Hey guys,
We need to be able to use the chat component exactly like chatgpt itself.
For example; We need chat messages to be remembered and we need different API endpoints to work dynamically without any problems. It says it would only work with Retool AI for now and making it hard to work with other APIs. And all the other things possible for a chat interface.
Thank you!
Hello @Onur_Bolaca!
Are you looking to remember chat threads for visual displaying or for training the model in real time with incoming data from message threads?
For chat messages to be remembered, they will need to be stored. For an AI to learn from data, the data must be stored in a vector DB.
If you want to display a message "history" you could query that conversation and feed it into the chat component with the 'Message History' input field and concatenate it with the current conversation which is temporarily stored in the frontend component's state
If you want to train a model on data from incoming threads, you could set up a workflow to add chat messages tagged to the user who sent the message into either a traditional DB to then modify/sanitize the data and then insert it into a Vector DB to train the LLM model so it can 'remember' past messages for context.
I am not sure what you mean by "different API endpoints to work dynamically without any problems", are you referring to API endpoints for a backend server that you have access to? Or making query requests from Retool with dynamic values in the API calls?