# HybridAI Chat Platform

> Multi-tenant, hybrid multi-LLM ai-agent-as-a-service platform with RAG, workflow automation, and advanced AI capabilities.

## About This File

This file follows the llms.txt specification (https://llmstxt.org/) to help AI systems understand this project. It provides structured information about the HybridAI Chat Platform for potential customers, partners, and AI-assisted integration.

## Project Overview

HybridAI Chat Platform is a comprehensive chatbot-as-a-service solution that enables businesses to deploy intelligent AI agents powered by multiple LLM providers. The platform offers enterprise-grade features including retrieval-augmented generation (RAG), workflow automation, and multi-channel messaging.

**Primary Use Cases:**
- Deep AI Integration with Multi-LLM support
- Hybrid AI approach - orchestration of multiple LLM calls 
- Custom trained and fine-tuned AI models
- Small Language Models for Financial Use-Cases like Incoterms, VAT Automation etc.
- Customer support automation with AI-powered chatbots
- Knowledge base integration via RAG (Retrieval-Augmented Generation)
- Multi-channel messaging (Web widget, WhatsApp, Email)
- Workflow automation through n8n integration
- Virtual bot personas for market research and surveys
- Human-in-the-loop takeover for live chat scenarios

## Technology Stack

- **Backend:** Python 3.13+, Flask, Celery
- **Databases:** MySQL, Redis, Weaviate (vector search)
- **Infrastructure:** Docker, GCP, GitHub Actions CI/CD
- **Testing:** pytest, Playwright

## LLM Provider Integration

The platform supports multiple LLM providers through a unified abstraction layer:

### Supported Providers
- **OpenAI:** GPT-4, GPT-4 Turbo, GPT-3.5 Turbo, fine-tuned models
- **Anthropic:** Claude 3.5 Sonnet, Claude 3 Opus, Claude 3 Sonnet, Claude 3 Haiku
- **Google:** Gemini Pro, Gemini Pro Vision
- **Mistral AI:** Mistral Large, Mistral Medium, Mistral Small
- **Grok:** X.AI Grok models
- **Local Models:** vLLM for on-premise deployment

### OpenAI-Compatible API

The platform exposes an OpenAI-compatible API, allowing seamless integration with existing OpenAI client libraries while routing to any supported provider. This enables easy migration and multi-provider strategies.

## Core Features

### 1. Multi-LLM Chat Backend
Intelligent routing to different LLM providers based on model selection, with streaming support, function calling, and multimodal capabilities (images, documents).

### 2. Retrieval-Augmented Generation (RAG)
- Weaviate vector database integration for semantic search
- Automatic document and website crawling
- Contextual responses based on your knowledge base
- Support for PDF, Word, and web content

### 3. Function Calling / Tools
LLM-powered function calls enable bots to:
- Send emails
- Store and retrieve user-specific memory
- Trigger external workflows
- Search web content (Perplexity integration)
- Query WordPress sites
- Execute n8n workflows

### 4. Virtual Bots
AI personas for market research, surveys, and opinion gathering:
- Survey bots for structured data collection
- Opinion bots for qualitative feedback
- Headline generators for content testing
- Evaluation bots for assessment tasks
- A/B testing with multiple personas

### 5. Workflow Automation (n8n)
Native integration with n8n for complex multi-step workflows triggered directly from conversations. Connect your chatbot to hundreds of services.

### 6. Human Takeover
Real-time dashboard for monitoring conversations and enabling seamless human agent intervention when needed.

### 7. Sensing System
Automatic message analysis for:
- Sentiment analysis (positive/negative/neutral)
- Content moderation
- Business intent extraction
- Lead qualification

### 8. Bot Memory
Persistent knowledge storage for individual bots, enabling information retention across conversations and personalized user experiences.

### 9. Multi-Channel Support
- **Web Widget:** Embeddable chat widget for websites
- **WhatsApp:** Business messaging integration
- **Email:** Email-based conversation handling
- **API:** Direct API access for custom integrations

## API Overview

### OpenAI-Compatible API
```
POST /api/openai/v1/chat/completions - Chat completions (streaming supported)
GET  /api/openai/v1/models           - List available models
```

### Bot Memory API
```
POST   /api/bot-memory/{bot_id}/memories      - Store memory
GET    /api/bot-memory/{bot_id}/memories      - List all memories
GET    /api/bot-memory/{bot_id}/memories/{key} - Get specific memory
PUT    /api/bot-memory/{bot_id}/memories/{key} - Update memory
DELETE /api/bot-memory/{bot_id}/memories/{key} - Delete memory
```

### Virtual Bots API
```
POST /api/virtual-bots                  - Create virtual bot persona
GET  /api/virtual-bots                  - List virtual bots
POST /api/virtual-bots/{bot_id}/opinion - Get opinion from virtual bot
POST /api/virtual-bots/ab-test          - A/B test with multiple personas
```

### Chat Endpoints
```
GET  /answer                - Chat completions
POST /api/text-to-speech    - Text-to-speech conversion
POST /api/voice_transcribe  - Voice transcription (speech-to-text)
```

For complete API documentation, see [docs/API.md](docs/API.md).

## Security & Compliance

- SQL injection prevention with parameterized queries
- XSS protection with automatic template escaping
- Secure session management
- Rate limiting per endpoint and user
- Environment-based secret management
- HTTPS enforced in production
- GDPR-compliant data handling

## Internationalization

The platform supports multiple languages:
- English (EN)
- German (DE)
- French (FR)

All user-facing text is translatable via Flask-Babel.

## External Integrations

- **LLM Providers:** OpenAI, Anthropic, Google, Mistral, Grok, vLLM, Ollama
- **Vector Database:** Weaviate
- **Workflow Automation:** n8n
- **Communication:** Brevo (SMS/Email), Twilio
- **CRM:** Brevo, HubSpot
- **Payments:** PayPal, Stripe
- **Authentication:** Google OAuth
- **Monitoring:** Langfuse (LLM observability), Sentry (error tracking)
- **Messaging:** WhatsApp (WHAPI), Discord

## Documentation & API Reference

**Live API Documentation (OpenAPI/Swagger):**
- OpenAI-Compatible API: `https://hybridai.one/v1/openai/openapi.json`
- Bot Memory API: `https://hybridai.one/v1/bot-memory/openapi.json`
- Virtual Bots API: `https://hybridai.one/api/v1/virtual-bots/openapi.json`

**Product Documentation:**
- `https://hybridai.one/documentation` - Full documentation

## License

Proprietary - All rights reserved

## Contact

- **Company:** HybridAI
- **Email:** support@hybridai.one
- **Website:** https://hybridai.one