| OPENAI_API_KEY= | |
| MODEL_NAME="gpt-realtime" | |
| # Local vision model (only used with --local-vision CLI flag) | |
| # By default, vision is handled by gpt-realtime when the camera tool is used | |
| LOCAL_VISION_MODEL=HuggingFaceTB/SmolVLM2-2.2B-Instruct | |
| # Cache for local VLM (only used with --local-vision CLI flag) | |
| HF_HOME=./cache | |
| # Hugging Face token for accessing datasets/models | |
| HF_TOKEN= | |
| # To select a specific use case with custom instructions and tools, to be placed in src/demos/<mydemo>/__init__.py | |
| DEMO="stone" |