Absolutely fantastic run through of how to integrate with OpenAI. Your video has saved me many hours of trial and error - and it's great to see that you also address security issues within the approach. One slight enhancement I did was to use Air Alert to alert the current status to the user after each OpenAI run poll to see what the status was. Although you don't get partial responses back, at least the user can see that it's doing something in the background. I've also used gpt-4o-mini as the model rather than full-on gpt-4 to reduce costs of the chatbot
Im trying to implement a feature where after the user uploads a set of data, they are then able to ask the AI insights about it. However i keep getting an error (no response from the AI) when the user asks a question. My JSON formatting is correct and tried everything, but do u think it has something to do with privacy issues/ admin roles? (when uploading data manually, everything works fine). Pls help its for a project
@@zubair_nocode i used the file uploader to let the user upload a csv file, after that i use a parser to convert it into a JSON format, saved it into my database and then in the API call, i send the saved data from my database as part of a prompt and another prompt for the user's question regarding the uploaded data
@@zubair_nocode i used the file uploader to let the user upload a csv file, after that i use a parser to convert it into a JSON format, saved it into my database and then in the API call, i send the saved data from my database as part of a prompt and another prompt for the user's question regarding the uploaded data
@@zubair_nocode i used the file uploader to let the user upload a csv file, after that i use a parser to convert it into a JSON format, saved it into my database and then in the API call, i send the saved data from my database as part of a prompt and another prompt for the user's question regarding the uploaded data
@@zubair_nocode i used the file uploader to let the user upload a csv file, after that i use a parser to convert it into a JSON format, saved it into my database and then in the API call, i send the saved data from my database as part of a prompt and another prompt for the user's question regarding the uploaded data
Would be amazing if you could make a video on 'function calling' with Assistants API. Good example could be building a function on retrieving accurate information from Pinecone while maintaning chat history :)
I've tried following these instructions but find myself struggling with loop function where we use the "do ever x seconds"-only when state is in progress. To me it seems like this runs whenever and not just for the condition. Any ideas?
Thanks for throwing these together. They are informative. However, I went down the polling method, and even though it only runs when the assistants call status is in progress, it still pings every second. I have to imagine that’s not ideal and that you’re using at least a workload unit every second With that method, right?
@@aisaluk are you asking the AI to return some words in bold? Add in the prompt to return in bbcode. That is the default markup language that bubble uses in text element. Then it should show bold in the front end
How much does this cost in terms of Workload Units? Let's say someone has a 100 back and forth messages with an assistant on a given day, how much would this be in workload units?
🤝 Work with me: www.azkytech.com/book-a-call
Absolutely fantastic run through of how to integrate with OpenAI. Your video has saved me many hours of trial and error - and it's great to see that you also address security issues within the approach. One slight enhancement I did was to use Air Alert to alert the current status to the user after each OpenAI run poll to see what the status was. Although you don't get partial responses back, at least the user can see that it's doing something in the background. I've also used gpt-4o-mini as the model rather than full-on gpt-4 to reduce costs of the chatbot
@@IanMatthews-s7z thank you so much for the kind words.
And yes I love using air alert in various places as well for better UX and feedback
Great video! thanks for sharing your knowledge!
Thanks! This is a much needed video.
u saved my life
How would you set up the in-progress - completed loop in a way that won’t break my WU?
Thanks
Im trying to implement a feature where after the user uploads a set of data, they are then able to ask the AI insights about it. However i keep getting an error (no response from the AI) when the user asks a question. My JSON formatting is correct and tried everything, but do u think it has something to do with privacy issues/ admin roles? (when uploading data manually, everything works fine).
Pls help its for a project
@@TechlarDigital how do you upload the file to openai?
@@zubair_nocode i used the file uploader to let the user upload a csv file, after that i use a parser to convert it into a JSON format, saved it into my database and then in the API call, i send the saved data from my database as part of a prompt and another prompt for the user's question regarding the uploaded data
@@zubair_nocode i used the file uploader to let the user upload a csv file, after that i use a parser to convert it into a JSON format, saved it into my database and then in the API call, i send the saved data from my database as part of a prompt and another prompt for the user's question regarding the uploaded data
@@zubair_nocode i used the file uploader to let the user upload a csv file, after that i use a parser to convert it into a JSON format, saved it into my database and then in the API call, i send the saved data from my database as part of a prompt and another prompt for the user's question regarding the uploaded data
@@zubair_nocode i used the file uploader to let the user upload a csv file, after that i use a parser to convert it into a JSON format, saved it into my database and then in the API call, i send the saved data from my database as part of a prompt and another prompt for the user's question regarding the uploaded data
Really helpful. Thanks!
Thank you Zubair bhai.
Is there a reason you chose a front-end workflow vs a backend workflow?
@@LafeSpace for the sake of the tutorial.
For a production app , backend workflows likely better.
But those are only on the paid plans
@@zubair_nocode I would love to see that process. All the tuts I've seen only show front-end workflows.
thanks! a lot, is that possible to have a quick tutorial to display the chunks in the last part about streaming? thank you
Would be amazing if you could make a video on 'function calling' with Assistants API. Good example could be building a function on retrieving accurate information from Pinecone while maintaning chat history :)
@@aisaluk that's quite an interesting use case.
General function calling has come up before but haven't had the time to do it yet
@@zubair_nocode Near bullet-proof responses. Worth a video imo, especially when Assistants API itself is a little tricky to get around.
Shukriya!
Oh bro, thank u so much! You have saved my life 🥰🥰
@@DioggGS thank you so much for the kind words
Thanks for this! Are you aware of anyone coaching or stamping templates for 3rd party servers for this use case?
not sure I understand 'stamping templates'. but there are plugins that have a small monthly fee and they maintain a 3rd party server for streaming
Man I dont understand how to use Tools along bubble using Assistants. How to do that? Could you record a video?
I've tried following these instructions but find myself struggling with loop function where we use the "do ever x seconds"-only when state is in progress. To me it seems like this runs whenever and not just for the condition.
Any ideas?
Hmm. I suggest slowly checking using step by step debug mode to see what went wrong
Thanks for throwing these together. They are informative. However, I went down the polling method, and even though it only runs when the assistants call status is in progress, it still pings every second. I have to imagine that’s not ideal and that you’re using at least a workload unit every second With that method, right?
@@DonnieBalliet yea WU are used with the polling method. But only when needed.
Been getting issues with formatting. Instead of getting bolded words, it returns responses in **.
What would be the best solution for this?
@@aisaluk are you asking the AI to return some words in bold?
Add in the prompt to return in bbcode. That is the default markup language that bubble uses in text element. Then it should show bold in the front end
@@zubair_nocode Yeap, in bold?
You got any videos/tutorials at hand on bbcode?
Thanks!
@@zubair_nocode I thought I had sent this already ... question was if you had any quick tutorial videos around this issue at hand?
How much does this cost in terms of Workload Units? Let's say someone has a 100 back and forth messages with an assistant on a given day, how much would this be in workload units?
hard to tell without doing a test. depends on prompt & length of response..
Could you share the server script at the final of the video? Appreciate your content a lot, thnx in advance
pastebin.com/RqTSiTVe
how do we make this so its per user account?
@@decentralisedproperties8066 save the thread ID in the database and add a field to relate it to the user
Great video! Is there a way to download this as a template?
@@brodyjohnston04 afraid not yet
Need help. Reach out to me please
Please post the server script 🥲
pastebin.com/RqTSiTVe
@@zubair_nocode thanks!