No video
OpenAI Assistants API - Part 3: Code Interpreter with run steps walk through
Рет қаралды 2,315
9 ай бұрынWho says GPT 4 can't do math? In this video, we show you how you can use the code interpreter tool to get LLMs to write and execute it's own code to complete tasks that LLMs normally would have trouble completing - such as mathematics!
This a 3 part series to explain and demo how OpenAI's Assistant API works:
- Part 1: Information Retrieval ( • OpenAI Assistants API ... )
- Part 2: Adding Agents with Function Calling ( • OpenAI Assistants API ... )
- Part 3: Code Interpreter with Run Steps walk through ( • OpenAI Assistants API ... )
OpenAI announced the brand new Assistants API on DevDay recently. The Assistant API enables you to easily build a custom chatbot on top of the GPT4 LLM that knows how to retrieve additional information from files that you upload, write/execute code and also can take actions (like agent functionality) by being able to call up to 128 functions that you give it. You would've heard about "GPTs" from DevDay - this is the same functionality but exposed via API
@fheralanis 7 ай бұрын
your video has helped me too much to find the source of the messages that the assistant generates when the 'run' is 'in-progress' using the 'code interpreter'. thank you very much for sharing !!
@JamesRBentley 8 ай бұрын
Great demo, and Lily is awesome, I dropped her a line on LinkedIn and she kindly shared the colab workbooks.
@rammss 8 ай бұрын
Very nice video series, thank you.
@stevefocht5520 8 ай бұрын
You mentioned that there will be a follow up video showing how to implement this on the Azure infrastructure. Is that published yet?
@hubel-labs 8 ай бұрын
Not yet! Will let you know when I get to it!
@aarush5925 8 ай бұрын
Thanks for the great value but I m concerned alot about whole Open Ai drama what would happen if whole chatgpt is gonna be shut down
@hubel-labs 8 ай бұрын
yeah, my guess is that Microsoft will find a way to help keep things running one way or another - it already has an instance of the OpenAI foundation models on Azure for enterprise use. In any case, the whole field is moving so fast - the models we use today will likely be different from what we will be talking about in 12 months time.
@nazihfattal974 8 ай бұрын
Quick question: when building an interface for an assistant, showing the (user, assistant) conversation is obviously important. The issue I am facing right now is that when printing a thread, it will print everything in it. So, every time a new run takes place, showing the conversation requires that you print the complete thread once more..... looks redundant and disorganized. There does not seem to be a way to only print the outcome (new messages added) per run. The only way to do this is to create code that tracks the messages and only print out the newly added messages to the thread. Any easier way of achieving this? Thanks.
@hubel-labs 8 ай бұрын
If you look at the “List messages” API here, you will see that there is an optional “after” parameter. So if you store the last message id that you printed to screen, you can fetch everything after that message using that “after” parameter so you are only printing each incremental set of messages out. I haven’t tried it myself but that’s what I might try to fix that issue. platform.openai.com/docs/api-reference/messages/listMessages Also, I believe that openAI said that they will be working on releasing streaming - so maybe that will help too when it comes out.
@nazihfattal974 8 ай бұрын
@@hubel-labs Thank you very much! I will give it a try.
@nazihfattal974 8 ай бұрын
Thanks for another great video. OpenAI will probably not go away, but I am looking for other source where I can use the same assistant functionality (RAG and function calling) as OPenAI's. I use HuggingFace (HF) models most of the times, but what drew me to OpenAI was the AI assistant (with mainly function calling and possibly integrated RAG instead of using Langchain or Llamaindex for RAG). Wondering if you came across something similar to OpenAI's assistant. I tried HF agent, but it is nothing like OpenAI's assistant.
@mayurmoudhgalya3840 8 ай бұрын
Have you tried Autogen? I've had luck with Autogen ret agents more than open AI so far. Simply my 2 cents
@nazihfattal974 8 ай бұрын
@@mayurmoudhgalya3840 Thanks for the pointer. I have heard about autogen but have not had the chance to try it. Just checked their implementation of function calling, Microsoft (being an investor in OpenAI) just ported the functionality into their own autogen. It also seems to be more efficient in collecting the function output than OpenAI.
@hubel-labs 8 ай бұрын
I was thinking of looking for the same thing myself - I will try to find some more time to test out autogen and see what else is out there.
@defaultdefault812 8 ай бұрын
Azure AI studio is what you want to be learning for enterprise solutions.
@nazihfattal974 8 ай бұрын
@@defaultdefault812 Thanks for the pointer. Azure AI at the moment is obviously the enhanced version of OpenAI's playground interface, and AutoGen is the equivalent of OpenAI's Assistant API. I can see Azure AI Studio being used as a preparation step to possibly finetune a model if needed to be called from Autogen. The limitation here is that Azure Ai Studio has access to a lot more models than Autogen. It makes sense to have Autogen API access all the models available in AI Studio. Something Microsoft may need to work on.
@MikeRhodesIdeas 8 ай бұрын
Oscar Wilde, “Imitation is the sincerest form of flattery” & all that! Maybe put your logo/watermark more prominently on the slides - and in multiple code cells as comments? Just make it difficult for them... but ultimately people will do this anyway & there's really no way to stop them completely :(
@hubel-labs 8 ай бұрын
Thanks! I will try to put some comments in the code as you suggested!
@JOHNSMITH-ve3rq 8 ай бұрын
Text very distracting !! Killer content otherwise tho.
Hubel Labs
Рет қаралды 10 М.
Hubel Labs
Рет қаралды 6 М.
Henrik Kniberg
Рет қаралды 1,8 МЛН