Hello friends, I created a GPT model and uploaded a JSON file containing 1,000 records of support tickets. Each record includes the ticket number, the name of the person who initiated the ticket, the supporter assigned to the ticket, the date of the ticket, a description of the problem, and additional details. However, when I ask the GPT to provide all tickets made by a specific customer or to list all tickets associated with a particular supporter, I only receive partial responses. I’ve tried changing the model from 4O to 4O Mini, adjusting the temperature, and modifying the prompt, but none of these changes have resolved the issue.
Any ideas?
Files are not fully placed into the AI context. They are used in semantic search, which is best for knowledge and natural language.
Files have document extraction if needed, and texts are broken into 800 tokens (if similar to the documented API), a similar number of words or less. Then a vector store database using AI embeddings is what powers a search, returning entire ranked chunks based on what has similar language and meaning.
The kind of thing you ask for needs a tool beyond what is provided, such as a SQL server you provide as GPT “actions”, a tool where an AI can put in query fields and your hosted server’s API (that you provide) is connected to and fulfills the data request.
Or just paste the whole file into a prompt.
(P.S. you didn’t make a model. A GPT is just textual instructions the AI uses, that you can see)