Humile - Translator Agent

Overview

Humile is the agent responsible for translation and response normalization in the TKM AI Agency Platform. It ensures consistent communication across different languages by handling message translation, language detection, and response formatting for all other agents in the system.

Directory Structure

Backend/CRM/Humile/
├── data/                # Response logs and data storage
├── humile.py           # Main agent implementation
├── api_humile.py       # FastAPI endpoints
├── tools.py            # Chat and translation utilities
├── tools_schema.py     # Data models and schemas
└── tools_definitions.py # Constants and definitions

Main Components

HumileAgent Class

The main class that handles chat interface operations:

• LLM provider integration (Groq)
• Message translation
• Response normalization
• User configuration management

Processing Pipeline

1. Message reception
2. Source agent identification
3. Language detection/configuration
4. Response processing/translation
5. Message normalization
6. Delivery to conversation

Key Features

Message Processing

• Multi-language support
• Response normalization
• Source agent handling
• User preferences

Translation Capabilities

• Language detection
• Target language configuration
• Translation services
• Fallback handling

Configuration Management

• User language preferences
• Timezone settings
• LLM configurations
• Agent-specific settings

API Operations

Normalized Response Request

# Request Format
{
    "conversation_id": str,
    "user_id": str,
    "organization_id": str,
    "normalized_response": bool,
    "data": dict
}
 
# Response Format
{
    "success": bool,
    "data": dict,
    "metadata": {
        "processed_at": str,
        "source_agent": str,
        "target_language": str
    }
}

Integration

LLM Integration

• Groq API integration
• Model configuration
• Temperature settings
• Token management

Agent Communication

• Atta: Message storage and retrieval
• Prometheus: System responses
• Carina: Search responses
• Scalaris/Hova/Orion: Normalized responses

Performance Features

Response Optimization

• Asynchronous processing
• Timeout handling
• Response caching
• Error recovery

Configuration Caching

• User preferences cache
• Language settings
• Timezone information
• Performance optimization

Error Handling

• Translation errors
• LLM failures
• Configuration issues
• Timeout management
• Comprehensive logging

Data Models

Message Format

{
    "timestamp": str,
    "source_agent": str,
    "type": str,
    "content": {
        "original_input": str,
        "processed_content": dict
    },
    "tokens": dict
}

User Configuration

{
    "user": {
        "default_language": str,
        "timezone": str
    },
    "llm": {
        "provider": str,
        "model": str,
        "temperature": float,
        "max_tokens": int
    }
}

Language Support

• Default language configuration
• User-specific language settings
• Automatic language detection
• Translation quality control

Future Enhancements

• Additional LLM providers
• Enhanced translation capabilities
• Advanced response formatting
• Real-time translation features