Image Upload
Functionality
This module makes us of the capabilities of some LLM to answers questions out of an Uploaded Image. This module uses the entirety of the conversation history (up to the model’s token limit). It also support using a summary history instead if correct tool definition is provided and the functionality is enabled from config.
Input
An Image with valid extension + user query. It’s helpful to trigger this module when specifying in the user query that you need to get information from uploaded image.
Example: ”Please describe the content of the uploaded image.”
Output
Answer to the question based on the image and conversation history.
Configuration settings (technical)
Tool definition
{
"type": "function",
"function": {
"name": "ImageUpload",
"parameters": {
"type": "object",
"required": [
"query",
"conversation_history"
],
"properties": {
"query": {
"type": "string",
"description": "Smart query based on the user request but augmented with the context of conversation if needed."
},
"image_name": {
"type": "string",
"description": "Name of the image file to be analyzed [Optional]."
},
"conversation_history": {
"type": "string",
"description": "Summary of the history of the conversation to provide context for the query."
}
}
},
"description": "This tool is designed to analyze, extract, and summarize information from images. It should be invoked whenever the user inquires about images."
}
}
Global structure of config
Parameter | Description |
---|---|
| used GPT model Default: |
| Instance of service |
| List of supported image extensions. Default:
|
chatHistory
Parameter | Description |
---|---|
| Whether to use full conversation history. If False, attempt to use conversation summary from tool call. Default: |
| Max history messages to consider Default: |
| Whether to use model token limit or not Default: |
| Default token limit if model token limit is not available. Default: |
| Upper limit of tokens to keep from history. Default: |
Example Json Config
{
"ImageUploadConfig": {
"chatHistory": {
"enabled": true,
"maxMessages": 10
},
"languageModel": "AZURE_GPT_4o_2024_0513"
}
}
Template Json Config (for developers)
{
"ImageUploadConfig": {
"supportedExtensions" : [
".jpg",
".jpe",
".jpeg",
".png",
".gif",
".bmp"
],
"chatHistory": {
"enabled": true,
"maxMessages": 10,
"useModelTokenLimit": true,
"defaultTokenLimit": 4096,
"percentageOfMaxTokens": 0.5
},
"languageModel": "AZURE_GPT_4o_2024_0513"
}
}
Author | @Rami Azouz |
---|
© 2024 Unique AG. All rights reserved. Privacy Policy – Terms of Service