Thanks Leo! of course this is now going to require Assistant V2 in the headers. And retrieval tool has been changed to file search which I assume will impact the future videos. Really looking forward to the next parts!
@@leomcmillion Would be amazing to give openAI Assistant access to an Authenticated users' files in Firebase so that the user can chat with their own documents using our FF assistant. I wonder if that's a good candidate for a future video on your channel.
what is the thread id? I believe before you would just pass in the entire conversation everytime but I think there may be token overflows at some point? I think I'd have to limit it to 50 messages every time for each chat or something or there could be way too much data being sent to chatgpt.
Thanks for the usefull video it really helps me! I am working on developing a GPT-powered tool to extract questions and answers from PDF files. However, I'm encountering some challenges in the implementation.
Hi, very interesting! I am following the tutorial :) I have a question regarding the Actions for the button (send). In variables, I choose message but then for choosing the value. I can´t see "input (Label here...) option. How can I add it? This doesn´t appear in my options. Thank you!
Have you checked so the value in your "message" datatype is a string? Also check so the input field is set to "text "and not to "number". It could also be that you renamed the text field to something else. It should pop up as a option, so it's probably something smal like this. Hope this helped somewhat at least.
do you guys have a 403 error message when trying to get messages step ? I don't understand, everything runs just fine before for creating thread, add message create run and get run.... I don't understand...-_-
Hello Leo! I made an AI assistant in Flutter Flow and I added also a vectoring data base to make the assistant make the answers from a specific knowledge base. The problem is that the assistant is using source annotations and they appear at the end of the message with weird brackets and symbols (e.g: [\d+\], or 4:1 â sourcã). Any idea how I can clear the messages in Flutter Flow?
To make incoming strings to flutterflow look as normal text instead of having those weird symbols you can create a custom function in flutterflow that decodes a string to be UTF8. This should work since I think they are using UTF8 encoding. Here is the code you can try: String decodeUtf8String(String encodedString) { List bytes = encodedString.codeUnits; return utf8.decode(bytes); } Hope this is what you were asking for, if I misunderstood your question please let me know.
@@leomcmillion hei Leo. Sadly it did not work :(. I sent you an email with some screenshots just to make myself clearer. Whenever you have the time to check it out! Thank you so much for your help!
Have you found a way to display past assistant chat history for users to reference back too? I also want to add 20+ assistants for various user functions. is there a way to do that without creating 30+ pages?
Yes, if you store the thread ids to the user account you can just loop through them and use the "getMessages" api call from the video (but grab all the messages instead of only the latest one) to get previous conversations. You can create several assistants on the same page by creating a thread id for each one, and use the specific assistant ids for each assistant. Otherwise the flow is basically the same, but you will need a seperate "messages" list for each one. Hope that helped.
Define the API calls individually instead. This will require you to fill in the headers and API separately for each API call. But not that much work since there were just around 5 API calls to define.
does adding the api call this way expose the open ai key to the public? do you have any videos or advice I can use to help keep the key private from preying eyes. thanks, your vids are awesome.
I explained in the video the things I changed. It was basically switching from a periodic action to a normal loop. I also switched the index of the data list in the "getMessageS" API call. Try watching the last part again, where I show this.
How did you get the "messages - (Lists < Data (message) >)"?
Your videos are very useful
Great vid Leo. Thank you so much.
Thanks Leo!
of course this is now going to require Assistant V2 in the headers. And retrieval tool has been changed to file search which I assume will impact the future videos. Really looking forward to the next parts!
Yep, that is right! Actually though of releasing an updated version after the coming video.
@@leomcmillion I will be watching.
@@leomcmillion
Would be amazing to give openAI Assistant access to an Authenticated users' files in Firebase so that the user can chat with their own documents using our FF assistant. I wonder if that's a good candidate for a future video on your channel.
@@KJPsyXX Absolutely! Have considered a video like this as well. But will be later down the line when I get some other videos out first.
what is the thread id? I believe before you would just pass in the entire conversation everytime but I think there may be token overflows at some point? I think I'd have to limit it to 50 messages every time for each chat or something or there could be way too much data being sent to chatgpt.
