GLPI Knowledge Base RAG Pipeline with Google Gemini and PostgreSQL

This n8n workflow demonstrates a Retrieval Augmented Generation (RAG) pipeline designed to interact with a GLPI knowledge base using Google Gemini for language understanding and PostgreSQL with PGVector for vector storage. It allows users to ask questions via a chat interface, and the AI agent will retrieve relevant information from the knowledge base to formulate answers.

What it does

This workflow automates the following steps:

1. Listens for Chat Messages: The workflow is triggered when a new chat message is received from a user.
2. Manages Conversation History: It utilizes a simple memory buffer to maintain the context of the ongoing conversation, allowing the AI to understand follow-up questions.
3. Processes User Input: The user's chat message is fed into an AI Agent.
4. Generates Embeddings: Google Gemini's embedding model is used to convert the user's query into a vector representation.
5. Retrieves Relevant Information: The generated embedding is used to query a PostgreSQL database (configured with PGVector) to find the most semantically similar documents or knowledge base articles.
6. Generates Response with Google Gemini: The AI Agent, powered by Google Gemini Chat Model, combines the user's query with the retrieved context from the knowledge base to generate a comprehensive and relevant answer.
7. Responds to the User: The AI agent's generated response is sent back to the user via the chat interface.

Prerequisites/Requirements

To use this workflow, you will need:

• n8n Instance: A running n8n instance.
• Google Gemini API Key: An API key for Google Gemini (for both Embeddings and Chat Model).
• PostgreSQL Database with PGVector Extension: A PostgreSQL database instance with the pgvector extension installed and configured. This database will store your GLPI knowledge base content as vector embeddings.
• GLPI Knowledge Base Content (Pre-embedded): Your GLPI knowledge base articles or relevant text content should already be embedded and stored in your PostgreSQL PGVector database. This workflow focuses on the RAG query part, not the initial embedding of the knowledge base.
• Chat Platform Integration: A chat platform configured with n8n to send and receive messages (e.g., Telegram, Slack, Discord, etc., connected to the Chat Trigger node).

Setup/Usage

1. Import the Workflow: Download the provided JSON and import it into your n8n instance.
2. Configure Credentials:
   • Google Gemini: Create or select an existing Google Gemini credential for both the Embeddings Google Gemini and Google Gemini Chat Model nodes.
   • PostgreSQL: Create or select an existing PostgreSQL credential for the Postgres PGVector Store node, providing the necessary database connection details (host, port, database name, user, password).
3. Configure Chat Trigger: Set up the When chat message received node to connect to your preferred chat platform (e.g., Telegram Bot, Slack, etc.) where users will interact with the RAG pipeline.
4. Configure PGVector Store:
   • Ensure the Postgres PGVector Store node is configured with the correct table name and column names where your GLPI knowledge base embeddings are stored.
5. Activate the Workflow: Once all credentials and configurations are set, activate the workflow.

Now, when a user sends a message to your configured chat platform, the workflow will trigger, process the query, retrieve relevant information from your GLPI knowledge base via PostgreSQL, and respond with an AI-generated answer.