RAG Search

Harnessing AIChat and RAG with a Public GitHub Repository

This guide demonstrates how to leverage AIChat’s built-in vector database and full-text search capabilities to create a powerful knowledge base from an entire public GitHub repository. This enables Retrieval-Augmented Generation (RAG), allowing you to use AIChat to answer questions and provide assistance based on the repository’s content.

Creating a RAG Knowledge Base from a GitHub Repository:

The following steps outline how to create a RAG knowledge base using AIChat, using the ilima-snippets example:

Initialize RAG: Use the .rag command followed by a name for your RAG knowledge base.
```
> .rag ilima-snippets
```
Configure Embedding Model: Select an appropriate embedding model. The example uses gemini:text-embedding-004. Note the model’s token limits, batch size, and pricing (if applicable).
```
> Select embedding model: gemini:text-embedding-004 (max-tokens:2048; max-batch:100; price:0)
```
Set Chunk Size and Overlay: Configure the chunk size and overlay for optimal retrieval. A chunk size of 1500 and an overlay of 100 are used in the example. Experiment with these values to find what works best for your repository.
```
> Set chunk size: 1500
> Set chunk overlay: 100
```
Add Documents: Specify the GitHub repository’s document path using a wildcard to include all files within the specified directory. In this example, all documents within the docs directory of the unapologetic-snippets repository are added.
```
> Add documents: https://github.com/igorlima/unapologetic-snippets/tree/main/docs/**
```
Exit RAG Configuration: Use the .exit rag command to finalize the RAG creation process.
```
@ilima-snippets> .exit rag
```

Managing Your RAG Knowledge Base:

AIChat provides several commands for managing your RAG knowledge base:

View Citation Sources: Use .sources rag to display the sources used in the last query. This helps verify the accuracy and relevance of the information.
```
> .sources rag
```
- Show citation sources used in last query
Edit Documents: Use .edit rag-docs to add or remove documents from the existing RAG knowledge base.
```
> .edit rag-docs
```
Rebuild RAG: Use .rebuild rag to rebuild the knowledge base after making changes to the documents. This ensures that the RAG reflects the latest content.
```
> .rebuild rag
```
Show RAG Information: Use .info rag to display information about the RAG knowledge base, such as the number of documents and the embedding model used.
```
> .info rag
```

Further Resources:

AIChat RAG Guide: For a more in-depth explanation, refer to the official AIChat RAG Guide ⁺.
Leveraging AIChat RAG with Your RC File Configuration Assistance Made Easy
Leveraging AIChat RAG with ILIMA THOUGHTS