With ThinkAutomation it's easy to create context-aware ChatGPT powered chat bots. The bot can lookup articles from your knowledge store to provide context.
By what if a user asks something like:
'What is the current status of the API service?'
The user wants to know the current real-time status of some service (this could apply to any question where the answer requires some sort of real time lookup), but the Knowledge Store isn't going to have this.
The way to do this is to create a Knowledge Store article, with text along the lines of:
If the user asks a question similar to:
What is the current service status?
What is the status of the web API?
What is the status of the public API?
Is the web API active?
Is the gateway active?
Is the API service running?
If the user asks the above questions or about the service status, or gateway status, or API status respond with '[STATUS]' only.
When the user asks their question - the above article will get added to the ChatGPT context (because it is related). The article tells ChatGPT to only respond with '[STATUS]' if the user asks about the current service status.
Inside our Automation, after the 'Ask ChatGPT To Respond To A Prompt' action we can check the response. If the response is equal to '[STATUS]' we can then execute other automation actions to perform some sort of service check (lookup from a database, or do a HTTP GET etc). We then return the result of that to the chat. If the response from ChatGPT is NOT '[STATUS]' we return the ChatGPT response as normal.
So Knowledge Store articles can be used to both provide Knowledge AND to instruct ChatGPT to respond in a certain way. We can use these tag-style responses in our Automation to perform additional actions.