Skip to content

Installation

Requirements

  • Python 3.10, 3.11, or 3.12
  • Tesseract OCR with Nepali language pack (for legacy/scanned PDFs)

Install LamiSema

pip install lamisema

This installs the core pipeline with pdfplumber and PyMuPDF. Tesseract is not bundled — it is a system binary installed separately.

Install with S3/Minio support

pip install "lamisema[s3]"

This adds boto3 for persistence to S3-compatible object storage.

Install all extras

pip install "lamisema[all]"

This includes Tesseract support, EasyOCR, and S3 persistence drivers.

Install from source

git clone https://github.com/sanjiblamichhane/lamisema
cd lamisema
pip install -e .

Install Tesseract

Tesseract is required for legacy-encoded and scanned PDFs. The Nepali language pack (nep) must be installed alongside the binary.

brew install tesseract tesseract-lang
tesseract --list-langs | grep nep   # verify

sudo apt-get install tesseract-ocr tesseract-ocr-nep
tesseract --list-langs | grep nep   # verify

Download the installer from the Tesseract GitHub releases. During installation, select Nepali from the additional language packs list.

Verify installation

from lamisema import LamiSema

lamisema = LamiSema()
print(lamisema.ocr_backend)   # TesseractBackend or EasyOCRBackend or None
print(lamisema.storage)       # InMemoryStorage (default)

Or start the API server and hit the health endpoint:

lamisema serve
# → http://localhost:9001/

The response shows which libraries and storage backends are active:

{
  "status": "online",
  "libraries": {
    "PyMuPDF": true,
    "pdfplumber": true,
    "pytesseract": true,
    "easyocr": false
  },
  "ocr_backend": "tesseract",
  "storage_backend": "InMemoryStorage"
}

Running the API server

# via installed CLI (Recommended)
lamisema serve

# via Python module
python -m lamisema.api.app

# via uvicorn directly (with persistent storage fallback test)
export LAMI_STORAGE_TYPE=s3
uvicorn lamisema.api.app:app --port 9001 --reload

Interactive docs: http://localhost:9001/docs