I want to integrate the knowledge of the employees information of the company into my Bot’s API. The idea is that people can ask ‘who is person X,’ and ChatGPT will provide the answer based on this information.
Is there a way to do this without filling the prompt with the information of each employee?
I’m considering some kind of integration with a database or something similar, but I haven’t found content on the internet that meets my needs yet.
You can do this with a function call feature now offered via API (not ChatGPT, which is the chat website)
@ description: query company employee database by name for more info
@@ function: employee_database
@@ parameter: employee_name
One must then actually make the API for a database search engine, tolerant of what the bot emits.
Another way to do this is to use embeddings.
- Prepare a list of employees and their descriptions using a spreadsheet. Export the file as a CSV file.
- Use embeddings API to transform the CSV file into vector data. Save the vector data to a database.
- Prepare a front-end where users can submit inquiries. When a user submits an inquiry, send it to the embeddings API.
- Process the result with the saved vector data from your file. Send the result to Chat API to summarize and give a conversational response.
Using function calling is I think the best solution to your use case but using embeddings will probably be the fastest to deploy.