So, quick overview so far. For the setup: > Create thread > Create assistant Then for interacting: > Create message thread > List messages > Run Assistant
Just went through your channel to see if you have contents about the newly updated API'w from open ai wow 🎉🎉🎉🎉 you've gone everything and more thank you so much 🎉🎉
We've now released 3 parts to this mini series. Part 3 is exclusive to our members and considered different approaches to fetching the new AI generated messages. (www.planetnocode.com/tutorial/plugins/openai/openai-assistant-bubble-io-api-connector-part-3/). In some ways it is really worth weighing up if the assistant based chat is worth the extra complexity compared to the Chat Competition endpoint. (platform.openai.com/docs/api-reference/chat/create)
Thank you, really good to the point info! How do you approach the issue that now we have to keep calling OpenAI API - List Messages untill we get the answer? I have an Event running on the page every 3 sec but i realize it's not the best solution :S
Great point! We have also considered that solution and are fully aware of potential issues with Workload Units and API rate limiting. Perhaps until OpenAI offer webhooks the ‘old way’ is better for many use cases.
Part 2 is exclusively available to our members on our website: www.planetnocode.com/tutorial/plugins/openai/openai-assistant-bubble-io-api-connector-part-2/ We will record Part 3 this morning and release it on our website for our members.
Yes, with gptvision api. Here nocodeplanet video on that (only the back, not the front). Hoping they release another one with the interface part. Link: ua-cam.com/video/7pe91cGJLj8/v-deo.html
Just in case this is what you're thinking of... ChatGPT (Custom GPTs) are not accessible with the API. But it sounds like you mean Assistants in the dev tools part of the OpenAI website right? Yes you can make API calls to specific Assistant by using the Assistant ID when you use the Run endpoint (platform.openai.com/docs/api-reference/runs/createRun)
Hey man can you make a video on setting up Twilio live conference API on Bubble? Or any live streaming video calls on Bubble? The Bubble documentation is very limited and I think you'll be the #1 in explaining it. I've delayed the development of my app for a week for this and can't find an answer.
Any reason why the gpt 3.5 version is sooooo stupid compared to the gpt 4 model? No matter the size of files and the instructions length and no matter how well organized it is, it always somehow hallucinations, does half of the work, or simply not follow the instructions where gpt4 is like a good student. Is there a specific thing I need to make sure to enable so it follows the instructions? Heck, even ChatGPT follows instructions better. (GPT 4 is so expensive it's out of the question, it's like 10 cents per message where GPT 3.5 is like a cent per message)
What temperature are you using? We find 0.7 is a good starter. Have you tried using the seed parameter to create more consistent outputs. Internally we’ve found the GPT-4 is needed more often than not now.
We don't actually recommend using the OpenAI Assistant API yet as it is still in beta and doesn't play well with Bubble. You can still create a very effective AI chat app, our Udemy course is the most detailed explanation: www.udemy.com/course/build-a-chatgpt-clone-with-no-code-bubble/?referralCode=AEEDAE8F3E9A20FFC79F
Part 2 is available exclusively to our members on our website www.planetnocode.com/tutorial/plugins/openai/openai-assistant-bubble-io-api-connector-part-2/ Part 3 will be recorded this morning and also released on our website for our members.
So, quick overview so far.
For the setup:
> Create thread
> Create assistant
Then for interacting:
> Create message thread
> List messages
> Run Assistant
Just went through your channel to see if you have contents about the newly updated API'w from open ai wow 🎉🎉🎉🎉 you've gone everything and more thank you so much 🎉🎉
Well done! Easy to follow and great instructions. Looking forward to the next video where I add this to the front-end
Great work thanks 🙏
hey! thanks for this great video ! Can you guide us on how to receive the assistant's message when taking a while?
We've now released 3 parts to this mini series. Part 3 is exclusive to our members and considered different approaches to fetching the new AI generated messages. (www.planetnocode.com/tutorial/plugins/openai/openai-assistant-bubble-io-api-connector-part-3/).
