Images and attachments for AI agents
Give agents image context and use supported image or document attachments with private Team Chat agents.
What it is
Supported AI features can use image content as well as text. Image messages retain their captions and references, and the agent can revisit a referenced image with its image tool. Private Team Chat agents also accept document input when the configured provider supports it.
Who uses it and why
- Visitors supply photos to explain what they need.
- Operators share images or supported documents with a Team Chat agent.
- Administrators choose compatible AI accounts and enable image tools where available.
Where to find it
For supported Autopilot features, open Settings › AI › Features and use Enable Image Tool. For a standalone agent, attach files in its Team Chat conversation. Team Chat attachment input is separate from the Autopilot image-tool switch.
How it works
When image analysis is enabled, INSIDE prepares a visitor image message for the AI model and keeps any caption with it. An agent can later ask its image tool to open a referenced image again.
If analysis is disabled or the image cannot be loaded, INSIDE keeps a text description and image reference in the AI context. That description says analysis is unavailable or not enabled; it does not mean that the model inspected the image.
For Team Chat agents, INSIDE checks uploaded attachments and builds input from recent private conversation messages. Image support depends on model capability. Document support depends on the provider. If ingestion is unsupported, the file stays in Team Chat and a private error explains the limitation.
Configure it
With the Settings role and the AI permission, or the Developer role, open the supported feature and enable Enable Image Tool. It starts off. Use a model that can accept images.
For Team Chat documents, ask Powerfront to confirm a compatible account and provider. Supported input paths include OpenAI Responses, Google AI Studio and AWS Bedrock, but the provider still determines which file formats it can read.
Ask Powerfront to change restrictions on external image sources. Do not assume that allowing an upload type also makes it readable by every AI model.
Settings
| Setting | Where | Applies to | Default | Who can change it | What it does |
|---|---|---|---|---|---|
| Enable Image Tool | Settings › AI › Features | website | Off | Settings role, Developer role | Allows supported AI flows to analyse image messages and call the image viewing tool. |
Permissions
| Permission | Lets you |
|---|---|
| Settings | Configure image tools when AI is allowed under Settings Permissions. |
| AI | Allows AI configuration within Settings Permissions. |
| Developer | Configure AI features and their image capabilities. |
| Allow AI Suggestions | Allows an operator with Chat access to use eligible Team Chat agents, including their supported attachment input. |
Limits and good to know
- The general image viewer limits externally downloaded images to 5 MiB and checks that the returned content is an image.
- Team Chat AI input uses at most 60 recent messages and 100,000 text characters.
- Team Chat includes up to 2 MiB per file and 4 MiB of file content in total for the input it prepares. Larger files may remain in chat without being included for the model.
- Team Chat accepts up to ten distinct attachments per message, uploaded by that operator for the same website. Visitor-owned files are not accepted through this path.
- A file visible in the conversation is not proof that the AI received its contents.
- Document attachment support in Team Chat does not establish document input for every visitor-facing AI feature.
Depends on and works with
View all connections as a list
This feature uses
- Roles and permissionsConfigured by
AI configuration and operator permissions control setup and access.
- Attachment file types and virus scanningConfigured by
Allowed attachment types affect which Team Chat uploads are accepted.
- AI feature settings and change historyConfigured by
Image analysis is configured on the supported feature.
- AI run limitsRequires
AI execution prepares image input and supplies image viewing tools.
- AI accounts and shared defaultsConfigured by
The selected account and provider determine supported attachment input.
- AI model catalogueRequires
Image input requires a model that supports it.
- Attachments in conversationsEnhances
Conversation attachments supply uploaded visitor images for AI analysis.
- File attachments for visitorsEnhances
Visitors can supply images through supported attachment controls.
Used by
- INSIDE AiFlow test dashboard · managed by PowerfrontEnhances
Scenarios can include image attachments.
- Ask AI agents in Team ChatEnhances
Supported attachments can provide image and document context.
- AI model provider selectionEnhances
Supported providers can accept image and document context.
- OpenAI API connection optionsEnhances
Supported models can receive image and file context.
- Google AI Studio model connectionEnhances
Requests can include image and file content when supported by the model.
- AWS Bedrock model connectionEnhances
Supported request modes can include image or document content.
FAQ and troubleshooting
Why can the agent see my caption but not describe the photo?
Image analysis may be off, the model may lack image support, or loading may have failed. Captions and references can remain available without visual analysis.
Why did my document remain in Team Chat with an error?
The selected provider may not support document input for that agent. Ask Powerfront to confirm the account's supported input types.
Why does a large attachment seem to be ignored?
Team Chat limits the file content sent to the AI. A file above the per-file limit or beyond the combined input budget can remain in chat without its contents being sent to the model.
Does enabling the image tool let the agent read every document?
No. The image tool handles images. Document input is a separate capability on supported Team Chat agent paths.