API assistant VS GPTs: what's the main difference?

I asked a custom GPT if its performing vector retrieval. This was the response:

No, I don’t perform vector retrieval on the documents you upload. Instead, I use a different approach to handle and analyze the documents. Here’s a brief overview of how I work with your uploaded documents:

  1. Document Upload and Storage: When you upload documents, they are stored securely in a way that I can access them during our conversation.

  2. File Browsing Tool: I use a specialized tool called myfiles_browser to interact with your documents. This tool allows me to search for specific terms within the documents, navigate through the document, and quote relevant sections to provide context to my responses.

  3. Manual Analysis: I don’t automatically process or analyze the entire content of the documents you upload. Instead, I rely on your queries to guide me. When you ask a question or request information, I use the myfiles_browser tool to search for relevant information in the documents.

  4. Contextual Understanding: My responses are based on the context of your queries and the information contained in the documents. I use the information from the documents, combined with my trained knowledge base, to provide accurate and relevant answers.

  5. Limitations: I adhere to the information contained in the documents and my training. I don’t speculate or provide information beyond what’s available in the documents and my training scope.

In summary, I manually search and interpret the content of your uploaded documents based on your specific questions and use this information, along with my baseline knowledge, to provide accurate and relevant responses.

2 Likes

It would be interesting to know how is performed this search: i don’t think is a simple full-text search :slight_smile:

Is it that end user’s context being captured in GPT through OAuth and sent to Actions easily? I did not saw similar capability with Assistant, without handling it at code level?