File Indexing and Local RAG

ChatFrame's powerful Local RAG (Retrieval-Augmented Generation) feature allows you to securely chat with your own documents and code. This process relies on local file indexing.

How Local RAG Works

  1. File Input: You add your local files (PDFs, text files, code files, etc.) to a project or a dedicated indexing area within ChatFrame.
  2. Local Processing: ChatFrame performs all necessary file parsing and vector indexing on your local machine. No data ever leaves your computer.
  3. Vector Store: The indexed data is converted into numerical vectors and stored in a local vector database.
  4. Context Retrieval: When you ask a question in the chat, ChatFrame first searches this local vector store for relevant snippets from your files.
  5. Augmented Generation: These snippets are then combined with your prompt and sent to the selected Large Language Model (LLM) as context, enabling the AI to generate a response that is grounded in your private data.

Supported File Types

ChatFrame is designed to handle a variety of file types, including:

  • Documents: PDF, TXT, Markdown (.md)
  • Code: Various programming language files (e.g., .js, .py, .java, .ts)

Privacy and Security

The local nature of the file indexing and RAG process is a core security feature:

"All file parsing, vector indexing, and RAG operations happen locally on your computer—no data ever leaves your machine." [1]

This ensures that your proprietary code, private documents, and sensitive information are never exposed to third-party services or stored in the cloud.