Skip to main content

Media Generation (Optional)

The tool server can call Google AI Studio / Vertex AI models to generate images and videos. When these variables are unset the agents fall back to DuckDuckGo image search (for images) and skip video generation entirely.

Image Generation Variables

VariableDescription
IMAGE_GENERATE_GCP_PROJECT_IDGoogle Cloud project that owns the Vertex or GenAI resources for image generation.
IMAGE_GENERATE_GCP_LOCATIONRegion where the model is deployed (e.g., us-central1).
IMAGE_GENERATE_GCS_OUTPUT_BUCKETBucket where generated assets are written before they are handed back to the agent.
IMAGE_GENERATE_GOOGLE_AI_STUDIO_API_KEYAPI key issued by Google AI Studio that authorizes direct calls to the image model.

Video Generation Variables

VariableDescription
VIDEO_GENERATE_GCP_PROJECT_IDProject ID that hosts your video-generation pipelines.
VIDEO_GENERATE_GCP_LOCATIONRegion for the video model (us-central1 works for most deployments).
VIDEO_GENERATE_GCS_OUTPUT_BUCKETBucket for intermediate/exported video files.
VIDEO_GENERATE_GOOGLE_AI_STUDIO_API_KEYGoogle AI Studio key that has access to video models.

Setup Checklist

  1. Enable the Vertex AI API inside the specified project(s) and grant the service account referenced by GOOGLE_APPLICATION_CREDENTIALS permission to use the models and write to the buckets.
  2. Create the GCS buckets listed above (enable uniform bucket-level access and versioning if you need stronger controls).
  3. Generate an API key inside Google AI Studio that has the correct project/region scope and paste it into the relevant variable(s).
  4. Restart the tool server after updating the environment file; verify generation by asking the agent for a new image/video and checking the bucket for outputs.