Apppublic

CMacD/AIC_PHASE1_POC

sourceHugging Faceupdated 4mo agoView on Hugging Face

0likes

App README

Assortment Intelligence Classifier (AIC)

Streamlit application for automated product attribute mapping. Analysts upload a flat-file assortment and the pipeline maps each product to validated MDM attribute values using historical lookup and ML classifiers, outputting a QC-ready Excel workbook. This gets passed to Phase 2 and 3 which perform data cleaning/transformations and analyst qc.

Deployment

The app runs as a Docker container exposing port 7860.

bash

docker build -t aic .
docker run -p 7860:7860 aic

Entry point: 1_Phase_1_Attribute_Mapping.py (Streamlit multipage app).

Repository structure

1_Phase_1_Attribute_Mapping.py   # Phase 1 — attribute mapping (Streamlit page 1)
pages/
  2_Phase_3_Pipeline_and_QC.py   # Phase 2 & 3 — processing pipeline and QC (page 2)
phase3_package/                  # Phase 3 backend modules
Ensemble.py                      # Predictor combiner and QC annotator
MappingLookup.py                 # Fuzzy historical match engine
TextMatch.py                     # BM25 text retrieval predictor
RandomForest_XGB.py              # XGBoost TF-IDF classifier
_write_results.py                # Excel output writer
aic_utils.py                     # Shared utilities
Dockerfile
requirements.txt
.streamlit/config.toml
test/                            # Unit tests

Requirements

Python 3.11. All dependencies in requirements.txt.

CI

.github/workflows/sync.yml syncs the main branch to the Hugging Face Spaces demo on every push. Requires HF_TOKEN secret configured in the repository.