BiChat Module

The BiChat module provides enterprise-grade AI chatbot capabilities for multi-tenant applications. It enables users to create intelligent dialogues with LLM-powered responses, maintain conversation history, and seamlessly integrate AI assistance into business workflows.

Overview

BiChat enables:

Multi-turn Dialogues: Maintain conversation context across multiple exchanges
LLM Integration: Support for OpenAI and extensible provider architecture
User-centric Design: Conversations scoped to individual users within tenants
Message History: Complete conversation history with timestamps
Domain-driven Architecture: Clean separation between domain logic and infrastructure

Module Location

modules/bichat/
├── domain/
│   ├── entities/
│   │   ├── dialogue/          # Dialogue conversation entity
│   │   ├── llm/               # LLM abstraction layer
│   │   ├── embedding/         # Embedding vectors
│   │   └── prompt/            # Prompt management
│   └── repositories/          # Repository interfaces
├── infrastructure/
│   ├── persistence/           # Database layer
│   ├── llmproviders/          # LLM provider implementations
│   └── cache/                 # Caching layer
├── services/
│   ├── dialogue_service.go    # Conversation management
│   ├── embeddings_service.go  # Vector embeddings
│   └── prompt_service.go      # Prompt templates
├── presentation/
│   ├── controllers/           # HTTP handlers
│   ├── templates/             # UI templates
│   └── dtos/                  # Data transfer objects
└── module.go                  # Module registration

Key Features

Dialogue Management

Create new conversations with initial user message
Add and retrieve messages within conversations
Set conversation labels for organization
User-scoped access (tenant isolation enforced)

LLM Provider Integration

OpenAI: GPT-4, GPT-4 Turbo, GPT-3.5 Turbo support
Extensible Architecture: Abstract provider interface for easy addition of other LLMs
Streaming Responses: Real-time response streaming for better UX
Configurable Parameters: Temperature, max_tokens, system prompts

Message Management

User and assistant message types
Message timestamps and ordering
Support for multi-modal content
Complete conversation history

Document Map

This BiChat documentation includes:

Business Requirements - Use cases, problem statement, and business rules
Technical Architecture - Module structure, services, repositories, and LLM integration

Integration Points

BiChat integrates with:

Core Module: User authentication and tenant isolation
Database: PostgreSQL for persistence
LLM Services: OpenAI API for chat completions
Web Frontend: HTMX-powered UI for interactions

Architecture Highlights

Domain-Driven Design

Dialogue aggregate with immutable message handling
LLM abstraction decoupled from infrastructure
Repository interfaces for persistence
Clean domain logic with no external dependencies

Multi-tenant Isolation

All queries filtered by tenant_id
User-scoped conversations
Database-level isolation enforced
Secure access patterns via composables

Streaming Architecture

Real-time response streaming via OpenAI API
Server-sent events (SSE) for frontend updates
Progressive message rendering
No blocking on long-running LLM calls

Getting Started

Review Business Requirements to understand the domain
Check Technical Architecture for implementation details
Explore the module code at modules/bichat/

Security Considerations

All dialogues scoped to authenticated users
Tenant isolation enforced at repository layer
Secure API credential management
Input validation on all user messages
Rate limiting recommended for LLM API calls

Core Module - Authentication and user management
Website Module - Public-facing AI chat functionality