An AI agent that helps users:

1. enter their job history, education, skills, and contact details
2. paste a job description
3. get an explainable qualification / match score + gap analysis
4. generate a tailored resume + cover letter for that job
5. download tailored outputs as PDF / DOCX, and generate LaTeX (Overleaf) code for easy editing

LLM Provider: OpenAI
Backend: FastAPI
Agent Workflow: LangChain + LangGraph

• User enters: job history, education, skills, summary, projects, contact details
• (Dev option) Use a local PDF resume → extract text → produce structured JSON

• Parse job description into structured requirements
• Compute match score with breakdown (stable and explainable)
• Identify gaps + missing evidence

• Tailored resume (ATS-friendly)
• Tailored cover letter (role-specific)
• Guardrails: do not invent experience

• Export tailored resume / cover letter as PDF and/or DOCX
• Generate LaTeX compatible with Overleaf

• FastAPI + Uvicorn
• LangChain (LLM calls, tools)
• LangGraph (workflow orchestration)
• Pydantic (strict schemas)
• PDF text extraction: pypdf
• DOCX: python-docx
• PDF: reportlab (or docx→pdf conversion later)
• Database (recommended): Postgres
• Vector DB (recommended): pgvector (inside Postgres)

.
├── backend/
│   └── app/
│       ├── main.py
│       ├── config.py
│       ├── models/
│       │   ├── __init__.py
│       │   ├── schemas_resume.py
│       │   ├── schemas_job.py
│       │   └── schemas_outputs.py
│       ├── resume/
│       │   ├── __init__.py
│       │   ├── pdf_extract.py
│       │   ├── structurer.py
│       │   └── latex_generator.py
│       ├── match/
│       │   ├── __init__.py
│       │   ├── scorer.py
│       │   └── gap_insights.py
│       ├── tailor/
│       │   ├── __init__.py
│       │   ├── resume_tailor.py
│       │   └── cover_letter_tailor.py
│       ├── export/
│       │   ├── __init__.py
│       │   ├── docx_export.py
│       │   └── pdf_export.py
│       └── agent/
│           ├── __init__.py
│           └── graph.py
├── data/
│   └── resumes/
├── .env
└── requirements.txt

Important: Ensure each package folder contains an __init__.py so Python imports work.

python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip

pip install fastapi "uvicorn[standard]" python-dotenv pydantic \
  langchain langgraph openai langchain-openai \
  pypdf python-docx reportlab beanie \
  sqlalchemy "psycopg[binary]" pgvector alembic

Create a .env file in the project root:

OPENAI_API_KEY=your_key_here
OPENAI_CHAT_MODEL=gpt-4.1-mini

# Local dev folder for PDFs (only reads PDFs from here)
BASE_RESUME_DIR=./data/resumes

# MongoDB (resume-ai database)
MONGODB_URI=mongodb://admin:your_password@localhost:27017
MONGODB_DB_NAME=resume-ai

# Optional DB (later)
DATABASE_URL=postgresql+psycopg://user:pass@localhost:5432/resume_tailor

Verify the key is visible to Python:

python -c "import os; print('OPENAI_API_KEY set?', bool(os.getenv('OPENAI_API_KEY')))"

uvicorn backend.app.main:app --reload --port 8000

curl http://127.0.0.1:8000/health

1) Put your resume PDF in data/resumes/:

mkdir -p data/resumes
cp /path/to/your_resume.pdf data/resumes/resume.pdf

2) Call the endpoint:

curl "http://127.0.0.1:8000/resume/format?pdf_path=resume.pdf"

The response typically includes:

• source_pdf
• extracted_chars
• resume (structured JSON)

Note: This endpoint reads only from BASE_RESUME_DIR to avoid path traversal issues.

• POST /profile — Create/update candidate profile from form input (job history, education, skills, contact)
• GET /profile/{id} — Fetch profile

• GET /resume/format?pdf_path=resume.pdf — Local PDF → structured resume JSON

• POST /job/parse — Parse job description into structured JobProfile

• POST /match — Returns match score + breakdown + gaps + evidence mapping

• POST /tailor/resume — Generates a tailored resume (structured + text)
• POST /tailor/cover-letter — Generates a tailored cover letter

• GET /export/resume.pdf?run_id=...
• GET /export/resume.docx?run_id=...
• GET /export/cover-letter.pdf?run_id=...
• GET /export/cover-letter.docx?run_id=...

• GET /export/resume.tex?run_id=... — Returns LaTeX source to paste into Overleaf

A controlled workflow is recommended (predictable + debuggable):

1. Parse candidate profile (or structure resume PDF/text)
2. Parse job description
3. Compute deterministic match score (Python)
4. Generate gap insights (LLM, structured output)
5. Tailor resume (ATS-friendly, evidence-based)
6. Tailor cover letter
7. Validate truthfulness (no fabricated claims)
8. Export: PDF/DOCX/LaTeX