aws-samples / aws-samples/generative-ai-use-cases
Discrepancy between the maximum number of images that can be attached to a chat and actual behavior
- Dominant language
- TypeScript
- Stars
- 1.4k
- Forks
- 433
- Avg merge
- 2h 35m
- Merged PRs (30d)
- 1
Description
Converse API allows up to 20 image attachments, and it correctly displays an error when trying to attach more than 20 images on the chat screen.
> You can include up to 20 images. Each image's size, height, and width must be no more than 3.75 MB, 8,000 px, and 8,000 px, respectively.
https://docs.aws.amazon.com/bedrock/latest/userguide/conversation-inference-call.html
https://github.com/aws-samples/generative-ai-use-cases/blob/201dfc19be3af42a955060985d8d10283ff527b7/packages/web/src/pages/ChatPage.tsx#L35
While Claude is fine, Pixtral has a token-based limit rather than a limit on the number of images, Llama3.2 has a limit of 1 image (though this is not documented), and Llama 4 has an upper limit of 8 images, with an error being returned if this limit is exceeded.

(I haven't thought of a good idea about what to do.)
Contributor guide
Research direction
Start with packages/web/src/pages/ChatPage.tsx at the referenced line and compare its 20-image check with the Converse API behavior described in the issue. Review the linked Bedrock conversation-inference documentation and the reported Claude, Pixtral, Llama 3.2, and Llama 4 limits. Done means the chat behavior and displayed errors accurately reflect the applicable model limits.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- aws, react, typescript
- Domain
- ai, backend-api-design, frontend
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 38/100