Makefile

# ==============================================================================
# Foundry Makefile
# ==============================================================================

REGISTRY ?= ghcr.io/infernet-org/foundry
# Default model (can be overridden: make run MODEL=hermes-4.3-36b)
MODEL ?= qwen3.5-9b
MODEL_TAG ?= $(REGISTRY)/$(MODEL)
PORT ?= 8080
MODELS_DIR ?= $(HOME)/.cache/foundry

.PHONY: help build run run-profile test benchmark monitoring down push push-all clean clean-models download

help: ## Show this help
	@echo "Available models: qwen3.5-9b (default), qwen3-coder-30b-a3b, hermes-4.3-36b, qwen3.5-35b-a3b"
	@echo "Usage: make run MODEL=qwen3-coder-30b-a3b"
	@echo ""
	@grep -E '^[a-zA-Z_-]+:.*?## .*$$' $(MAKEFILE_LIST) | sort | \
		awk 'BEGIN {FS = ":.*?## "}; {printf "\033[36m%-20s\033[0m %s\n", $$1, $$2}'

# --- Build -------------------------------------------------------------------

build: ## Build the model image
	@cp scripts/entrypoint.sh models/$(MODEL)/entrypoint.sh
	docker build \
		-t $(MODEL_TAG):latest \
		models/$(MODEL)/

# --- Run ---------------------------------------------------------------------

up: ## Start via docker compose (detatched)
	FOUNDRY_MODEL=$(MODEL) docker compose up -d

monitoring: ## Start via docker compose with full monitoring stack
	FOUNDRY_MODEL=$(MODEL) docker compose --profile monitoring up -d

down: ## Stop all docker compose services
	docker compose --profile monitoring down

run: ## Run the model container directly (auto-detect GPU)
	@mkdir -p $(MODELS_DIR)
	docker run --gpus all \
		--shm-size 2g \
		--ulimit memlock=-1:-1 \
		--ulimit stack=67108864:67108864 \
		--sysctl net.core.somaxconn=4096 \
		--sysctl net.ipv4.tcp_keepalive_time=60 \
		-p $(PORT):8080 \
		-v $(MODELS_DIR):/models \
		--name foundry-$(MODEL) \
		--rm \
		$(MODEL_TAG):latest

run-profile: ## Run with explicit profile (PROFILE=rtx5090)
	@mkdir -p $(MODELS_DIR)
	docker run --gpus all \
		--shm-size 2g \
		--ulimit memlock=-1:-1 \
		--ulimit stack=67108864:67108864 \
		--sysctl net.core.somaxconn=4096 \
		--sysctl net.ipv4.tcp_keepalive_time=60 \
		-p $(PORT):8080 \
		-v $(MODELS_DIR):/models \
		-e FOUNDRY_PROFILE=$(PROFILE) \
		--name foundry-$(MODEL) \
		--rm \
		$(MODEL_TAG):latest

# --- Test --------------------------------------------------------------------

test: ## Smoke test: start container, wait for health, send one request
	@echo "Starting container..."
	@mkdir -p $(MODELS_DIR)
	@docker run --gpus all -d \
		--shm-size 2g \
		--ulimit memlock=-1:-1 \
		--ulimit stack=67108864:67108864 \
		--sysctl net.core.somaxconn=4096 \
		--sysctl net.ipv4.tcp_keepalive_time=60 \
		-p $(PORT):8080 \
		-v $(MODELS_DIR):/models \
		--name foundry-test-$(MODEL) \
		$(MODEL_TAG):latest
	@echo "Waiting for server to be ready..."
	@for i in $$(seq 1 60); do \
		if curl -sf http://localhost:$(PORT)/health > /dev/null 2>&1; then \
			echo "Server ready after $$i seconds"; \
			break; \
		fi; \
		if [ $$i -eq 60 ]; then \
			echo "Timeout waiting for server"; \
			docker logs foundry-test-$(MODEL); \
			docker rm -f foundry-test-$(MODEL); \
			exit 1; \
		fi; \
		sleep 1; \
	done
	@echo "Sending test request..."
	@curl -s http://localhost:$(PORT)/v1/chat/completions \
		-H "Content-Type: application/json" \
		-d '{"model":"$(MODEL)","messages":[{"role":"user","content":"Say hello in one sentence."}],"max_tokens":64}' \
		| python3 -m json.tool
	@echo ""
	@echo "Test passed. Cleaning up..."
	@docker rm -f foundry-test-$(MODEL)

# --- Download ----------------------------------------------------------------

download: ## Download the GGUF model file
	./scripts/download-model.sh --model $(MODEL)

# --- Benchmark ---------------------------------------------------------------

benchmark: ## Run benchmark against a running server (PORT=8080)
	python3 scripts/benchmark.py --url http://localhost:$(PORT) --mode all

# --- Push --------------------------------------------------------------------

push: ## Push model image to GHCR
	docker push $(MODEL_TAG):latest

push-all: ## Push all tags to GHCR
	docker push --all-tags $(MODEL_TAG)

# --- Clean -------------------------------------------------------------------

clean: ## Remove local images
	-docker rmi $(MODEL_TAG):latest

clean-models: ## Remove downloaded models
	rm -rf "$(MODELS_DIR)"/*.gguf