AI Assistant

AI Assistant is the unified interface for AI application-user interaction, implementing intelligent routing and multi-Agent collaboration based on LangGraph architecture. It is responsible for routing decisions, intelligent dialogue, and workflow control, supporting visual orchestration, complex business logic, and one-click integration capabilities.

The AI Assistant element hierarchical structure is Meta (aiassistants.Meta) → Type (aiassistants.NormalType) → Instance. Developers can quickly create AI Assistant instance elements through JitAI's visual development tools.

Of course, developers can also create their own Type elements, or override the official aiassistants.NormalType element provided by JitAI in their own App to implement their own encapsulation.

Quick Start

Creating Instance Elements

Directory Structure

Recommended Directory Structure

aiassistants/
├── myassistant/
│   ├── e.json          # Element definition file
│   └── config.json     # Business configuration file

e.json File

aiassistants/myassistant/e.json

{
  "backendBundleEntry": ".",
  "frontBundleEntry": "./config.json",
  "icon": "AIzhushou1",
  "title": "Intelligent Customer Service",
  "type": "aiassistants.NormalType",
  "extendType": "self"
}

Business Configuration File

aiassistants/myassistant/config.json

{
  "apiKey": "{{llmApiKey}}",
  "baseUrl": "{{llmBaseUrl}}",
  "model": "{{llmName}}",
  "temperature": 0.7,
  "saverDb": "databases.Default",
  "sessionTimeout": 3600,
  "nodes": {
    "start": {
      "id": "start",
      "type": "start",
      "title": "Start",
      "position": {"x": 90, "y": 100}
    },
    "aiagent_001": {
      "id": "aiagent_001",
      "type": "aiagent",
      "title": "Intelligent Customer Service",
      "agent": "aiagents.CustomerService",
      "position": {"x": 680, "y": 100}
    }
  },
  "edges": {
    "start-right@aiagent_001-left": {
      "id": "start-right@aiagent_001-left",
      "sourceNode": "start",
      "targetNode": "aiagent_001",
      "sourcePort": "start-right",
      "targetPort": "aiagent_001-left",
      "argMapping": {
        "input_data": "start.userInput"
      }
    }
  },
  "welcomeMessage": "Happy to serve you! How can I help you?",
  "prologues": [
    "What is the element mechanism?",
    "Does JitAI App support private deployment?"
  ]
}

Usage Example

Using AI Assistant

# Get AI Assistant instance
assistant = app.getElement("aiassistants.myassistant")

# Start conversation
result = assistant.run(
    message="Hello, I want to learn about product features",
    chatId="chat_001",
    streamMode=True
)

# Handle streaming response
for chunk in result:
    print(chunk)

Element Configuration

e.json Configuration

Field	Description	Required
backendBundleEntry	Backend entry, fixed as "."	Yes
frontBundleEntry	Frontend entry, points to config.json	Yes
type	Element type, fixed as "aiassistants.NormalType"	Yes
title	Assistant display name	Yes
icon	Icon identifier	No
extendType	Extension type, generally "self"	No

config.json Configuration

Basic Configuration

Configuration	Type	Native Type	Required	Description
apiKey	Stext	str	Yes	LLM API key
baseUrl	Stext	str	Yes	LLM service address
model	Stext	str	Yes	Model name
temperature	Numeric	float	No	Temperature parameter (0-1)
maxTokens	Numeric	int	No	Maximum token count
saverDb	Stext	str	Yes	Checkpoint database
sessionTimeout	Numeric	int	No	Session timeout (seconds)
welcomeMessage	Stext	str	No	Welcome message
prologues	JitList	list	No	Preset question list

Node Configuration

Supported Node Types:

• start: Start node, workflow entry point
• router: Router node, implements intelligent routing
• aiagent: AI agent node, executes specific tasks
• humanaction: Human intervention node
• end: End node, workflow exit point

Edge Configuration

Configuration	Type	Native Type	Required	Description
id	Stext	str	Yes	Unique identifier for the edge
sourceNode	Stext	str	Yes	Source node ID
targetNode	Stext	str	Yes	Target node ID
sourcePort	Stext	str	Yes	Source port
targetPort	Stext	str	Yes	Target port
argMapping	JitDict	dict	No	Parameter mapping relationship

Methods

run

Initiate AI Assistant conversation, supporting both streaming and non-streaming modes.

Parameter Details

Parameter	Type	Native Type	Required	Description
message	Ltext	str	Yes	User input message
chatId	Stext	str	No	Session ID, auto-generated if not provided
attachments	File	list	No	Attachment list
caller	Stext	str	No	Caller identifier
streamMode	Boolean	bool	No	Whether to use streaming mode, default True

Return Value

• Streaming Mode: Returns generator object, outputs processing results step by step
• Non-streaming Mode: Returns final processing result dictionary

Usage Example

Streaming Conversation Example

assistant = app.getElement("aiassistants.myassistant")

# Streaming processing
for chunk in assistant.run(
    message="Help me analyze this problem",
    chatId="chat_001",
    streamMode=True
):
    print(f"Output: {chunk.get('content', '')}")
    if chunk.get('finished'):
        print("Conversation completed")
        break

Non-streaming Conversation Example

# Non-streaming processing
result = assistant.run(
    message="Simple query question",
    streamMode=False
)
print(f"Final result: {result}")

resume

Resume interrupted workflow execution, used to continue processing after human intervention.

Parameter Details

Parameter	Type	Native Type	Required	Description
chatId	Stext	str	Yes	Session ID
nodeId	Stext	str	Yes	Resume node ID
interruptResume	-	dict	No	Resume data

Return Value

Returns generator object, outputs resumed processing results step by step.

Usage Example

Resume Workflow Example

# Resume execution of specified node
for chunk in assistant.resume(
    chatId="chat_001",
    nodeId="aiagent_001",
    interruptResume={"user_confirm": True}
):
    print(chunk)

terminate

Terminate workflow execution for the specified session.

Parameter Details

Parameter	Type	Native Type	Required	Description
chatId	Stext	str	Yes	Session ID to terminate

Return Value

Returns boolean value indicating whether the termination operation was successful.

Usage Example

Terminate Session Example

success = assistant.terminate(chatId="chat_001")
if success:
    print("Session terminated successfully")

getChatStatus

Get the current status of the specified session.

Parameter Details

Parameter	Type	Native Type	Required	Description
chatId	Stext	str	Yes	Session ID

Return Value

Returns string indicating session status:

• "running": Running
• "terminated": Terminated
• "completed": Completed
• None: Session does not exist

Usage Example

Query Session Status Example

status = assistant.getChatStatus(chatId="chat_001")
print(f"Session status: {status}")

Properties

None

AI Assistant element does not currently provide public property access interfaces.

Advanced Features

Streaming Processing

AI Assistant supports real-time streaming output, suitable for long conversations and complex reasoning scenarios.

Session Management

Built-in session state management, supports session resumption, termination, and status queries.

Workflow Control

Implements complex workflow orchestration based on LangGraph, supporting conditional branches, parallel processing, and human intervention.