> ## Documentation Index
> Fetch the complete documentation index at: https://fliqr.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Analyze images with AI Vision

> Use OpenAI, Google Gemini, or Claude Vision in Fliqr AI Flows to analyze images customers send — damage photos, screenshots, receipts, and more.

Vision lets your automation understand images. Collect a photo in a Flow, send it to a vision-capable model, and continue with the analysis stored in a custom field.

## Providers

| Provider             | Typical action label   |
| -------------------- | ---------------------- |
| **OpenAI**           | Analyze image / Vision |
| **Google Gemini**    | Gemini Vision          |
| **Anthropic Claude** | Claude Vision          |

Connect the provider first: [AI Providers](/docs/core-concepts/ai-agents/ai-providers).

## Basic setup in a Flow

<Steps>
  <Step title="Collect the image">
    Add **User Input / Get User Data** and save the image to a custom field.
  </Step>

  <Step title="Add the provider action">
    Insert the OpenAI / Gemini / Claude action block. Choose **analyze image** (wording varies by provider).
  </Step>

  <Step title="Pass the image field">
    Select the custom field that holds the image. Add an optional text prompt (“Describe the defect and severity”).
  </Step>

  <Step title="Save the output">
    Write the model’s description into another custom field.
  </Step>

  <Step title="Continue the Flow">
    Show the analysis, Condition on keywords, create a ticket, or hand over to a human.
  </Step>
</Steps>

## Advanced: analyze unexpected images

You can analyze images even when they arrive outside a dedicated “please upload” step — for example in a Default Reply path or while an AI Agent is handling the thread. Pattern:

1. Detect that the inbound message includes an image
2. Run the Vision action on that media
3. Reply with the summary or route by Condition

## Prompt tips for Vision

* Ask for structured output (`damage: yes/no`, `category`, `confidence`)
* Tell the model what to ignore (background clutter)
* Keep PII rules explicit if images may contain IDs or faces

## Next

<CardGroup cols={2}>
  <Card title="Read documents" icon="file-lines" href="/docs/core-concepts/ai-agents/read-documents">
    Extract text/insights from documents contacts send.
  </Card>

  <Card title="Image generation" icon="image" href="/docs/core-concepts/ai-agents/image-generation">
    Generate or edit images from text prompts.
  </Card>
</CardGroup>