# WP38 — reproducible pipeline (run from the project root; requires the authorised snapshot in data/raw/, see data/raw/README.md)
PY := .venv/bin/python
.PHONY: all env ingest clean match validate external features models robust tables figures paper test lit

all: ingest clean match validate categories features models robust tables figures paper

env:
	uv venv --python 3.12 .venv && uv pip install --python $(PY) -e .

ingest:
	$(PY) -m src.ingest.load_snapshot

external:        # re-download public covariates and geography (manifest in data/external/)
	$(PY) -m src.ingest.external

clean:
	$(PY) -m src.clean.listings && $(PY) -m src.clean.siblings && $(PY) -m src.clean.listings && $(PY) -m src.clean.stores && $(PY) -m src.reports.data_dictionary

match:
	$(PY) -m src.features.match_products

categories:      # A-002 classifier (needs cached labels in data/processed/; API keys only to regenerate)
	$(PY) -m src.features.category_model embed && $(PY) -m src.features.category_model train

validate:        # LLM annotations need OPENAI_API_KEY / ANTHROPIC_API_KEY in .env; labels are kept in data/processed/*.csv (never overwritten)
	$(PY) -m src.validation.category_audit report && $(PY) -m src.validation.match_validation report && $(PY) -m src.reports.phase2_audit

features:
	$(PY) -m src.features.region_category && $(PY) -m src.features.diversity

models:
	$(PY) -m src.models.h1_raa && $(PY) -m src.models.h2_scale && $(PY) -m src.models.h3_longtail && $(PY) -m src.models.h4_dynamics \
	&& $(PY) -m src.models.h5_hedonic && $(PY) -m src.models.h7_complexity && $(PY) -m src.models.qdaci && $(PY) -m src.models.summary $${WP38_LABEL:-rule}

robust:
	$(PY) -m src.validation.robustness && $(PY) -m src.validation.external_validation

tables figures:
	$(PY) -m src.reports.make_tables && $(PY) -m src.reports.make_figures

paper:
	cd paper && latexmk -pdf -interaction=nonstopmode main.tex

test:
	$(PY) -m pytest -q

lit:
	$(PY) -m src.literature.verify_refs && $(PY) -m src.literature.make_bib