Bitrix24 AI-Powered RAG Chatbot for Open Line Channels

Description

Overview
Transform your Bitrix24 Open Line channels with an intelligent chatbot that leverages Retrieval-Augmented Generation (RAG) technology to provide accurate, document-based responses to customer inquiries in real-time.

Use Case
This workflow is designed for organizations that want to enhance their customer support capabilities in Bitrix24 by providing automated, knowledge-based responses to customer inquiries. It's particularly useful for:

Customer service teams handling repetitive questions
Support departments with extensive documentation
Sales teams needing quick access to product information
Organizations looking to provide 24/7 customer support

What This Workflow Does

Smart Document Processing
Automatically processes uploaded PDF documents
Splits documents into manageable chunks
Generates vector embeddings for semantic understanding
Indexes content for efficient retrieval

AI-Powered Responses
Utilizes Google Gemini AI to generate natural language responses
Constructs answers based on relevant document content
Maintains conversation context for coherent interactions
Provides fallback responses when information is not available

Vector Database Integration
Stores document embeddings in Qdrant vector database
Enables semantic search beyond simple keyword matching
Retrieves the most relevant information for each query
Maintains a persistent knowledge base that grows over time

Webhook Handler
Processes incoming messages from Bitrix24 Open Line channels
Handles authentication and security validation
Routes different types of events to appropriate handlers
Manages session and conversation state

Event Routing
Intelligently routes different event types:
ONIMBOTMESSAGEADD: Processes new user messages
ONIMBOTJOINCHAT: Handles bot joining a conversation
ONAPPINSTALL: Manages application installation
ONIMBOTDELETE: Handles bot deletion

Document Management
Organizes processed documents in designated folders
Tracks document processing status
Moves indexed documents to appropriate locations
Maintains document metadata for reference

Interactive Menu
Provides menu-based options for common user requests
Customizable menu items and responses
Easy navigation for users seeking specific information
Fallback to operator option when needed

Technical Architecture

Components
Webhook Handler: Receives and validates incoming requests from Bitrix24
Credential Manager: Securely manages authentication tokens and API keys
Event Router: Directs events to appropriate processing functions
Document Processor: Handles document loading, chunking, and embedding
Vector Store: Qdrant database for storing and retrieving document embeddings
Retrieval System: Searches for relevant document chunks based on user queries
LLM Integration: Google Gemini model for generating natural language responses
Response Manager: Formats and sends responses back to Bitrix24

Integration Points
Bitrix24 API**: For bot registration, message handling, and user interaction
Ollama API**: For generating document embeddings
Qdrant API**: For vector storage and retrieval
Google Gemini API**: For AI-powered response generation

Setup Instructions

Prerequisites
Active Bitrix24 account with Open Line channels enabled
Access to n8n workflow system
Ollama API credentials
Qdrant vector database access
Google Gemini API key

Configuration Steps

Initial Setup
Import the workflow into your n8n instance
Configure credentials for all services
Set up webhook endpoints

Bitrix24 Configuration
Create a new Bitrix24 application
Configure webhook URLs
Set appropriate permissions
Install the application to your Bitrix24 account

Document Storage
Create a designated folder in Bitrix24 for knowledge base documents
Configure folder paths in the workflow settings
Upload initial documents to be processed

Bot Configuration
Customize bot name, avatar, and description
Configure welcome messages and menu options
Set up fallback responses

Testing
Verify successful installation
Test document processing pipeline
Send test queries to evaluate response qu