AI Image - ChatGPT Image Analysis

Description

ChatGPT, the innovative AI service from OpenAI, has expanded its capabilities! While it remains a master of responding to your instructions and generating creative content, it can now also analyze images.

Unlock a New World of Insights:

Identify objects and scenes: Upload an image and let ChatGPT tell you what it sees.
Extract information: Need details from a picture? ChatGPT can help you extract text, recognize objects, and understand the context.
Gain deeper understanding: ChatGPT can analyze the image and provide insights beyond the surface level.

API URL

Access to the following URL is needed:

https://api.openai.com/v1/chat/completions

Configuration Parameters

Workflow-Name: ai_imageanalysis

Key (* = required)	Type	Default	Description	Example
ai_imageanalysis_openai_apiKey*	string		The ChatGPT API Key
ai_imageanalysis_openai_apiUrl	string	https://api.openai.com/v1/chat/completions	The ChatGPT API Url
ai_imageanalysis_openai_apiOrganizationId	string		The ChatGPT Organization ID. For more information see: https://platform.openai.com/docs/api-reference/organizations-and-projects-optional
ai_imageanalysis_openai_apiModelName	string	gpt-4-vision-preview	The name of the ChatGPT model to use. At the time writing this manual there only exists "gpt-4-vision-preview"
ai_imageanalysis_openai_apiPrompt*	string		The ChatGPT prompt to use.	What does the image show?
ai_imageanalysis_openai_apiDetail	string	auto	Determines in which resolution the model should receive the image. For more information see: https://platform.openai.com/docs/guides/vision/low-or-high-fidelity-image-understanding	high
ai_imageanalysis_openai_apiMaxTokens	long	300	The maximum number of tokens that can be generated in the chat completion.
ai_imageanalysis_openai_apiTimeout	long	30	The maxmimum number of seconds we wait for ChatGPT sending the result. If the request to ChatGPT takes longer than the value set here then the Workflow will result in a timeout error.
ai_imageanalysis_openai_apiJsonBody	string	${ai_imageanalysis_openai_apiJsonBody_stringTargetAttribute}	The JSON Body send to the ChatGPT API endpoint. For convenience we provide two JSON Bodies which contain different prompts regarding to whether the result should be tags only (like when the target attribute is an enum field) or whether the result should be plain text (like when the target attribute is a string/clob field): Use "${ai_imageanalysis_openai_apiJsonBody_stringTargetAttribute}" for plain-text results Use "${ai_imageanalysis_openai_apiJsonBody_enumTargetAttribute}" for tags-only results
ai_imageanalysis_targetAttribute*	string		The DAM target attribute identifier where the result from ChatGPT should be stored at. Depending on the apiJsonBody (see above) this should be either an enum or a string/clob field. It is recommended to use a language dependent field here. That way the Workflow can write the result to the corresponding language.
ai_imageanalysis_assetFormat	string	original	The format of the image to send to ChatGPT for analysis. It is recommended to use a lower resolution like "previewCoarse" here since it will reduce costs. See also: https://platform.openai.com/docs/guides/vision/calculating-costs Valid values are: original previewOriginal previewCoarse previewThumbnail	previewCoarse
ai_imageanalysis_omnTargetLanguageId	language (long)		The OMN Language which will be used to write to a language dependent target attribute. By default this language will also be used for the ChatGPT result. In case you want to use a different language for ChatGPT, please set the parameter "targetLanguageTag" (see below). This way you can for example generate spanish texts even there is no OMN Language Spanish available on the system. If no language is specified then "de-DE" will be used as language tag
ai_imageanalysis_openai_targetLanguageTag	string		The language tag which will be used for the generated ChatGPT result. By default this is set automatically to the language tag which results from the parameter "omnTargetLanguageId" (see above). But you can overwrite this in case you want the ChatGPT result in a different language.	es-ES
ai_imageanalysis_openai_targetLanguageTagMappings	string		The language tag mappings which will be used to map the OMN Language Tag (resulting from parameter "omnTargetLanguageId") to the language tag ChatGPT will use (the parameter "targetLanguageTag"). By default this is an empty string-map and the OMN Language Tag will simply be mapped 1:1 to the ChatGPT Language Tag. The mappings you provide here will override this behaviour.	de-DE:es-ES,en-EN:en-GB
ai_imageanalysis_configPreset	string	stringTarget	The config preset to use. The available config presets automatically set the "apiJsonBody" to fit an enum or string target attribute. Valid values are: stringTarget enumTarget	enumTarget

