XWorker AI Integration Solution Documentation

XWorker is suitable for writing symbiotic systems of human-computer collaboration, that is, systems that can be modified by humans and AI together.

1. Metamodel New Attributes

Attribute Name	Type	Description
ai_needGenerate	Boolean	Marks whether content needs to be generated. When the value is true, the AI generation task is executed
ai_promptContainsVars	Boolean	Controls whether the prompt word contains context variables (such as SWT button event variables in the Groovy script model)
ai_content_attribute	String	Specify the name of the attribute that stores the AI generated results
ai_promptSystem	String	System-level role prompt words, placed at the beginning of the `role=system` message
ai_promptRequirement	String	User requirement description, placed in the middle of the `role=system` message
ai_getPromptFormat	String	Output format requirements, placed at the end of the `role=system` message

2. Metamodel New Methods

Core Control Methods

boolean ai_isNeedGenerate()
Check whether content needs to be generated. By default, read the value of the ai_needGenerate attribute
void ai_generate()
Execute content generation:
1. When ai_isNeedGenerate() returns true, it is triggered
2. By default, ChatService is called (stream mode)
3. The result is passed to ai_onResult(content)processing

Prompt word processing method

String ai_getPrompt()
Generate a complete system prompt word, default splicing:
System prompt + user requirements + format requirements + variable inclusion identifier
String ai_getPromptSystem()
Get role settings, default return ai_promptSystem value
String ai_getPromptRequirement()
Get user requirements, default return ai_promptRequirement value
String ai_getPromptFormat()
Get format requirements, default return ai_getPromptFormat value

Result processing method

void ai_onResult(String content)
Process AI return results:
By default, save to the specified attribute of ai_content_attribute

Service configuration method

ChatService ai_getChatService()
Get the chat service instance, and use the system default service when null is returned
String ai_getContent()
Get the edited content, and return ai_content_attribute specifies the value of the attribute

III. Typical application scenarios

1. AI automatic generation mode

Applicable scenarios: Multi-level content generation (such as novel generation system)
Execution process:

Model tree traversal: deep traversal from the root node
Generation detection: execute ai_isNeedGenerate()
for each node
Content generation: call ai_generate()
for the node to be generated
Loop detection: when generation occurs, wait for completion and then traverse again
Termination condition: End the process when there is no generation demand in the whole tree

2. AI editing mode

Applicable scenarios: Single node content modification
Interaction process:

Editor initialization:
- System prompt = ai_getPrompt()
- Edit content = ai_getContent()
User interaction:
- Enter the modification instruction as the user message
Result processing:
- After the user confirms, call ai_onResult() to save the result

Fourth, Extension Description

This solution achieves deep integration of AI capabilities by extending the metamodel, supporting two modes: fully automatic generation and interactive editing. Developers can flexibly customize AI behaviors through attribute configuration and method rewriting.

Custom Generation: You can achieve non-chat service generation (such as preset code injection) by rewriting ai_generate()
Variable Context: Automatically inject environment variable description when ai_promptContainsVars=true
Service Extension: You can access different AI service providers by implementing ai_getChatService()