---
title: AI Vision Processor
slug: litmusedge/ai-vision-processor
docTags: 
createdAt: 2025-04-01T17:42:45.572Z
---

The AI Vision Processor sends an image through an Analytics flow to an AI model and returns the model's response. Use it when a flow needs to interpret an image, for example a photo captured from an industrial camera or an image included in an incoming payload.

## Why use the AI Vision Processor

Use the AI Vision Processor when the data you want to interpret is an image rather than text or numeric values, for example inspecting a captured photo for a defect or reading a value off a physical gauge.

The processor's parameters let you tune the model's behavior for your use case:

- Set a **System Prompt** to enforce a consistent tone, persona, or domain expertise across every response.
- Adjust **Temperature** to trade consistency for creativity or the reverse.

## Parameters

:::hint{type="info"}
**Note:** File Path always takes priority over a base64 image in the incoming payload. To process an image from the payload, leave File Path blank.
:::

| **Parameter**         | **Details**                                                                                                                                                                                                                          |
| --------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| Name                  | The unique identifier for this instance of the AI Vision Processor.                                                                                                                                                                  |
| AI Model              | The AI model used to generate responses.                                                                                                                                                                                             |
| HTTP Timeout          | The maximum time (in seconds) to wait for a response from the model before the request times out and is ignored.                                                                                                                     |
| File Path For Image   | The path to the image file to process. Accepts `jpg`, `png`, and `gif` files. If left blank, the model expects a base64-encoded image from the input payload instead. If set, File Path always supersedes an incoming payload image. |
| System Prompt         | A predefined instruction that shapes the model's tone, persona, or domain expertise for every response.                                                                                                                              |
| Temperature           | Controls the randomness of responses. Lower values (for example, `0.2`) produce more focused, deterministic output. Higher values (for example, `1.8`) produce more varied, creative output. Default: `1.0`.                         |
| Max Completion Tokens | Limits the number of tokens the model can generate in a single response. If not set, the model's default applies.                                                                                                                    |
| Timer Interval        | Sets a timer (in milliseconds) that continuously publishes the output when no event triggers a response. Enter `0` to disable the timer.                                                                                             |
| Pass Through Value    | Determines whether the processor's output is the model's response or the original unchanged input value.                                                                                                                             |

## Related topics

- [Analytics Flows and Processors](docId\:xuoHHgbpBUH0NPTI1xtwx) 
- [AI Models](docId\:wPy1bCI1y2GXJTZfaVDm9)