Attachments and Media Generation

Attachments give the Agent user assets; media generation lets the Agent call authorized image or video models. Both must be explicitly enabled.

Configure Attachments

Enable attachments in the Agent's Orchestration > Files & Images. Default limits:

Default extensions include PDF, Word, PowerPoint, Excel, TXT, Markdown, CSV, JSON, and HTML. Attachments are saved as Assets; the full text is not automatically injected into the conversation. The Agent calls the parsing capability on demand. Images go directly to the model only when the chat model supports vision.

Configure Image Generation

1. Turn on Image Generation and select an image model, or use the system default model.
2. The default size is 1024×1024, with width and height ranging 256-4096.
3. A single run generates up to 4 images by default, configurable from 1-10.
4. Allow Reference Images is on by default; keep it on when editing or reference-based generation is needed.

Configure Video Generation

1. Turn on Video Generation and select a video model.
2. The default duration is 5 seconds and the default single-run maximum is 10 seconds; both range from 1-30 seconds.
3. The default aspect ratio is 16:9.
4. The status polling interval defaults to 3000ms and the synchronous wait cap defaults to 120 seconds. After a timeout, a task may return a still-processing status rather than a failure.

Security and Failure Handling

• Image/video models must be authorized for the team and enabled.
• Reference images and generated artifacts are subject to Asset access control; do not copy unauthorized internal download URLs.
• Files that are too large, disallowed extensions, or insufficient vision capability return clear errors before sending or from the backend.
• When a video provider task fails, check the tool results and provider errors in the Agent logs.