Build a ServiceNow Knowledge Chatbot with OpenAI and Qdrant RAG

1. Data Ingestion Workflow (Left Panel – Pink Section)

This part collects data from the ServiceNow Knowledge Article table, processes it into embeddings, and stores it in Qdrant.

Steps:

Trigger: When clicking ‘Execute workflow’

The workflow starts manually when you click Execute workflow in n8n.

Get Many Table Records

Fetches multiple records from the ServiceNow Knowledge Article table. Each record typically contains knowledge article content that needs to be indexed.

Default Data Loader

Takes the fetched data and structures it into a format suitable for text splitting and embedding generation.

Recursive Character Text Splitter

Splits large text (e.g., long knowledge articles) into smaller, manageable chunks for embeddings. This step ensures that each text chunk can be properly processed by the embedding model.

Embeddings OpenAI

Uses OpenAI’s Embeddings API to convert each text chunk into a high-dimensional vector representation. These embeddings are essential for semantic search in the vector database.

Qdrant Vector Store

Stores the generated embeddings along with metadata (e.g., article ID, title) in the Qdrant vector database. This database will later be used for similarity searches during chatbot interactions.

2. RAG Chatbot Workflow (Right Panel – Green Section)

This section powers the Retrieval-Augmented Generation (RAG) chatbot that retrieves relevant information from Qdrant and responds intelligently.

Steps:

Trigger: When chat message received

Starts when a user sends a chat message to the system.

AI Agent

Acts as the orchestrator, combining memory, tools, and LLM reasoning. Connects to the OpenAI Chat Model and Qdrant Vector Store.

OpenAI Chat Model

Processes user messages and generates responses, enriched with context retrieved from Qdrant.

Simple Memory

Stores conversational history or context to ensure continuity in multi-turn conversations.

Qdrant Vector Store1

Performs a similarity search on stored embeddings using the user’s query. Retrieves the most relevant knowledge article chunks for the chatbot.

Embeddings OpenAI

Converts user query into embeddings for vector search in Qdrant.