Most Rasa chatbots fetch data from an API. E.g. in the financial-demo, a user can see their latest transactions. Here, we have a dummy API returning restaurant results.
Using the ChatGPT API, we then allow users to ask questions about the data they've just been shown. We have an existing feature called Knowledge Base Actions which does the same thing. In comparison, using ChatGPT is much simpler, but also occasionally makes up wrong answers!
Here's the data which the API has access to:
Restaurants | Rating | Has WiFi | cuisine | |
---|---|---|---|---|
0 | Mama's Pizza | 4.5 | False | italian |
1 | Tacos el Cabron | 4.6 | True | mexican |
2 | China Palace | 3.2 | False | chinese |
3 | The Cheesecake Factory | 4.2 | True | american |
4 | Pizza Hut | 3.8 | True | italian |
5 | Biryani Bowl | 4.9 | False | indian |
6 | Burger King | 2.6 | True | american |
7 | Taco Bell | 3.5 | True | mexican |
8 | Freshii | 4.7 | True | healthy |
9 | Panda Express | 3.9 | False | chinese |
10 | McDonald's | 2.3 | True | american |
11 | Sushi Roku | 4.8 | True | japanese |
12 | Olive Garden | 4 | True | italian |
13 | Chipotle | 4.4 | True | mexican |
14 | Little Caesars | 3.1 | False | italian |