Resume Processing Microservice - Production Build

AI-powered resume and cover letter generation API optimized for high-concurrency deployment.

Overview

This microservice provides a JSON-only API that takes job descriptions and user data as input, and returns:

• Structured job analysis
• Tailored resume (JSON format)
• Customized cover letter text

No PDF generation. No Google Drive integration. Pure JSON I/O.

Architecture

build/
├── api.py                    # FastAPI application entry point
├── orchestrator.py           # Async pipeline orchestration
├── modules/                  # Core processing modules
│   ├── structuring/          # Job description analysis
│   ├── enhancing/            # Resume enhancement (coordinator, bullets, profile, skills)
│   └── cover_letter/         # Cover letter generation
├── deployment/               # Deployment files
│   ├── Dockerfile
│   ├── .dockerignore
│   ├── deploy_cloudrun.sh
│   └── deploy_cloudrun.ps1
└── requirements.txt          # Python dependencies

Features

• High Concurrency: 4 Uvicorn workers, 100 concurrent requests per instance
• Async Pipeline: Maximized parallelization (3 parallel bullet generations)
• Auto-scaling: 1-3 instances on Cloud Run (300 max concurrent users)
• JSON Output: Resume and cover letter returned as structured JSON
• Multi-language: Supports English and French job descriptions
• ATS Optimized: Keyword matching and scoring for applicant tracking systems

API Endpoints

POST /process

Main endpoint - Full pipeline processing with dynamic experience configuration

Request:

{
  "job_text": "Job description text here...",
  "user_json": {
    "personal": { "name": "John Doe", "title": "Software Engineer" },
    "contact": { "email": "john@example.com", "phone": "+33..." },
    "projects_database": {
      "Project A": { "contexte": "...", "technologies": [...], "metiers": [...], "realisations": [...] },
      "Project B": { "..." },
      "Project C": { "..." }
    },
    "skills_database": { "skills": {...}, "essential_skills": [...] },
    "experiences_config": [
      {
        "candidate_projects": [0, 1],
        "role_strategy": "enhanced",
        "content_strategy": "enhanced"
      },
      {
        "candidate_projects": [1, 2],
        "role_strategy": "enhanced",
        "content_strategy": "direct"
      },
      {
        "candidate_projects": [2],
        "role_strategy": "direct",
        "content_strategy": "direct"
      }
    ],
    "education": [...],
    "languages": [...]
  },
  "config_json": { /* configuration */ }
}

Experience Configuration Options:

• candidate_projects: List of project indexes (0-based) referencing projects_database order
• role_strategy:
  • "direct" - Use role title exactly as-is from project
  • "enhanced" - Adapt role title for ATS optimization
• content_strategy:
  • "direct" - Use bullet points directly from project (no AI enhancement)
  • "enhanced" - Generate AI-enhanced bullet points for job fit

Response:

{
  "success": true,
  "structured_job": { /* analyzed job data */ },
  "resume": {
    "personal": { "name": "...", "title": "..." },
    "contact": { "email": "...", "phone": "..." },
    "profile": "Professional summary text...",
    "experience": [
      {
        "role": "Software Engineer",
        "company": "Company Name",
        "location": "Paris, France",
        "start_date": "2023-01",
        "end_date": "2024-12",
        "bullets": ["Bullet 1", "Bullet 2", "..."],
        "is_direct": false
      }
    ],
    "skills": {
      "technical": ["Python", "FastAPI", "..."],
      "soft": ["Leadership", "Communication", "..."]
    },
    "education": [...],
    "certifications": [...],
    "languages": [...]
  },
  "cover_letter": "Cover letter text...",
  "metadata": {
    "processing_time_seconds": 12.5,
    "language": "en",
    "experiences": {
      "total": 3,
      "enhanced": 2,
      "direct": 1,
      "projects_used": ["Project A", "Project B", "Project C"]
    },
    "average_ats_score": 85.3
  }
}

POST /structure

Job description structuring only

GET /health

Health check endpoint

GET /

Service information

Local Development

Prerequisites

• Python 3.13+
• OpenAI API key

Setup

# Install dependencies
pip install -r requirements.txt

# Set environment variables
export OPENAI_API_KEY="your-openai-key"
export API_SECRET_KEY="your-secret-key"

# Run locally
python api.py

API will be available at:

• API: http://localhost:8080
• Docs: http://localhost:8080/docs

Deployment to Google Cloud Run

Prerequisites

• Google Cloud SDK installed and configured
• Docker Desktop installed
• Environment variables set:
  • OPENAI_API_KEY
  • API_SECRET_KEY

Deploy (Windows)

cd deployment
.\deploy_cloudrun.ps1

Deploy (Linux/Mac)

cd deployment
./deploy_cloudrun.sh

Manual Deployment

# Build Docker image locally
docker build -f deployment/Dockerfile -t gcr.io/castal-job-tracker/resume-processor .

# Configure Docker auth
gcloud auth configure-docker gcr.io

# Push image
docker push gcr.io/castal-job-tracker/resume-processor

# Deploy to Cloud Run
gcloud run deploy resume-processor \
  --image gcr.io/castal-job-tracker/resume-processor \
  --region europe-west9 \
  --memory 8Gi \
  --cpu 4 \
  --timeout 600 \
  --concurrency 100 \
  --min-instances 1 \
  --max-instances 3 \
  --set-env-vars "OPENAI_API_KEY=${OPENAI_API_KEY},API_SECRET_KEY=${API_SECRET_KEY},WORKERS=4" \
  --allow-unauthenticated

Configuration

Concurrency Settings

• Workers per instance: 4
• Requests per instance: 100
• Min instances: 1 (always warm, no cold starts)
• Max instances: 3 (budget constraint: <$100/month)
• Total capacity: 300 concurrent users

Cost Estimation

• Baseline: ~$40/month (1 instance always running)
• Typical: ~$60-80/month (1-2 instances average)
• Maximum: ~$120/month (3 instances 24/7)

Testing

# Health check
curl https://your-service-url.run.app/health

# Full pipeline test
curl -X POST "https://your-service-url.run.app/process" \
  -H "Authorization: Bearer your-api-key" \
  -H "Content-Type: application/json" \
  -d @test_request.json

Performance

Expected metrics:

• Processing time: 10-20s per request
• Success rate: >95%
• Median latency: <30s
• Throughput: ~5-10 req/s (with auto-scaling)

Security

• API key authentication required (Bearer token)
• Secrets passed via environment variables (never in code)
• No file storage (stateless service)
• HTTPS only (enforced by Cloud Run)

Monitoring

View logs and metrics:

# Logs
gcloud run services logs tail resume-processor --region europe-west9

# Metrics dashboard
https://console.cloud.google.com/run/detail/europe-west9/resume-processor/metrics

Troubleshooting

High latency

• Check OpenAI API response times
• Verify worker count is appropriate
• Check instance memory usage

Rate limiting

• Increase max-instances if needed
• Check OpenAI API quotas

Out of memory

• Reduce concurrency per instance
• Increase memory allocation

Support

For issues or questions, refer to the main project documentation.

Version

v3.0.0 - Dynamic Experience Configuration (January 2025)

• Dynamic number of experiences (not limited to 3)
• Per-experience project candidate pools
• Direct vs Enhanced content strategies
• Direct vs Enhanced role strategies
• Intelligent coordinator that respects constraints