A full-stack web application that optimizes LaTeX resumes for specific job descriptions using AI and generates tailored cover letters.

• LaTeX Resume Upload - Drag-and-drop your .tex resume
• AI-Powered Optimization - Tailors your resume to match job descriptions
• Smart Skill Targeting - Prioritizes a small set of missing high-value technical skills
• Cover Letter Generation - Creates personalized cover letters
• PDF Export - Compiles optimized LaTeX to downloadable PDF
• ATS-Friendly - Keeps formatting compatible with applicant tracking systems
• Persistent Per-User Base Resume - each authenticated user uploads once, then reuses their own baseline resume

haraesume/
├── backend/                 # Go API server
│   ├── go.mod
│   └── main.go              # Routes, LLM integration, PDF compilation
├── frontend/                # React SPA
│   ├── package.json
│   ├── vite.config.ts
│   └── src/
│       ├── App.tsx          # 4-step wizard UI
│       └── index.css        # Theme styles
└── sample_resume.tex        # Test resume

flowchart LR
  UI["React Frontend\n(frontend/src/App.tsx)"] --> API["Go HTTP Server\nbackend/main.go"]

  API --> ROUTES["Route Handlers\n/upload-resume, /optimize,\n/cover-letter, /generate-application-package"]
  ROUTES --> STATE["Per-User In-Memory App State\nbase resume, active resume,\ncover letter, job description"]
  ROUTES --> LLM["LLM Adapter\nOpenRouter chat completions"]
  ROUTES --> PDF["LaTeX Compiler Service\nlatexmk/tectonic/xelatex/pdflatex"]
  ROUTES --> STORE["Filesystem Storage\nstate/user_resumes/<user_hash>/base_resume.tex\napplications/<user_hash>/<company>/"]

  LLM --> OR["OpenRouter API"]
  PDF --> FS["Generated Artifacts\nresume.pdf, cover_letter.pdf"]

flowchart LR
  classDef card fill:#fbf7e6,stroke:#6f6a55,stroke-width:1.5px,color:#222,rx:8,ry:8;
  classDef result fill:#e8f4ea,stroke:#3f6f4a,stroke-width:1.5px,color:#1f3d26,rx:8,ry:8;

  A["Resume Input\n(base .tex)"]:::card
  B["Job Description Input"]:::card
  C["Resume Optimizer Agent\n(LLM)"]:::card
  D["Cover Letter Agent\n(LLM)"]:::card
  E["Company Name Extractor Agent\n(LLM + fallback heuristic)"]:::card
  F["Applications Folder Writer\napplications/<user_hash>/<company>/"]:::card
  G["PDF Compiler Node"]:::card
  H["Cleanup Node\nremove generated .tex"]:::card
  I["Final Artifacts\nAkbari, Deep.pdf\nCV_Deep.pdf"]:::result

  A --> C
  B --> C
  C --> D
  B --> E
  C --> G
  D --> G
  E --> F
  G --> F
  F --> H
  H --> I

• Go (1.22+)
• Node.js (20+)
• TeX Live (for pdflatex)
• OpenRouter API key
• Auth0 tenant with:
  • One SPA application (for frontend login)
  • One custom API (for audience/access tokens)

git clone https://github.com/Ackberry/haraesume.git
cd haraesume

cd backend
export OPENROUTER_API_KEY="your-openrouter-api-key"
export AUTH_PROVIDER="auth0"
export AUTH0_DOMAIN="your-tenant.us.auth0.com"
export AUTH0_ISSUER_BASE_URL="https://your-tenant.us.auth0.com"
export AUTH0_AUDIENCE="https://haraesume.ackberry.dev"
go run .
# Server: http://localhost:3001

cd frontend
npm install
export VITE_AUTH0_DOMAIN="your-tenant.us.auth0.com"
export VITE_AUTH0_CLIENT_ID="your-auth0-spa-client-id"
export VITE_AUTH0_AUDIENCE="https://api.haraesume.com"
export VITE_AUTH0_REDIRECT_URI="http://localhost:5173"
npm run dev
# App: http://localhost:5173

1. Open http://localhost:5173
2. Sign in (or create account) via Auth0
3. Upload a .tex resume (use sample_resume.tex to test)
4. Paste the job description
5. Click Optimize Resume
6. Download PDF or generate a cover letter

After the first upload, the backend persists your base resume for that authenticated user. On later sessions, that same user can go straight to the job description step.

You can keep a single root .env at haraesume/.env.

• Frontend reads root env via frontend/vite.config.ts (envDir: '..')
• Backend auto-loads .env from either backend/.env or root ../.env

Auth0 app settings should include:

• Allowed Callback URLs: http://localhost:5173
• Allowed Logout URLs: http://localhost:5173
• Allowed Web Origins: http://localhost:5173

Resume upload is .tex only. PDF is generated as output via /api/generate-pdf.

• Railway ephemeral disk is cleared on redeploy/restart unless you use a volume.
• Mount a Railway volume (example mount path: /data).
• Set:
  • RESUME_STORE_PATH=/data/user_resumes
  • APPLICATIONS_ROOT_PATH=/data/applications
• If these vars are unset and /data exists, the backend now defaults to:
  • /data/user_resumes for base resumes
  • /data/applications for generated application artifacts
• With that setup:
  • New authenticated users start with no base resume.
  • After upload, each user keeps their own base resume across restarts/redeploys.

This repo now includes a dedicated LangChain + LangGraph multi-node agent flow for resume checking and job matching:

• Location: backend/agents/
• Install deps: pip install -r backend/agents/requirements.txt
• Entry point:
  • cd backend
  • python -m agents --resume-file ../sample_resume.tex --jd-file ./job_description.txt

• extract_keywords - keyword extraction for resume/JD text
• ats_lint_resume - ATS format and completeness checks
• parse_job_requirements - required/preferred skill extraction
• compute_match_score - required/preferred/overall match scores
• identify_skill_gaps - matched vs missing skill detection

1. validate -> checks required inputs
2. resume_check -> ATS lint + resume keywords
3. job_analysis -> JD requirements + keywords
4. matching -> coverage scoring + gap detection
5. recommendations -> actionable improvements
6. finalize -> final report synthesis

Edges:

• START -> validate
• validate -> resume_check (if valid) else validate -> finalize
• resume_check -> job_analysis -> matching -> recommendations -> finalize -> END

cd backend
go test ./...
go run .

cd frontend
npm run dev
npm run build
npm run preview

MIT