Chat

The main API endpoint that powers your AI chatbots with intelligent responses, conversation memory, and dynamic knowledge base integration.

What it is

The Chat workflow provides:

• Intelligent AI responses formatted for chatbot interactions with text, mood, and animation
• Conversation memory that maintains context across chat sessions
• Knowledge base integration for accurate, contextual answers
• Scenario-based customization with different personalities and behaviors
• Rate limiting protection to manage API usage
• Structured output optimized for frontend chatbot displays

How to use it

API Endpoint

Send POST requests to your webhook URL with:

Copy
{
  "message": "Hello, how can you help me?",
  "scenario": "customer-service",
  "sessionId": "user-123-session-456"
}

Response Format

Receive structured responses ready for your chatbot interface:

Copy
{
  "messages": [
    {
      "text": "Hello! I'm here to help you.",
      "animation": "Talking",
      "mood": "Happy"
    },
    {
      "text": "What can I assist you with today?",
      "animation": "Listening",
      "mood": "Curious"
    }
  ]
}

How it works

1. Rate Limiting - Checks if the user has exceeded usage limits
2. Scenario Loading - Retrieves personality, prompts, and behavior rules for the selected scenario
3. Knowledge Retrieval - Searches the vector database for relevant information
4. AI Processing - Generates contextual responses using conversation memory
5. Response Formatting - Structures output with text, mood, and animation data
6. Memory Storage - Saves conversation context for future interactions

Key Features

Conversation Memory

• Redis-powered persistence maintains chat history across sessions
• 30-day retention keeps conversations accessible for a month
• Session-based isolation ensures user privacy and context separation

Knowledge Base Integration

• Vector search finds relevant information from your knowledge bases
• Automatic retrieval pulls context when users ask questions
• Multi-collection support different scenarios can use different knowledge bases

Intelligent Response Generation

• Multi-message responses breaks long answers into digestible parts
• Mood detection adds emotional context to responses
• Animation guidance suggests visual behaviors for avatar-based chatbots
• JSON structure ready for frontend consumption

Error Handling

• Rate limit responses (HTTP 429) when usage limits exceeded
• Scenario validation (HTTP 404) when invalid scenarios requested
• Graceful fallbacks for missing data or API failures

Customization Options

Language Model

The workflow uses OpenAI GPT-4.1-mini by default, but you can replace it with:

• Claude (Anthropic)
• Gemini (Google)
• Ollama (Open-source local models)
• Grok (xAI)
• Other GPT models (OpenAI)

Response Structure

Customize the output format by modifying:

• Message structure - Add custom fields beyond text/mood/animation
• Response length - Adjust sentence splitting behavior
• Mood categories - Define custom emotional states
• Animation types - Create scenario-specific animations

Memory Configuration

• TTL settings - Adjust conversation retention period
• Session keys - Customize user identification logic
• Memory scope - Control what information persists

Integration Points

• Rate Limiter - Controls API usage and prevents abuse
• Get Scenario Data - Loads chatbot personalities and configurations
• Knowledge Bases - Generated by the Generate Knowledge Base workflow

Best Practices

• Session Management - Use consistent sessionId formats for reliable memory
• Scenario Organization - Create focused scenarios for different use cases
• Knowledge Base Maintenance - Keep vector databases updated with fresh content
• Response Testing - Verify mood and animation mappings match your frontend
• Memory Cleanup - Monitor Redis usage and adjust TTL as needed

This workflow serves as the core intelligence layer for your chatbot applications, handling everything from user input processing to contextual response generation.