In some ways it is really worth weighing up if the assistant based chat is worth the extra complexity compared to the Chat Competition endpoint. (platform.openai.com/docs/api-reference/chat/create)
Thank you, really good to the point info! How do you approach the issue that now we have to keep calling OpenAI API - List Messages untill we get the answer? I have an Event running on the page every 3 sec but i realize it's not the best solution :S
Great point! We have also considered that solution and are fully aware of potential issues with Workload Units and API rate limiting. Perhaps until OpenAI offer webhooks the ‘old way’ is better for many use cases.
thanks for this video! really appreciate it and very helpful. excited for part 2 :)
Part 2 is exclusively available to our members on our website: www.planetnocode.com/tutorial/plugins/openai/openai-assistant-bubble-io-api-connector-part-2/
We will record Part 3 this morning and release it on our website for our members.
Thank you for you video. Any tutorial about math expression display on a bubble chat app?
Brilliant! Waiting for PT2… Thanks a million!
Thanks for this great tuto !
Do you know if it is possible to send an image from Bubble and ask gpt some questions about this image ?
Yes, with gptvision api. Here nocodeplanet video on that (only the back, not the front). Hoping they release another one with the interface part.
Link: ua-cam.com/video/7pe91cGJLj8/v-deo.html
Thanks for sharing!
Hi. If I ask my assistant to plot me a graph how do I display the graph?
by connecting the api in those following steps, how would that connect the assistant api you trained on the OpenAI platform?
Just in case this is what you're thinking of... ChatGPT (Custom GPTs) are not accessible with the API.
But it sounds like you mean Assistants in the dev tools part of the OpenAI website right? Yes you can make API calls to specific Assistant by using the Assistant ID when you use the Run endpoint (platform.openai.com/docs/api-reference/runs/createRun)
Would streaming be a way to easily resolve the issues in video 3?
Sadly no, as the Bubble API Connector does not support streaming. There are plugins that may help with this.
Do you offer consulting services! Great Video, BTW!
Please feel free to drop us a message on our website
@@planetnocode cool. ✌️
Hey man can you make a video on setting up Twilio live conference API on Bubble? Or any live streaming video calls on Bubble?
The Bubble documentation is very limited and I think you'll be the #1 in explaining it. I've delayed the development of my app for a week for this and can't find an answer.
For my app, it's essential that users can modify the behavior of the assistants
How do I add different assistance such as giving my bots personalities
When you create an assistant, you get an assistant ID. This can be used to involve a particular assistant in a conversation.
Any reason why the gpt 3.5 version is sooooo stupid compared to the gpt 4 model?
No matter the size of files and the instructions length and no matter how well organized it is, it always somehow hallucinations, does half of the work, or simply not follow the instructions where gpt4 is like a good student.
Is there a specific thing I need to make sure to enable so it follows the instructions? Heck, even ChatGPT follows instructions better.
(GPT 4 is so expensive it's out of the question, it's like 10 cents per message where GPT 3.5 is like a cent per message)
What temperature are you using? We find 0.7 is a good starter. Have you tried using the seed parameter to create more consistent outputs. Internally we’ve found the GPT-4 is needed more often than not now.
Hi
lol parts 2 and 3 cost three figures
We don't actually recommend using the OpenAI Assistant API yet as it is still in beta and doesn't play well with Bubble. You can still create a very effective AI chat app, our Udemy course is the most detailed explanation: www.udemy.com/course/build-a-chatgpt-clone-with-no-code-bubble/?referralCode=AEEDAE8F3E9A20FFC79F
When is PT2 coming?
Part 2 is available exclusively to our members on our website www.planetnocode.com/tutorial/plugins/openai/openai-assistant-bubble-io-api-connector-part-2/
Part 3 will be recorded this morning and also released on our website for our members.