# Copy to .env. Supply credentials only for the pipelines you run.
# Never commit .env or credentials. Exported shell variables take precedence.

# VideoDB Activity: required for hosted understanding/indexing/search.
# Uses the SDK's public API endpoint by default.
VIDEO_DB_API_KEY=

# TwelveLabs: required for Marengo embeddings and hosted retrieval.
TWELVE_LABS_API_KEY=

# Hugging Face: needed for gated/private model acquisition or dataset access.
# Public model downloads and already-acquired local inference need no token.
# HF_TOKEN=

# AWS: choose your normal credential chain (SSO/profile, VM role, or keys).
# Prefer an existing authenticated profile or VM role; do not fill both methods.
# AWS_PROFILE=my-profile
# AWS_ACCESS_KEY_ID=
# AWS_SECRET_ACCESS_KEY=
# AWS_SESSION_TOKEN=             # Required only for temporary key credentials.
# AWS_SHARED_CREDENTIALS_FILE=/absolute/path/credentials
# AWS_CONFIG_FILE=/absolute/path/config
# AWS_DEFAULT_REGION=us-east-1   # CLI default; benchmark uses aws.region in config.

# Optional local GPU/cache settings. Inference uses acquired local weights.
# CUDA_VISIBLE_DEVICES=0
# HF_HOME=/absolute/path/huggingface-cache
# HF_HUB_OFFLINE=1               # Inference only; unset for downloads.
# TRANSFORMERS_OFFLINE=1         # Inference only; unset for downloads.

# Model paths, runtime interpreters, dataset paths, provider resource IDs,
# sampling, concurrency and budgets belong in run configs, not this file.
# The runner loads the configured env_file. External hf/aws/curl tools and
# robotics-bench-aws setup commands use shell credentials or their own login;
# they do not automatically read this file.