RAG Knowledge Base

RAG (Retrieval-Augmented Generation) allows the AI to search a local knowledge base for relevant training data before generating responses.

What Is RAG?

RAG works like this:

1. You ask the AI a question.
2. RTILA X searches the local knowledge base for relevant training examples.
3. The relevant examples are included in the AI’s context.
4. The AI generates a more accurate response based on the examples.

When RAG Is Enabled

RAG is automatically enabled when:

• Using OpenRouter as the AI provider
• Using a Custom Provider
• Using a Custom Local Model (not the built-in RTILA Lite model)

Note: The built-in RTILA Lite 1.5 Assistant model is already fine-tuned for web automation and does not need RAG.

Configuration

Setting	Description	Default
Number of Results (Top K)	How many training examples to retrieve	14
Similarity Threshold	How strict the matching should be (0–1). Higher = stricter.	0.2

Server Status

The RAG settings show the status of the local Qdrant vector database:

Status	Description
🟢 Server Running	The vector database is active and ready
🟡 Starting…	The server is starting up
🔴 Server Stopped	The server is not running

The server starts automatically when RAG is enabled. You can also open the Qdrant Data Folder using the folder icon next to the settings.

How It Improves Responses

Without RAG:

“Generate a scraper for Amazon.” → AI generates a generic configuration.

With RAG:

“Generate a scraper for Amazon.” → AI retrieves training examples of Amazon scrapers and generates a more accurate, specific configuration.

Managing the Knowledge Base

The knowledge base is stored locally in the Qdrant vector database. You can:

• View the data folder by clicking the folder icon in RAG settings.
• Disable RAG by switching to the built-in RTILA Lite model.
• Adjust Top K and Threshold to control how much context is retrieved.