Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
You can enable users of your agent to upload files and images. Your agent can analyze these files and use them to provide responses.
When to allow file input
Allow users to upload files to elicit agent responses in the following scenarios:
- Questions that require information from your configured knowledge sources.
- Questions that can be answered from an image or uploaded supported file types, such as text, CSV, or PDF files.
When you allow users to upload files and your agent to generate responses based on those files, follow best practices for prompt engineering.
To get the most out of this feature, review the following tips:
Upload images that are clear and high-quality. Agents might struggle to interpret grainy, poorly lit, or pixelated images.
Only upload supported file types:
- DOCX
- CSV
- TXT
- JPG
- PNG
- WebP
- nonanimated GIF files
Only upload files that meet these size restrictions:
- Individual file size: 15 MB
- Text character limit:
- Without code interpreter: 30,000 characters per file and total when there's a multiple file upload
- With code interpreter: There's no character limit
If a file triggers content moderation filtering, restart the conversation. The agent uses the current conversation history to generate answers and continues to return content moderation filtering errors if objectionable content is part of the conversation history.
Note
If you're having problems generating high-quality responses from files, turn on the Allow ungrounded responses setting in the Knowledge section of your agent's Generative AI settings. This setting is required if there are no relevant knowledge sources configured. For more information, see Allow ungrounded responses.
If you publish your agent to a SharePoint channel, users can't upload files.
If your agent resides in a customer managed key enabled environment, adding files as input is allowed, but the agent doesn't process the files.
Support for XLSX and PPTX file types is currently only available in experimental mode. If you want to use these file types, reach out to Microsoft support or your Microsoft representative.
Turn on file uploads
Makers can allow users of their agent to upload files during chat interactions with their agent.
Go to the Settings page for your agent, and select Generative AI.
Navigate to the File processing capabilities section.
Turn on File uploads.
(Optional) Select your preferred level of strictness for content moderation.
Select Save.
Test your agent's file analysis
After you turn on File uploads, you can test your agent's capability by uploading files and images and verifying the agent's responses. To upload a file, select the paper clip icon in the Test your agent chat box. Thoroughly test your agent's accuracy by using various files and images with various queries to validate the agent's responses before you publish.
Responsible AI
Microsoft is committed to strong responsible AI practices. Review the responsible AI guidelines for Copilot Studio before using any AI features. Like all generative answers, content is checked twice: first during user input and again before the agent generates an output response. If the system finds harmful, offensive, or malicious content at either the input or output stage, the agent doesn't respond.