Can I use my GPT-4 API key to run a 0-shot prompt designed to initiate a consultation with a user?
I’ll need to set tuning parameters like temperature and frequency penalty too.
Please give some more detailed description of your use case, we will see how we can help you using Juji.
With that said, Juji does leverage OpenAI APIs in many features of our product, and you do not need your own OpenAI API key to use these features.
Hi Hyang!
I’m looking to create a career counselor chatbot. My zero-shot prompt initiates a consultation with the user by asking multiple choice questions, one at a time, and then recommending career paths.
I need to set the temp, max tokens, and presence penalty.
I’d like to include a tipping option if possible.
The only other feature is a Feedback form.
Need anything else?
Juji can be configured to create a career counselor chabot, where users have full control on what the chatbot says. We even have option to generate a chatbot automatically by GPT, yet still allow user to customize it. All these can be done in Juji’s user interface without writing code.
However, we are not simply a thin wrapper of OpenAI’s APIs. Our platform integrates GPT deeply in the system and is not exposing it directly to the users, for we retain the option to switch to better APIs when they become available. If you are looking for one of those GPT wrappers, I’d suggest to look elsewhere.
Am I the ‘user’ when you mention user, or the person using the chatbot?
In any case, where would I put my prompt to see if Juji will carry it out?
(Also, where can I set those tuning parameters?)
As @hyang mentioned, parameters and direct prompt to LLM are not exposed. The “user” @hyang refered to is the designer/creator of the chatbot not the conversation participant.
Good question. I am trying to do something similar. Can I do the following with Juji:
- Fine-tune an LLM (could be my ChatGPT account) with data related to a specific topic area.
- Use Juji to provide an interface for end users to ask free form queries and get responses from the AI.
The idea here is to leverage the analytical abilities of a LLM around a certain topic area. Specifically things like sentiment analysis, concordance/word counts, etc.
Juji supports retrieval augmented generation (RAS). You can upload your data as a PDF file, or as a URL if it is publicly accessible. After data is added to a chatbot, it supports questions answering about that data using LLM.
We do not support fine-tune an LLM at this point. Please try uploading the data first and see it does what you need.
Juji already has sentiment analysis available as attribute in the UI. Please see documentation here Chatbot Monitoring - Juji Documentation
Juji also have an extensive library of NLP functions, please see documentation here System Functions - Juji Documentation
1 Like