| # Slide Skill OpenEnv — Environment Variables | |
| # | |
| # Copy this file to .env and fill in the values. | |
| # Never commit .env to version control. | |
| # --------------------------------------------------------------------------- | |
| # Required | |
| # --------------------------------------------------------------------------- | |
| # Google Gemini API key — used by all three LLM roles: | |
| # Generator: Gemini 3 Flash (writes pptxgenjs JavaScript) | |
| # Evaluator: Gemini 3.1 Pro (scores the slide with vision) | |
| # Optimizer: Gemini 3.1 Pro (rewrites DESIGN_RULES.md) | |
| # Get your key at: https://aistudio.google.com/app/apikey | |
| GEMINI_API_KEY=your_gemini_api_key_here | |
| # Anthropic API key (optional — for Claude-based evaluation) | |
| ANTHROPIC_API_KEY=your_anthropic_api_key_here | |
| # Patronus API key (optional — for Patronus Judge MM gate evaluation) | |
| PATRONUS_API_KEY=your_patronus_api_key_here | |
| # --------------------------------------------------------------------------- | |
| # Optional — override defaults | |
| # --------------------------------------------------------------------------- | |
| # Maximum number of optimization steps per episode (default: 7). | |
| # Each step takes ~60-120s. At 7 steps, a full episode runs ~7-14 minutes. | |
| # Reduce for faster iteration during development; increase for deeper optimization. | |
| # SLIDE_SKILL_MAX_STEPS=7 | |
| # Root directory for session working directories (default: repo/tmp/ locally). | |
| # On HuggingFace Spaces the app directory is read-only, so set this to /tmp: | |
| # SLIDE_SKILL_SESSION_ROOT=/tmp | |
| # --------------------------------------------------------------------------- | |
| # HuggingFace Spaces (set these as Space secrets, not in .env) | |
| # --------------------------------------------------------------------------- | |
| # When deploying to HF Spaces, add GEMINI_API_KEY as a repository secret | |
| # via the Space settings UI. Do not hardcode it in the Dockerfile or source. | |