I found that there are two kinds of knowledge base query:
My question is that
What’s the difference between these two? I found that the 1st one would call the python code (if you click the icon), but I have no idea about the 2nd.
When will these two queries be triggered? Can I control these two queries? e.g. via prompts?
I copied and pasted it here, may be useful to others.
Let’s delve a bit deeper into how these two types of queries work in the context of the GPT interacting with uploaded Knowledge files:
Analyzing (Clickable Python Code):
Triggered When: This query type is triggered when your question or command directly relates to the content in the uploaded Knowledge files. GPT recognizes specific references or context that match the content of these files.
Process: GPT actively parses and analyzes the content of the uploaded file. This involves running specific Python code to extract and process information relevant to your query.
User Interaction: You can click on the “Analyzing” icon to see the Python code being executed. This gives you an insight into how GPT is processing the file to answer your question.
Control: While direct control over this process via prompts isn’t possible, framing your questions to be clearly related to the content of the uploaded files can increase the likelihood of triggering this type of query.
Searching my knowledge (Non-Clickable):
Triggered When: This is triggered for more general queries where the answer might be informed by a combination of the GPT’s pre-trained knowledge and the uploaded Knowledge files. It’s used when the query does not directly call for specific information from the uploaded files but could still benefit from them.
Process: GPT scans its internal knowledge base (which includes a vast array of pre-trained information) along with any insights it can draw from the uploaded files. This process is more opaque to the user, as it’s a part of GPT’s internal reasoning and decision-making.
User Interaction: There is no option to view the internal workings in this case. The GPT synthesizes an answer based on its comprehensive understanding.
Control: As with the Analyzing query, direct control isn’t possible. However, general, broader questions are more likely to trigger this type of response.
In summary, while you can’t directly control which query type GPT uses via specific prompts, the nature of your questions can influence it. More specific, targeted questions about the content of the Knowledge files lean towards the “Analyzing” type, while broader, more general queries engage the “Searching my knowledge” approach.
