# Preferred: gateway-agnostic configuration for any OpenAI-compatible endpoint
# (e.g. LiteLLM proxy, self-hosted OpenRouter mirror, or OpenRouter itself)
LLM_API_BASE_URL=https://openrouter.ai/api/v1
LLM_API_KEY=sk-or-REPLACE-ME

# Legacy fallback: Direct OpenRouter configuration (still supported for backward compatibility)
# If LLM_API_* vars are not set, the resolver will check these env vars before defaulting to OpenRouter
OPENROUTER_API_KEY=sk-or-REPLACE-ME
OPENROUTER_API_BASE=https://openrouter.ai/api/v1

# Optional: only for visual_object_replacement/visual_text_replacement data generation (REVE)
REVE_API_KEY=reve-REPLACE-ME

# Optional: only for Ollama-based inference (alternative to OpenRouter)
OLLAMA_API_KEY=REPLACE-ME

# Optional: only for HuggingFace model downloads
HUGGINGFACE_KEY=hf_REPLACE-ME