Key (* = required)

Type

Default

Description

Example

ai_imageanalysis_openai_apiKey*

string

The ChatGPT API Key

ai_imageanalysis_openai_apiUrl

string

https://api.openai.com/v1/chat/completions

The ChatGPT API Url

ai_imageanalysis_openai_apiOrganizationId

string

The ChatGPT Organization ID. For more information see: https://platform.openai.com/docs/api-reference/organizations-and-projects-optional

ai_imageanalysis_openai_apiModelName

string

gpt-4-vision-preview

The name of the ChatGPT model to use. At the time writing this manual there only exists "gpt-4-vision-preview"

ai_imageanalysis_openai_apiPrompt*

string

The ChatGPT prompt to use.

What does the image show?

ai_imageanalysis_openai_apiDetail

string

auto

Determines in which resolution the model should receive the image. For more information see: https://platform.openai.com/docs/guides/vision/low-or-high-fidelity-image-understanding

high

ai_imageanalysis_openai_apiMaxTokens

long

300

The maximum number of tokens that can be generated in the chat completion.

ai_imageanalysis_openai_apiTimeout

long

The maxmimum number of seconds we wait for ChatGPT sending the result. If the request to ChatGPT takes longer than the value set here then the Workflow will result in a timeout error.

ai_imageanalysis_openai_apiJsonBody

string

${ai_imageanalysis_openai_apiJsonBody_stringTargetAttribute}

The JSON Body send to the ChatGPT API endpoint. For convenience we provide two JSON Bodies which contain different prompts regarding to whether the result should be tags only (like when the target attribute is an enum field) or whether the result should be plain text (like when the target attribute is a string/clob field):

Use "${ai_imageanalysis_openai_apiJsonBody_stringTargetAttribute}" for plain-text results
Use "${ai_imageanalysis_openai_apiJsonBody_enumTargetAttribute}" for tags-only results

ai_imageanalysis_targetAttribute*

string

The DAM target attribute identifier where the result from ChatGPT should be stored at. Depending on the apiJsonBody (see above) this should be either an enum or a string/clob field. It is recommended to use a language dependent field here. That way the Workflow can write the result to the corresponding language.

ai_imageanalysis_assetFormat

string

original

The format of the image to send to ChatGPT for analysis. It is recommended to use a lower resolution like "previewCoarse" here since it will reduce costs.

Valid values are:

original
previewOriginal
previewCoarse
previewThumbnail

previewCoarse

ai_imageanalysis_omnTargetLanguageId

language (long)

The OMN Language which will be used to write to a language dependent target attribute. By default this language will also be used for the ChatGPT result. In case you want to use a different language for ChatGPT, please set the parameter "targetLanguageTag" (see below).

This way you can for example generate spanish texts even there is no OMN Language Spanish available on the system.

If no language is specified then "de-DE" will be used as language tag

ai_imageanalysis_openai_targetLanguageTag

string

The language tag which will be used for the generated ChatGPT result. By default this is set automatically to the language tag which results from the parameter "omnTargetLanguageId" (see above). But you can overwrite this in case you want the ChatGPT result in a different language.

es-ES

ai_imageanalysis_openai_targetLanguageTagMappings

string

The language tag mappings which will be used to map the OMN Language Tag (resulting from parameter "omnTargetLanguageId") to the language tag ChatGPT will use (the parameter "targetLanguageTag").

By default this is an empty string-map and the OMN Language Tag will simply be mapped 1:1 to the ChatGPT Language Tag. The mappings you provide here will override this behaviour.

de-DE:es-ES,en-EN:en-GB

ai_imageanalysis_configPreset

string

stringTarget

The config preset to use. The available config presets automatically set the "apiJsonBody" to fit an enum or string target attribute.

Valid values are: