.env.example

# Copy to .env (local) or set as HF Space secrets. Never commit real keys.
#
# ONE KNOB: set MODAL_WORKSPACE to your Modal workspace slug — all three endpoint
# URLs (vLLM, llama.cpp, FLUX) are derived from it (see endpoints.py). Switching
# workspaces is a one-line change here, not a code edit.
MODAL_WORKSPACE=your-workspace-slug

# Shared endpoint API keys (the server default is "local-dev-key" unless you set
# a *_API_KEY secret at deploy time).
MODAL_VLLM_API_KEY=change-me
MODAL_VLLM_MODEL=Qwen/Qwen3.5-27B-FP8
MODAL_LLAMACPP_API_KEY=change-me
MODAL_LLAMACPP_MODEL=minicpm5-1b
MODAL_FLUX_API_KEY=change-me

# --- Optional overrides: set a full URL to bypass MODAL_WORKSPACE derivation ---
# MODAL_VLLM_BASE_URL=https://<workspace>--small-hack-vllm-serve.modal.run/v1
# MODAL_LLAMACPP_BASE_URL=https://<workspace>--small-hack-llamacpp-serve.modal.run/v1
# MODAL_FLUX_URL=https://<workspace>--small-hack-flux-flux-generate.modal.run

# --- deploy-time (used by modal_app/*.py) ---
HF_TOKEN=
VLLM_MODEL=Qwen/Qwen3.5-27B-FP8
GGUF_REPO=openbmb/MiniCPM5-1B-GGUF
GGUF_FILE=MiniCPM5-1B-Q4_K_M.gguf
FLUX_MODEL=black-forest-labs/FLUX.1-schnell