Image Analysis in the AI Agent Node - ConvyAI
Overview
The AI Agent node supports attaching images that are analyzed as part of the prompt. The model processes both the text input and the attached images in the same LLM call, enabling visual classification, data extraction from documents or screenshots, and multimodal decision-making within the flow.
Supported formats by provider
Supported formats and size limits vary by the AI model selected in the node.
OpenAI and Azure OpenAI
| Property | Value |
|---|---|
| Formats | PNG, JPEG, WEBP, non-animated GIF |
| Maximum total size | 50 MB |
Google Gemini
| Property | Value |
|---|---|
| Formats | PNG, JPEG, WEBP, HEIC, HEIF |
| Maximum total size | 20 MB |
Amazon Bedrock
| Property | Value |
|---|---|
| Formats | PNG, JPEG, WEBP, non-animated GIF |
| Maximum total size | 3 MB |
note
These limits may change as providers update their APIs. Always verify against the current provider documentation for production use.
Detail option
When attaching images, a detail parameter controls the analysis depth:
| Value | Description |
|---|---|
auto | The model automatically selects the appropriate detail level based on image size and content. |
low | Faster analysis with reduced precision. Suitable for simple classification tasks. |
high | More in-depth analysis. Suitable for extracting fine-grained information from complex images. |