ikpu 0.1.0

IKPU/MXIK product classifier for Uzbekistan tax codes

ikpu

Python library for matching products with IKPU/MXIK codes (Uzbekistan tax classification) by product name or barcode.

Uses TF-IDF similarity search over the official tasnif.soliq.uz catalog.

Installation

pip install ikpu

With FastAPI server support:

pip install ikpu[server]

Quick Start

from ikpu import IKPUMatcher

matcher = IKPUMatcher()

# Download the catalog (first time only)
matcher.update()

# Load indexes
matcher.load()

# Search by product name
results = matcher.search("Coca-Cola 0.5", language="ru", top_k=5)
for r in results:
    print(f"{r['ikpu']} — {r['name']} ({r['similarity']:.2f})")

# Search by barcode
results = matcher.search_barcode("4600494600029", language="ru")

API Server

# Start the server
ikpu-server

# Or with uvicorn directly
uvicorn ikpu.api:app --host 0.0.0.0 --port 8044

Endpoints

Method	Path	Description
POST	/search	Search by name or barcode
GET	/health	Service status
POST	/rebuild-cache	Rebuild TF-IDF indexes
POST	/check-update	Check for catalog updates

POST /search

{
  "query": "Coca-Cola 0.5",
  "barcode": null,
  "language": "ru",
  "top_k": 5
}

CLI

# Download/update the IKPU catalog
ikpu-update

Configuration

Data is stored in ~/.ikpu/data/ by default. Override with the IKPU_DATA_DIR environment variable:

export IKPU_DATA_DIR=/path/to/custom/dir

Supported Languages

• ru — Russian
• uz_cyrl — Uzbek (Cyrillic)
• uz_latn — Uzbek (Latin)

License

MIT