Commit to Career

Paste a company name and role — get a researched, personalized recruiter outreach message in seconds.

What is this?

Commit to Career is an AI-powered job outreach tool. You give it a company name and a target role; it researches the company, finds relevant people to contact, and drafts a personalized cold outreach message — all automatically.

The workflow:

1. You enter a company name and job role in the UI
2. The backend runs a multi-agent pipeline (research → find people → draft message)
3. The UI polls for live progress and shows results as each stage completes
4. You review and send the drafted outreach message

Key features:

• Company research using live web search (Tavily)
• LinkedIn people finder to surface relevant contacts
• LLM-drafted personalized outreach (Groq)
• Resume-aware — the draft references your background
• Real-time progress updates via polling

Architecture

User Input (company + role)
        │
        ▼
  ┌─────────────┐        ┌──────────────────────────────────────────┐
  │  Next.js UI │──────▶ │              FastAPI Backend              │
  │  (port 3000)│◀────── │                                          │
  └─────────────┘  poll  │  load_resume → research → find_people   │
                         │                          → draft_message │
                         │                                          │
                         │  Tavily (web search)  Groq (LLM)        │
                         └──────────────────────────────────────────┘

The frontend polls /rows every 3 seconds to reflect live agent progress. Each stage updates the shared RecruitState as it completes.

Agent pipeline (backend/app/agents/):

• researcher.py — Dual Tavily search + Groq LLM → company overview and external links
• people.py — LinkedIn search via Tavily, LLM extracts structured employee list
• drafter.py — Groq LLM drafts personalized message using top contact and your resume

State flows through RecruitState (TypedDict in models/state.py). Status transitions: researching → finding_people → drafting → awaiting_review (or error).

Database

Postgres 17 — local, no Docker required.

jobs       thread_id (PK) · company · role · status [indexed] · tags[] [GIN indexed] · state_json (JSONB) · updated_at
profiles   id=1 (singleton) · profile_json (JSONB) · updated_at

status and tags are real columns with indexes for filtering. Everything else lives in state_json — the full JobState serialised by Pydantic. Store interface in backend/app/store/ is abstracted so the backing DB is a one-file swap.

Local Setup

Prerequisites

• Node.js 18+ and npm
• Python 3.11+
• PostgreSQL 17 running locally
• API keys for Tavily and Groq

1. Clone the repo

git clone https://github.com/neversettle17-101/commit-to-career.git
cd commit-to-career

2. Set up environment variables

Create backend/.env:

TAVILY_API_KEY=your_tavily_key_here
GROQ_API_KEY=your_groq_key_here
DATABASE_URL=postgresql://localhost/commit_to_career

3. Set up the database

createdb commit_to_career

4. Set up the Python backend

cd backend
python -m venv venv
source venv/bin/activate        # Windows: venv\Scripts\activate
pip install -r app/requirements.txt

5. Set up the frontend

cd ui
npm install

6. Run the app

Open two terminals from the project root:

# Terminal 1 — backend (port 8000)
./start.sh backend

# Terminal 2 — frontend (port 3000)
./start.sh frontend

Then open http://localhost:3000.

Running tests

source backend/venv/bin/activate
pytest backend/app/tests/

Tech Stack

Layer	Technology
Frontend	Next.js 16, React 19, Tailwind CSS v4, shadcn/ui
Backend	FastAPI, Python 3.11+
AI / Search	Groq LLM, Tavily web search
Database	PostgreSQL 17
Agents	Custom multi-agent pipeline