Hello Bwm561.
I tryed to feed data into the API prompt using dynamic text but did not work well. I used the search for things in the DB but when I tested (with or without this call) the results were pretty much the same… Still looking for guidance. It would be extremely valuable to have a video on that from BUBBLE felas.
@felipesantosdutra21 made a guide about it here.
You need to setup function calls to an open ai assistant. There’s calls are API call you your own bubble database
Hi - this is interesting and doesn’t look terribly complicated if it’s done through Buildship. I have some questions:
- If Buildship is not used (it looks great but I would prefer not to use another app if possible), how complicated is the set up and has anyone made a guide for that?
- If I use Buildship and use the chat widget, can the actual chat messages and threads be saved in the Bubble database?
- With Buildship, is there also the possibility of doing web scraping, and then having that information fed into the Bubble database? I’m trying to create an easy way for a) Users to pull information about their university into the Bubble database without needing to enter it manually, and b) having OpenAI search the database for results, creating a new partnership between universities, or updating university information
@bwm561 yeah Buildship is a nice add-on to Bubble!
- I think it can take some time to set up. You can use a plugin for having the API calls set up, but you will still need to check the status of each function calling you are doing and submit the output. I have guide here that shows how to set up a function calling to create an entry in your database only using Bubble.
- Yes, you can save the thread Id and the chat message in bubble database, but you would need to setup that step in Buildship.
- Yes, you can do scraping with Buildship and you could save that information in your Bubble database by setting up the backend workflow in bubble and using the Ai assistant to make the API call with the data. Check this video and demo for scraping using Buildship
I everyone I hope you are well.
Thank you for contributing to this topic. I am not certain any of the above solutions will work for me.
I currently have an application like thumbtack(mine niched to my location). The database I assume is not entirely different from mine. I want to include AI to the search part of the webapp, that reads the database and provide the user with results.
Example query: " I am looking for a house helper/maid, at Location X, at price X.
Then the output to be a list of the profiles close to the search.
Please help
Hi, did you succeed in this? I have tried severally to integrate OpenAI but it seems there is no light at the end of the tunnel.
At the end of the day, there are no other methods than:
- Define a “Do a Search for” operation for each of your data types, ensuring that “Ignore if empty” is checked and using constraints as variables.
- Define a function/tool in conjunction with your prompt that takes the user query in natural language and infers all the variables you need from step 1.
- Use the “Do a Search for” operation to return results based on the user’s query.
Any other approach—such as assuming that the AI will magically return results from your database without any search algorithm—is, at worst, a flawed design and, at best, highly inefficient (the other way around works also).
Giving it Data API schema and allowing it to specify its own constraints also works.
Did you have any issues with the AI not accurately following secondary keys?
No, with correct prompting it manages just fine.
This looks useful, however but is very slow when loading. Is there a way I can optimize loading time?
Thank you. I am going to try this out and get you feedback
Subscribe to broadband ?
Not yet
@machkorir30 @simnikiwe I made a plugin that handles all the setups of the AI assistants and functions.
The assistant can get information from your database, create a new entry, and answer back in natural language or in JSON. The JSON contains the Unique IDs in case you want to filter and RG.
It can perform also multiple actions. It can get data from one table first, then from another table, and finally, create a new entry.
It’s a private plugin for the moment, but I want to get some early feedback!
Let’s have a chat and see if it solves your use case
Just to throw another option in the ring, this time from Google:
Also, just having chatgpt formulate SQL queries works pretty well!
Thank you so much for this useful resource drfalken