Exactly what I was looking to do. However, one big problem -> the ability for it to answer questions about the KB couldn't be further away from the experience when I drag-and-drop a file into the prompt. Have you experienced this? This problem alone is keeping us out of production.
Ralf, quality content my man! I am a complete code newb and forked your Gadget Assistants API project. I really want to add the functionality of a user adding files to the thread (not to the Assistant). Any tips on how to do this?
Thanks, Ralf. This is the topic I wanted to know and great. It would be appreciated if you can make a tutorial to upload files with thread level not assistants, meaning that uploaded files are not taken over in new thread??
Thanks for a great walk through. Offhand, is there anything that prevents us from using Threads, to leverage the state management features, along with our own custom RAG, such as combining the existing Embeddings' api's?
I'm not sure about your use-case. But on the surface there shouldn't be anything that prevents you from inserting a step inbetween Assistants calls to e.g. fetch context from somewhere and pass it in with the next assistant call. Is there any reason why you wouldn't want to upload the same documents (context) to the Assistant and use knowledge retrieval instead?
I already have a lot of data stored in a vector store, and the files and retrieval seems limited in terms of how much you can store, and at additional expense, etc. I'd much rather continue using my own RAG, and just refactor the state persistence in, but that's just for this particular use case. I have no bias against the new retrieval in general. @@ralfelfving
Ok, makes sense. As mentioned, you should be able to use it by appending your desired context. But I'd compare pricing of Assistants API vs Chat completions if price is an issue, iirc Assistants is more expensive.
Hi Ralf, great video. VIDEO IDEA... You have another video detailing OpenAI Assistant Chat App using Gadget. Expanding on that video to add file uploads and management would be very useful.
This is great! I understand the use of "openai.beta.assistants.create" well enough, but I can't find any information on how to retrieve an existing Assistant (that I made in the Playground) via the API. Is there a "openai.beta.assistants.getAssistant"? My particular assistant needs access to the many files that I uploaded for it on the playground, but the user doesn't need to be able to upload their own. Edit: I figured it out. It's not called getAssistant, it's called retrieve. Use it like const response = openai.beta.assistants.retrieve("assistant id from playground");
Hi Ralf, thanks a lot for those incredible explanations. I'm working on a new project and I'm really new to this kind of stuff. I created a project and deployed it online by following a tutorial on TH-cam. Although I made some customizations, the project currently uses the chatGPT model and provides generic responses. I'm interested in having the conversation part of the project communicate with the assistant I created in OpenAI as you do in this video, but I'm not sure if this is possible. Could you please help me with this? I would greatly appreciate your assistance. Aside from this message, I will also send an email regarding this matter.
![[Part 1] ทารกน้อยถูกทิ้ง👶 หญิงชราเก็บขวดขายเลี้ยงดูเขาจนโต #โลกการละคร](http://i.ytimg.com/vi/nPTDvFCExks/mqdefault.jpg)
Super useful. Can not be more simplified. Thanks a ton for your efforts. Success !
Another great tutorial, thanks a lot!
Glad you liked it!
Exactly what I was looking to do. However, one big problem -> the ability for it to answer questions about the KB couldn't be further away from the experience when I drag-and-drop a file into the prompt. Have you experienced this? This problem alone is keeping us out of production.
Ralf, quality content my man! I am a complete code newb and forked your Gadget Assistants API project. I really want to add the functionality of a user adding files to the thread (not to the Assistant). Any tips on how to do this?
Thanks, Ralf. This is the topic I wanted to know and great. It would be appreciated if you can make a tutorial to upload files with thread level not assistants, meaning that uploaded files are not taken over in new thread??
Thanks for the suggestion, i have a few other tutorials lined up before i do that one. Keep your eyes peeled in the future.
Thanks for a great walk through. Offhand, is there anything that prevents us from using Threads, to leverage the state management features, along with our own custom RAG, such as combining the existing Embeddings' api's?
I'm not sure about your use-case. But on the surface there shouldn't be anything that prevents you from inserting a step inbetween Assistants calls to e.g. fetch context from somewhere and pass it in with the next assistant call. Is there any reason why you wouldn't want to upload the same documents (context) to the Assistant and use knowledge retrieval instead?
I already have a lot of data stored in a vector store, and the files and retrieval seems limited in terms of how much you can store, and at additional expense, etc. I'd much rather continue using my own RAG, and just refactor the state persistence in, but that's just for this particular use case. I have no bias against the new retrieval in general.
@@ralfelfving
Ok, makes sense. As mentioned, you should be able to use it by appending your desired context. But I'd compare pricing of Assistants API vs Chat completions if price is an issue, iirc Assistants is more expensive.
I'd still like to use Assistants, but mainly for Threads to manage the state. @@ralfelfving
Hi Ralf, great video. VIDEO IDEA... You have another video detailing OpenAI Assistant Chat App using Gadget. Expanding on that video to add file uploads and management would be very useful.
Bear in mind the token limit for large files. Also it hullucinates sometimes with pdf files.
how about in php? if I want to use curl?
This is great! I understand the use of "openai.beta.assistants.create" well enough, but I can't find any information on how to retrieve an existing Assistant (that I made in the Playground) via the API. Is there a "openai.beta.assistants.getAssistant"? My particular assistant needs access to the many files that I uploaded for it on the playground, but the user doesn't need to be able to upload their own.
Edit:
I figured it out. It's not called getAssistant, it's called retrieve. Use it like const response = openai.beta.assistants.retrieve("assistant id from playground");
You got it!
Hi Ralf, thanks a lot for those incredible explanations. I'm working on a new project and I'm really new to this kind of stuff. I created a project and deployed it online by following a tutorial on TH-cam. Although I made some customizations, the project currently uses the chatGPT model and provides generic responses. I'm interested in having the conversation part of the project communicate with the assistant I created in OpenAI as you do in this video, but I'm not sure if this is possible. Could you please help me with this? I would greatly appreciate your assistance. Aside from this message, I will also send an email regarding this matter.
Sorry, I missed this comment. Hope you figured it out!
You are the best! I'll watch the new one again and try to figure it out. Thanks a lot! @@ralfelfving
How about in Python?
There's a lot of people doing Python tutorials so I'm sure you can find some out there, I'll stick to Node.
What is use of these assistants?if we can't imploy
This is a tutorial of how to build with these APIs. The use is up to you to decide.
@@ralfelfving thank You.... Can we imploy an assistant into website?