Makefile

# Makefile for rag-retrieval-benchmark.
#
# Most commands run inside the Docker services. `make up` builds and starts the
# stack; the data/index/evaluate commands run one-off processes in the `api`
# image (which has the package and GPU access).

COMPOSE := docker compose
# Run one-off commands in a throwaway container based on the api service.
RUN := $(COMPOSE) run --rm api

.DEFAULT_GOAL := help

.PHONY: help up down logs jupyter build download-data prepare-data index retrieve evaluate test lint format

help: ## Show this help message
	@grep -E '^[a-zA-Z_-]+:.*?## .*$$' $(MAKEFILE_LIST) | \
		awk 'BEGIN {FS = ":.*?## "}; {printf "  \033[36m%-16s\033[0m %s\n", $$1, $$2}'

build: ## Build the Docker images
	$(COMPOSE) build

up: ## Build and start all services (qdrant + jupyter + api)
	$(COMPOSE) up -d --build

down: ## Stop and remove all services
	$(COMPOSE) down

logs: ## Follow logs from all services
	$(COMPOSE) logs -f

jupyter: ## Print the JupyterLab URL (token is disabled for local dev)
	@echo "JupyterLab: http://localhost:8888/lab"
	@$(COMPOSE) logs jupyter | tail -n 15 || true

download-data: ## Download the BEIR SciFact dataset into data/raw/
	$(RUN) python -m rag_retrieval_benchmark.data.download

prepare-data: ## Normalize raw data into data/processed/*.jsonl
	$(RUN) python -m rag_retrieval_benchmark.data.prepare

index: ## Embed the corpus and index it into Qdrant
	$(RUN) python -m rag_retrieval_benchmark.indexing.build_index

retrieve: ## Run a sample retrieval query (override with QUERY=..., METHOD=..., TOP_K=...)
	$(RUN) python -m rag_retrieval_benchmark.cli retrieve \
		--query "$(or $(QUERY),Does the claim have supporting evidence?)" \
		--method "$(or $(METHOD),dense)" \
		--top-k "$(or $(TOP_K),5)"

evaluate: ## Run the full benchmark and generate the report
	$(RUN) python -m rag_retrieval_benchmark.evaluation.runner

test: ## Run the test suite
	$(RUN) pytest

lint: ## Run the linter
	$(RUN) ruff check .

format: ## Auto-format the code
	$(RUN) ruff format .