Attachments

ChatFrame handles attachments in two primary ways: for Local RAG and for Multimodal Input.

1. Attachments for Local RAG

The primary function of attaching files is to enable Local RAG (Retrieval-Augmented Generation).

• Purpose: To allow the LLM to use the content of your private files (PDFs, code, text documents) as context for its responses.
• Process: Files are added to a local index within ChatFrame. The application performs all parsing and vector indexing locally.
• Supported Types: Documents (PDF, TXT, MD) and code files.
• Security: The content of these files is never sent to the LLM provider; only relevant snippets are extracted and sent as part of the context for a specific query.

2. Attachments for Multimodal Input

For models that support multimodal capabilities (e.g., image input), ChatFrame allows you to attach media files directly to your prompt.

• Purpose: To ask the LLM questions about the content of an image or other media.
• Process: The attached media file is sent to the LLM provider along with your text prompt.
• Supported Types: Depends on the specific LLM model selected (typically images like JPEG, PNG).

File Management

All files used for Local RAG are managed within ChatFrame's project or indexing interface, giving you control over which files are available for the AI's context.