AI Content Enrichment Task
Objective 3Sixty supports the use of Artificial Intelligence for processing document in a multitude of ways by using the AI Content Enrichment Task. Q&A, Metadata Extraction, Text Summarization and PII Detection are just a handful of many example use cases for this task. This task uses Chat Completion which is simply the umbrella term that is used to describe Artificial Intelligence features in a chat-like manner.
The Objective AI Content Enrichment Task can be used by any "OpenAI Like" servers and not just OpenAI. That is to say, any server providing an interface similar to that of OpenAI can be used with this task.
Configuration
To use this task go to the task tab in your job. Select the task from the drop down and click the plus circle to configure the task. Click done after making any changes to save.
Condition check
It will execute the task when the condition's result is 'true', 't', 'on', '1', or 'yes' (case-insensitive), or run on all conditions if left empty. This condition is evaluated for each document, determining whether the task should be executed based on the specified values.
Example: If I only want to run this task for PDF documents I would use the expression: equals('#{rd.mimetype}',"application/pdf")
Configuration
To perform content enrichment, the following parameters must be provided:
AI Authentication: Select the API Key Authentication connection that has the Server and API key you want to use for this task.
Completion Path: Change if yours is different from the default one (v1/chat/completions)
Model: If the server requires a model name (in case of servers supporting many models) it must be provided
System prompt: Set the overall behavior of the AI assistant (must be a 3Sixty expression language)
User prompt: Set the specific request to be asked from the AI assistant (must be a 3Sixty expression language)
JSON Format: Checking this box allows the user to request a valid JSON format response. This feature is model dependent, which means not all models will honor it. Depending on the Chat Completions API server and model you are using, you need to set or unset this.
Attach Media: This allows the original file (e.g. image file, description or classification tasks) to be sent with the query and analysed by the multi-modal LLMs.
Timeout: Server timeout in seconds
Result Field: Result field containing AI response to be added to the document
Top P: Optional. Controls diversity by sampling from the top tokens whose cumulative probability is p. Higher values include more possible words, lower values narrow the selection to top choices. Valid values: 0 to 1 Recommended: 0.8
Temperature: Optional. Controls how random or creative the response is. Lower values (0) make output more focused and predictable, higher values (up to 2) make it more diverse and creative. Valid values: 0 to 2 (OpenAI), typically 0.0 to 1.0 for other providers. Recommended: 0.7
Max Tokens: Optional. Specifies the maximum number of tokens (words or word pieces) to generate in the response. Limits how long the output can be. Valid values: Any positive integer. Example: 256, 500, 1000
Stop Sequence: Optional. Defines one or more sequences of characters that will stop the response when encountered. Valid values: Al
API Keys
Processor: openAiChatCompletionTask
|
Key |
Display Name |
Type |
|---|---|---|
| use_condition | Check a condition before executing this task. | Boolean |
|
task_condition |
Condition |
String |
|
task_stop_proc |
Stop Processing |
Boolean |
|
open_ai_server |
Server |
String |
|
open_ai_api_key |
API Key |
String |
|
open_ai_model |
Model |
String |
|
open_ai_system |
System Prompt |
String |
|
open_ai_user |
User Prompt |
String |
|
open_ai_json |
JSON Format |
Boolean |
|
open_ai_timeout |
Timeout |
Integer |
|
open_ai_result |
Result Field |
String |
|
open_ai_completions_path |
Completions Path |
String |
|
open_ai_media |
Attach Media |
Boolean |
|
open_ai_temperature |
Temperature |
Decimal |
|
open_ai_max_tokens |
Max Tokens |
Integer |
|
open_ai_top_p |
Top_P |
Decimal |
|
open_ai_stop_sequences |
Stop Sequences |
String |
Related Articles