sciagent/be0/CHAT_ASSISTANT_README.md

Chat Assistant Module

Overview

The Chat Assistant module provides a conversational AI interface for answering policy and compliance questions using Ollama.

Architecture

Backend (be0/src/chat_assistant.py)

The ChatAssistant class provides:

• Chat functionality: Conversational AI for policy questions
• Content verification: Verify content against compliance requirements
• Policy Q&A: Answer questions about policies and compliance

Frontend (fe0/src/features/chat/)

The frontend chat feature includes:

• Service layer: API communication with backend
• React hooks: Easy-to-use hooks for chat functionality
• Type definitions: TypeScript types for type safety

API Endpoints

1. Chat Endpoint

POST /api/v1/chat

Request Body:

{
  "message": "What are ISO 27001 requirements?",
  "conversation_history": [
    {
      "role": "user",
      "content": "Previous message"
    },
    {
      "role": "assistant",
      "content": "Previous response"
    }
  ],
  "context": "Optional context about policies"
}

Response:

{
  "message": "ISO 27001 is an information security management system...",
  "model": "gemma3:27b",
  "tokens_used": 150
}

2. Verify Content Endpoint

POST /api/v1/chat/verify

Form Data:

• field_name: Name of the field being verified
• content: Content to verify
• verification_criteria: (Optional) Specific criteria to check

Response:

{
  "message": "The content meets compliance requirements...",
  "model": "gemma3:27b",
  "tokens_used": 200
}

3. Policy Question Endpoint

POST /api/v1/chat/question

Form Data:

• question: The user's question
• policy_context: (Optional) Context about specific policies

Response:

{
  "message": "Answer to the policy question...",
  "model": "gemma3:27b",
  "tokens_used": 180
}

Features

1. Conversational Context

• Maintains conversation history for context-aware responses
• Keeps last 10 messages for context
• System prompt guides the assistant's behavior

2. Policy Expertise

• Specialized in IT governance and compliance
• Knowledgeable about ISO 27001, NIST, GDPR, etc.
• Provides accurate, actionable advice

3. Content Verification

• Analyzes content against compliance requirements
• Provides detailed feedback
• Suggests improvements

Usage

Backend

from src.chat_assistant import get_chat_assistant

# Get chat assistant instance
assistant = get_chat_assistant()

# Chat
request = ChatRequest(
    message="What is ISO 27001?",
    context="IT governance"
)
response = await assistant.chat(request)

# Verify content
response = await assistant.verify_content(
    field_name="Project Description",
    content="Our project implements security controls..."
)

Frontend

import { useChat } from '@/features/chat/hooks/useChat';

const { sendMessage, verifyContent, isLoading } = useChat();

// Send a message
const response = await sendMessage(
  "What are compliance requirements?",
  conversationHistory, // Optional
  "ISO 27001 context"   // Optional
);

// Verify content
const verification = await verifyContent(
  "Project Name",
  "Project content to verify"
);

Configuration

Model Selection

The default model is gemma3:27b. To change it:

# In chat_assistant.py
assistant = ChatAssistant(model_name="your-model-name")

System Prompt

The system prompt can be customized in the ChatAssistant.__init__ method to change the assistant's behavior and expertise.

Logging

All chat interactions are logged to:

• be0/logs/ChatAssistant.log

This helps with debugging and monitoring.

Error Handling

The module includes comprehensive error handling:

• Catches and logs all exceptions
• Returns user-friendly error messages
• Raises HTTPException for API errors

Testing

To test the chat assistant:

1. Start the backend:

   cd be0
   docker-compose up be0
   

2. Test via API:

   curl -X POST http://localhost:4402/api/v1/chat \
     -H "Content-Type: application/json" \
     -d '{"message": "What is ISO 27001?"}'
   

3. Test via Frontend:

   • Open the Dashboard
   • Use the ChatAssistant component
   • Ask questions or verify content

Integration

The ChatAssistant is integrated with:

• ChatAssistant.tsx: React component in the Dashboard
• useChat hook: React hook for chat functionality
• chatService: API service layer

Future Enhancements

Potential improvements:

1. Streaming responses for real-time text generation
2. Multi-turn conversation management
3. Document context injection
4. Voice input/output
5. Response rating and feedback
6. Conversation export
7. Custom model fine-tuning