Makefile

PYTHON := .venv/bin/python
RUN := PYTHONPATH=src $(PYTHON)
ARGS ?=

.DEFAULT_GOAL := help
.PHONY: help init setup up down status sync dataset-export dataset-check eval view test

help: ## Show available targets
	@grep -E '^[a-z-]+:.*?## ' $(MAKEFILE_LIST) | awk -F':.*?## ' '{printf "  \033[36m%-16s\033[0m %s\n", $$1, $$2}'

init: ## Install, start, and seed everything for a fresh clone
	@test -f .env || { \
		echo "No .env found. Run 'cp .env.example .env' and replace every change-me value." >&2; \
		exit 1; }
	@! grep -q change-me .env || { \
		echo ".env still has change-me placeholders; replace them before starting." >&2; \
		exit 1; }
	$(MAKE) setup
	$(MAKE) up
	$(MAKE) sync

setup: ## Install locked dependencies
	uv sync --frozen --all-groups

up: ## Start Langfuse and the Handy tracing proxy
	docker compose up --detach --build --wait

down: ## Stop the stack without deleting volumes
	docker compose down

status: ## Show local stack status
	docker compose ps

sync: ## Bootstrap Langfuse prompts and dataset from tracked seeds
	$(RUN) -m transcript_cleanup_bench.prompts
	$(RUN) -m transcript_cleanup_bench.dataset bootstrap

dataset-export: ## Refresh the tracked dataset snapshot atomically
	$(RUN) -m transcript_cleanup_bench.dataset export

dataset-check: ## Check the tracked snapshot for Langfuse drift
	$(RUN) -m transcript_cleanup_bench.dataset check

eval: ## Run diagnostic experiments (default concurrency 8)
	$(RUN) -m transcript_cleanup_bench.experiment $(ARGS)

view: ## Open Langfuse in the default browser
	open http://localhost:4001

test: ## Run the test suite
	.venv/bin/pytest