JobSearchAgent

Personal AI job search and opportunity tracker built around LangChain, LangGraph, LangSmith, FastAPI, Postgres, and a custom Agent Chat UI workspace.

The app lets you chat with an agent while keeping a structured opportunities table as the source of truth for roles, links, fit scores, application status, interview state, and notes.

Features

• Chat-based job intake from pasted links, job descriptions, CSVs, and tables.
• Opportunity table with search, score/status/applied filters, sorting, modal details, and quick status updates.
• 25-point fit scoring rubric with sub-scores stored in metadata.
• Resume/profile context tools so the agent can score roles against your background.
• URL extraction and optional Tavily web search for job/company research.
• Local Docker Compose stack for Postgres, migrations, FastAPI, LangGraph, and the UI.

Stack

• Backend: FastAPI, SQLModel, Alembic, PostgreSQL
• Agent: LangChain, Deep Agents, LangGraph, LangSmith, OpenAI, Tavily
• UI: Agent Chat UI, Next.js, Tailwind CSS, TanStack Query, TanStack Table

Local Setup

Start the full local app:

cp backend/.env.example backend/.env
docker compose up

Open http://localhost:13000.

If that port is already in use, run for example:

UI_PORT=13001 docker compose up

Or start services manually:

docker compose up -d postgres
cd backend
uv sync --extra dev
cp .env.example .env
uv run alembic upgrade head
uv run api

Start the agent server:

uv run langgraph dev --port 12024

Start the UI:

cd ui
nvm use
corepack enable
cp .env.example .env.local
pnpm dev --filter=web

More details: Development guide.

Project Docs

• Architecture plan
• Development guide
• Folder structure plan
• User stories
• Project checklist