Thanks for the usefull video it really helps me! I am working on developing a GPT-powered tool to extract questions and answers from PDF files. However, I'm encountering some challenges in the implementation.
Glad to hear you liked the video. What implementation challenges are you facing?
Thanks for this great video! Would you consider updating it to use FlutterFlow streaming API to work with OpenAI Assistants?
Absolutely! Had actually planned on recording an video about that in the autumn when I get back to recording videos again. Thanks for the suggestion!
Hi, very interesting! I am following the tutorial :) I have a question regarding the Actions for the button (send). In variables, I choose message but then for choosing the value. I can´t see "input (Label here...) option. How can I add it? This doesn´t appear in my options. Thank you!
Have you checked so the value in your "message" datatype is a string? Also check so the input field is set to "text "and not to "number". It could also be that you renamed the text field to something else. It should pop up as a option, so it's probably something smal like this. Hope this helped somewhat at least.
do you guys have a 403 error message when trying to get messages step ? I don't understand, everything runs just fine before for creating thread, add message create run and get run.... I don't understand...-_-
Hello Leo! I made an AI assistant in Flutter Flow and I added also a vectoring data base to make the assistant make the answers from a specific knowledge base. The problem is that the assistant is using source annotations and they appear at the end of the message with weird brackets and symbols (e.g: [\d+\], or 4:1 â sourcã). Any idea how I can clear the messages in Flutter Flow?
To make incoming strings to flutterflow look as normal text instead of having those weird symbols you can create a custom function in flutterflow that decodes a string to be UTF8. This should work since I think they are using UTF8 encoding. Here is the code you can try:
String decodeUtf8String(String encodedString) {
List bytes = encodedString.codeUnits;
return utf8.decode(bytes);
}
Hope this is what you were asking for, if I misunderstood your question please let me know.
@@leomcmillion hei Leo. Sadly it did not work :(. I sent you an email with some screenshots just to make myself clearer. Whenever you have the time to check it out! Thank you so much for your help!
@@ghitaneugeneduard4318 I did not get any screenshots via email. Please try to send again.
Great video
Thanks!
@leo Great video. Can you please share the app or the workflow that works?
Did the workflow shown in the video not work for you?
@@leomcmillion I got lost when you changed something after your 1st try gave an error. I would be happy if I could clone it and tweak to learn
didn’t work you changed something that was not shown
@@clinbrokers I will have another look myself and let you know if I missed to mention something.
Have you found a way to display past assistant chat history for users to reference back too? I also want to add 20+ assistants for various user functions. is there a way to do that without creating 30+ pages?
Yes, if you store the thread ids to the user account you can just loop through them and use the "getMessages" api call from the video (but grab all the messages instead of only the latest one) to get previous conversations.
You can create several assistants on the same page by creating a thread id for each one, and use the specific assistant ids for each assistant. Otherwise the flow is basically the same, but you will need a seperate "messages" list for each one.
Hope that helped.
@@leomcmillion I would need some help on direction. Would you be interested in a zoom call? or video call to help?
@@MatthewFusco-wx9pq you can contact me on my email: leomcmillion@gmail.com
just reached out thanks! @@leomcmillion
What would change if you don’t have the API grouping? It’s only available for pro users
Define the API calls individually instead. This will require you to fill in the headers and API separately for each API call. But not that much work since there were just around 5 API calls to define.
does adding the api call this way expose the open ai key to the public? do you have any videos or advice I can use to help keep the key private from preying eyes. thanks, your vids are awesome.
Have planned to release a video about this within a week. This is problem many developers are struggling with in flutetrflow.
didn’t work you change something that doesn’t trigger the assistant on the button press just send the input and no response
I explained in the video the things I changed. It was basically switching from a periodic action to a normal loop. I also switched the index of the data list in the "getMessageS" API call. Try watching the last part again, where I show this.
I will have another look myself and let you know if I missed to mention something.