Skip to content

Production Deployment Guide

LamiSema is designed to be deployed as a stateless containerized service with persistent object storage.

1. Local Production Stack (Docker Compose)

The easiest way to run the full stack (API + Minio) is using Docker Compose.

docker-compose up -d

This starts: - API Server: http://localhost:8001 - Minio Console: http://localhost:9001 (User: minioadmin, Pass: minioadmin)

All uploads and results will be persisted in a Docker volume named minio_data.

2. Environment Variables

Configure the API behavior using these variables:

Variable Default Description
LAMI_STORAGE_TYPE memory memory (volatile), disk (persistent /uploads), or s3 (persistent remote)
LAMI_S3_ENDPOINT — Minio/S3 endpoint (e.g. http://minio:9000)
LAMI_S3_ACCESS_KEY — S3 Access Key
LAMI_S3_SECRET_KEY — S3 Secret Key
LAMI_S3_BUCKET lamisema-vault Bucket name for storage
PORT 9001 Internal container port

3. Automatic Fallback Behavior

LamiSema is designed for high availability: - If LAMI_STORAGE_TYPE=s3 is set but the S3/Minio connection fails (e.g., Minio container is down), the API automatically falls back to disk storage in the /uploads directory. - This ensures the service remains functional even during storage outages. - You can check which backend is active via the GET / health endpoint (storage_backend field).

4. Kubernetes / Managed Cloud (S3)

To deploy to AWS/GCP without running Minio:

  1. Create an S3 bucket.
  2. Set LAMI_STORAGE_TYPE=s3.
  3. Provide the bucket credentials via ENV variables.
  4. If using AWS, you can leave LAMI_S3_ENDPOINT blank to use the default regional endpoint.

4. Manual Server Setup (No Docker)

If you must run on a bare Linux server:

  1. System Deps: 
    sudo apt-get install tesseract-ocr tesseract-ocr-nep libgl1
  2. Install: 
    pip install lamisema[s3]
  3. Run: 
    export LAMI_STORAGE_TYPE=s3
# Set other variables...
lamisema serve

5. Security Notes

  • Network: The API currently has no built-in auth. It should be deployed behind a reverse proxy (Nginx, Traefik) or in a private VPC.
  • Persistence: Results are stored in the results/ prefix of your bucket as JSON.
  • Cleanup: LamiSema does not automatically delete files from the storage backend. Implement bucket lifecycle rules to purge old files if needed.