This document describes the image analysis driver plugins and how they integrate with the perception server.

OpenAI image analysis driver

Plugin type:

fp_perception::OpenAIImageAnalysisDriver

This driver is REST-backed and uses the OpenAI Responses API to analyze an image and return text.

How requests are formed

The driver sends a request to https://api.openai.com/v1/responses with:

• an input_text containing the user prompt
• an input_image containing a base64-encoded data URL (for example data:image/png;base64,...)

The server passes both the image and the prompt to the driver.

Parameters

The driver reads parameters under driver.image_analysis.OpenAIImageAnalysisDriver.*:

• name (string): informational name
• model (string): model used for image analysis (current default gpt-5-mini)
• detail (string): image detail level (auto, low, high)
• test_file_path (string): image path used by the driver test() function
• test_prompt (string): prompt used by the driver test() function

REST settings (provided by RestBase):

• driver.image_analysis.OpenAIImageAnalysisDriver.rest.uri (string): should be https://api.openai.com/v1/responses
• driver.image_analysis.OpenAIImageAnalysisDriver.rest.method (string): POST
• driver.image_analysis.OpenAIImageAnalysisDriver.rest.ssl_verify (bool)
• driver.image_analysis.OpenAIImageAnalysisDriver.rest.auth_type (string): Bearer

Environment variables:

• OPENAI_API_KEY: required for authorization

Output

The driver returns a single string containing the model output text.

If the OpenAI response includes an error payload, the driver surfaces the error message as the returned string.