Sign inSign up

rezaq/libpostal

By rezaq

β€’Updated 10 months ago

Production-ready, multi-arch image for libpostal address parsing with automatic model management.

Image
Machine learning & AI
0

943

rezaq/libpostal repository overview

⁠rezaq/libpostal

A lightweight, multi-arch Docker image containing the compiled libpostal C library.
The image does not include the 2GB+ model data. Instead, the model is downloaded automatically on first run and stored in /data.

This design keeps the image small and lets you manage models parametrically using environment variables.


β πŸš€ Features

  • ⚑ Small image size – model not baked into the image
  • 🧩 Multi-architecture – supports amd64 and arm64
  • πŸ”§ Parametric model management – choose model via environment variable
  • πŸ“¦ Volume-friendly – persist downloaded data across restarts
  • πŸ— Perfect base layer for building your own services (Go, Python, etc.)

β πŸ“¦ First Run & Model Download

On the first startup, libpostal’s dataset is automatically downloaded into /data

To avoid re-downloading on every run: To avoid re-downloading on every run:

πŸ’‘ Always mount a volume to /data

Example:

docker run -it \
  -v /path/to/folder:/data \
  rezaq/libpostal:latest

⁠🧰 Parametric Model Management

By default, the libpostal model is used.You can override it with Senzing model which is performing better.

docker run -it \
  -e MODEL=senzing \
  -v /path/to/folder:/data \
  rezaq/libpostal:latest

The startup script detects whether the correct model already exists in /data. If not, it downloads and initializes it automatically.


⁠More Info

GitHub page⁠

Tag summary

Content type

Image

Digest

sha256:c593d6a09…

Size

40.2 MB

Last updated

10 months ago

docker pull rezaq/libpostal