
     │ OpenRouter LLM │ ← streaming SSE
     └─────────────┘
         ↓ (if job query)
     ┌────────────────┐  ┌────────────────┐
     │ Remotive MCP  │  │  Adzuna MCP  │ (HTTP calls)
     └────────────────┘  └────────────────┘
         ↓
     Merge results + normalize
         ↓
     Save SearchHistory + SearchResult (if authenticated)
         ↓
     Stream response to frontend (tokens + jobs payload)
         ↓
     Frontend renders: chat message + JobCards with badges + interaction buttons

Logging

Structured JSON logs via Pino:

Every request gets a requestId (UUID prefix)
Log levels: debug (dev), info (prod)
Timing metrics: LLM latency, MCP latency, total request time
Stages logged: CHAT_ROUTE, LLM_REQUEST, LLM_RESPONSE, LLM_STREAM, MCP_REMOTIVE, MCP_ADZUNA

Example log:

{
  "level": 30,
  "requestId": "a1b2c3",
  "stage": "LLM_RESPONSE",
  "model": "meta-llama/llama-3.1-8b-instruct:free",
  "duration": 2340,
  "hasContent": true,
  "toolCalls": ["search_remotive_jobs", "search_adzuna_jobs"],
  "timestamp": "2026-05-02T..."
}

Deployment (Render)

1. Push to GitHub

cd "C:\Users\krama\Downloads\Cohort-projects\TalentNode"
git add .
git commit -m "feat: complete alpha - auth, multi-platform, logging"
git push origin main

2. Render Services

Service	Type	Root Directory	Build	Start
talentnode-db	PostgreSQL	—	—	—
talentnode-backend	Web Service	`backend`	`npm install && npm run build`	`node dist/index.js`
talentnode-frontend	Static Site	`frontend`	`npm install && npm run build`	— (publish: `dist`)
talentnode-remotive	Web Service	`mcp-servers/remotive`	`npm install && npm run build`	`node dist/index.js`
talentnode-adzuna	Web Service	`mcp-servers/adzuna`	`npm install && npm run build`	`node dist/index.js`

3. Environment Variables

Backend: DATABASE_URL (from DB), OPENROUTER_API_KEY, JWT_SECRET, REMOTIVE_URL, ADZUNA_URL
Frontend: VITE_API_URL=https://talentnode-backend.onrender.com

Post-Alpha Roadmap

Real LinkedIn job data (API or scraping)
User profile + resume upload
Advanced search filters (salary, experience, industry)
Job application tracking (applied, interviewing, rejected)
Email alerts for new matching jobs
Mobile responsive design
Render production deployment (all services live)

License

MIT

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

TalentNode

Tech Stack

Project Structure

Features

Phase 1-6: Alpha Core ✅

Phase 7: Multi-Platform Job Search ✅

Phase 8: Authentication + Data Optimization ✅

Database Schema

Setup

Prerequisites

Backend

Frontend

MCP Servers (for local dev)

API Endpoints

Auth

Chat (streaming SSE)

Saved Jobs

Job Interactions

Request Flow

Logging

Deployment (Render)

1. Push to GitHub

2. Render Services

3. Environment Variables

Post-Alpha Roadmap

License

About

Uh oh!

Releases

Packages

Uh oh!

Contributors

Uh oh!

Languages

Name		Name	Last commit message	Last commit date
Latest commit History 16 Commits
backend		backend
frontend		frontend
mcp-servers		mcp-servers
.gitignore		.gitignore
README.md		README.md

Folders and files

Latest commit

History

Repository files navigation

TalentNode

Tech Stack

Project Structure

Features

Phase 1-6: Alpha Core ✅

Phase 7: Multi-Platform Job Search ✅

Phase 8: Authentication + Data Optimization ✅

Database Schema

Setup

Prerequisites

Backend

Frontend

MCP Servers (for local dev)

API Endpoints

Auth

Chat (streaming SSE)

Saved Jobs

Job Interactions

Request Flow

Logging

Deployment (Render)

1. Push to GitHub

2. Render Services

3. Environment Variables

Post-Alpha Roadmap

License

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages