OpenAI Assistants API coding tutorial

[@deoarlo]

Damn you're fast haha

[@ralfelfving]

I had some time and motivation:)

[@deoarlo]

@@ralfelfvingWould love to see more examples with Assistant API + Function calling. Perhaps in a free flow chatbot

[@jmaicaaan]

Do you think it is possible for an assistant to create a PDF of a music sheet based on the song I'd like to practice playing?

[@ralfelfving]

I'm not sure. I don't think plain vanilla OpenAI can generate a PDF. I don't know if code interpreter can generate PDFs, or if there's a plugin that you could hook in to do it.

[@jmaicaaan]

@@ralfelfving got it, thanks for checking!

[@张超-x1q]

I came across an error that the gpt-4-1106-preview does not exists, but I'm a plus user. Anyone has the same problem?

[@ralfelfving]

No, that's strange. It should be available.

[@andresh77]

Your video helped me earn $100 USD quickly, and it also taught me many other things that will be useful to me, thanks to you!!

[@ralfelfving]

Great, keep learning and earning!

[@solagbenro7295]

Thank you so much for putting this out so fast! Couldn't have been easy and we appreciate it!

[@ralfelfving]

Thanks for taking the time to check it out, hope it's useful!

[@santiago-ow6ou]

Thanks for this. I learned a lot. I have one problem though. Every time I run the assistant it creates a new one in my openai account. I just got done deleting like dozens of math tutor assistants.

[@ralfelfving]

Yes, there's definetely limitations to this simplistic code. I created a tutorial showing how to keep state of existing assistants, threads, and messages. You can watch it here: kzbin.info/www/bejne/mIesd6yeaZZsm9U

[@inck1107]

To me the Assistant API is just another wrapper around the original API (a bit like langchain). However, I would be curious: can we save cost because once initialized they won't be a need to send the character description of the bot with every API call?

[@ralfelfving]

Not sure I agree, but it depends on how much functionality you think it's fair to ascribe to the "wrapper". The Assistant API has some novel functionality like code interpreter or knowledge retrieval (of custom uploaded data). The cost structure of the Chat completion and Assistant APIs are very different, depending on what capabilities you use (code interpreter, knowledge retreival). So it could be cheaper, but it could also be more expensive.

[@Canna_Science_and_Technology]

CAUTION! FYI there is a failure of direct retrieval with GPT-4 using the new OpenAI Assistant API. GPT tokenizes text and creates its own vector embeddings based on its specific training data. The new terms and sequences may not connect well to the pretrained knowledge in GPT's weight tensors. There was no semantic similarity between the new API terms and GPT's existing vector space. This is a fundamental issue with retrieval augmentation systems like Rag - external knowledge is not truly integrated into the model's learned weights. Adding more vector stores cannot solve this core problem. The solution is to have multiple learned "knowledge planes" with trained weight tensors for specific tasks that can be switched in. This is better than just retrieving separate vector representations.

[@zahraahawie3707]

Thanks Ralf, very helpful. How can we integrate this assistant api in our website?

[@ralfelfving]

There's several ways of doing that, neither of which are super simple. My most recent tutorial is an example of how to build a web-based Assistant chat interface.

[@TheRealDOSmile]

I desperately want to see how to integrate Zapier functionality into the assistants beta api. I'm having a hard time understanding that.

[@mitcheldepalo6898]

Screw zappier learn how to use the API for whatever your are trying to automate and make it even better then you wont have to pay extra

[@ralfelfving]

In what way? Describe a use case and I might pick it up in the future (no guarantee).

[@ITworld-gw9iy]

what is the benefit of using AssistantsAPI using ChatGPT api?

[@elkabeatz135]

so much value, thank you!! Is it possible to add the ability for the chatbot to accept an uploaded file by the chatbot user, such as a CSV, so that it can analyze it using a code interpreter or data analysis tools?" This refers to the chatbot user being able to upload a CSV, not that it's pre-uploaded as part of its knowledge base.

[@ralfelfving]

Yes. I posted some example code at the very bottom of this blog post: medium.com/@ralfelfving/scrape-and-upload-any-site-to-a-gpt-in-2-minutes-1ec5ec127a81

[@EmirFurkanBilici-ei2ru]

Hello sir, what's your vscode color theme name?

[@ralfelfving]

I think I'm running Monokai Pro

[@EmirFurkanBilici-ei2ru]

Thank you. The video was also very helpful. I liked the way you explained all the things. Have a nice day!

[@conradohernanvillagil2764]

Excelent video. Thank you so much.

[@f2ni11]

are assistants cheaper than fine-tuning models for knowledge base custom chatbots?

[@ralfelfving]

I don't know yet, not looked into pricing at all.

[@michaeli5113]

great video.. my question is whats the point of Assistant API when we can just ask same questions in chat gpt

[@ralfelfving]

There's many examples of where ChatGPT isn't the best option. You may want to build a custom chat interface for your company/product/service. Say you run a webshop and want to offer a chat bot that personalize help and product offers from customers embedded into your website. ChatGPT can't do that.

[@fredrikt6980]

Whats the VS code theme?

[@ralfelfving]

Monokai Pro i beliene

[@dawid_dahl]

Thanks! So instead of Completion you use Threads?

[@ralfelfving]

Yes, Assistants make use of threads not completions.

[@aisprint]

Thank you for the great tutorial to start with! One question if we have assistants established through openAI platform and want to use it, only what we have to do is to replace assistant.id with the actual assistants ID??

[@ralfelfving]

Yes, that's one way. I have another tutorial on my channel that allows you to fork an entire frontend, backend and database to chat with assistants (it takes just two-three minutes). You could add a record to your database with the OpenAI assistant ID to chat with it in there.

[@aisprint]

@@ralfelfving Thanks! It would be really helpful and I'll check it.

[@monkeybiz101]

great video! gonna try out your code example now! 👍

[@Tomascastro04]

Thanks! Is it possible to create a bot in the playground and connect it with an API to a whatsapp or message app?

[@ralfelfving]

I haven't tried, but I don't believe so since you'd need to handle authentication to the third party messaging service.

[@Tomascastro04]

@@ralfelfving so is possible to get the api of your own GPTS

[@ralfelfving]

I don't know, I don't have access to GPTs yet.

[@redli2648]

love this tutorial. Can you kindly upload another video with Files? Say I want to upload a list of scientific literature and do Q&A with Assistants API on those papers. This can be an interesting use case!! Also, file is difficult to deal with

[@ralfelfving]

I have a few other videos I want to do first, but you can upload files manually to Assistants if you're ok with not doing it programatically at the moment. Just go to your platform.openai.com account, select the option "Assistants" in the left hand menu, find the assistant in list, and click the "..." menu to the far right. In the little sidebar menu that opens, you can upload a file manually.

[@francisco_williams]

@@ralfelfvingis it possible to upload files programmatically?

[@ralfelfving]

Yes, i shared a video with an example of how: kzbin.info/www/bejne/jJeudWiEgZmVq7ssi=-GJodzVBj-uY1us1 In the video description there's a link to code examples.

[@akka6193]

Hello, nice tutorial. Is it possible to make an assistant that takes fillings from the sec.gov api, analyzes data from, for example, the last quarter of the Tesla share, and the final summary would be displayed on my page?

[@DreamsAPI]

Thank you for the video, please when you can zoom in on your code for us to fully understand your thought process, appreciate very much you sharing your knowledge

[@ralfelfving]

Thanks for the feedback, I will consider this in the future.

[@nosult3220]

Thank you! Building a Mac ap with the AI assistant API.

[@ralfelfving]

Glad to hear, always be building! :)

[@luigitoniolo1719]

Ralf, firstly, thanks a lot to sharing your knowledge! I am building a complex service bot for restaurants and I already have a AI assistant with gpt-3.5-turbo and passing the whole instructions on the prompt like assistant name, description of it and the whole logical service of the restaurant. Do you know if with Assistants API will work better for my case?

[@ralfelfving]

Yes, Assistants would be much better. OpenAI stores the instructions for assistants, and you can upload and store files that contain more context. You can try to spike replacing your existing functionality in the OpenAI API playground at platform.openai.com by creating an assistant and uploading any docs, and conversing with it there. Note that assistants may be more expensive to run, so I suggest you have a look at pricing as well before making your decision.

[@luigitoniolo1719]

Thanks@@ralfelfving ! I'll do some playground on openai website to test it. I just want to know how many tokens I am using to compare with my current solution. Anyway, thank you very much for your work and your patience!

[@danielsddonan8384]

what is the difference between custom gpt and custom assistant

[@ralfelfving]

I'm sure there are several, but the biggest difference that comes to mind is that custom GPTs will be accessible in OpenAI's ChatGPT and distributable in their upcoming GPT store. I don't have access to GPTs yet, so I personally don't know of all the differences.

[@GrygD-d6w]

Thanks for this! Very informative. What approach do you recommend for storing the thread ids if say you build an app that will have multiple users?

[@ralfelfving]

I'm not sure what all options are. I'm currently building a second version of this where I store the assistant, thread, and message information in a DB

[@GrygD-d6w]

@@ralfelfving you mean storing the thread id in a DB? OpenAI takes care of storing the messages now with the Assistant API.

[@ralfelfving]

Yes, the thread IDs. You can store message payloads in the DB as well if you want to pull from your db vs pulling messages from OpenAI, e.g. when you go back to an old thread and want to load the history.

[@streamology]

thank you!

[@ralfelfving]

Glad you liked it! ♥️

[@AnyThingCreativex]

Exactly what I needed, thank you!

[@ralfelfving]

Check my latest tutorial, it's an upgrade -- better assistant handling! :)

[@airevolution23]

Great info!! Can you do one with GPT 4V and TTS API please

[@ralfelfving]

I will look into it, not sure if I'll get to it in the next week tho.

[@airevolution23]

@@ralfelfving appreciate it! And amazing info by the way! You gained a new follower

[@bigbadcatbigbcy2933]

This video was super helpful thank you so much

[@Nilaxification]

Awesome stuff, well done!

[@ralfelfving]

Thank you, glad you enjoyed it!

[@zelaxan]

Thank you so much for the tutorial

[@ralfelfving]

Thanks for taking the time to watch and comment!

[@brainrotjs]

eww commonjs! haha good video :)

[@ralfelfving]

Thanks. 😋 Went in unscripted, one take. Scrappy is the way I roll :)

[@robrita]

@@ralfelfving keep it up bro!! I do code in full enterprise best practices with typescript, linters, prettier, unit test and coverage.. but that's not the point here. Thanks for sharing the CONCEPT and ignore the naysayers :)

[@ralfelfving]

Thanks for sharing the positive vibes and encouragement @robrita, I think there's only been positive sentiment... So far :) I wouldn't be too bothered by negativity, I know some people find this content helpful so that's all I need to know.

[@GrygD-d6w]

What do you mean? All your cutting-edge new js libraries still use nodejs and React under the hood. This is still a very reliable and proven way to build things especially for production and scale.

[@mpaforoufakis]

can i do this in client side js ?

[@ralfelfving]

Yes you can, but I would strongly suggest you have a backend that do do the API calls to OpenAI or you'd expose your OpenAI API key publicly.

[@sauravkharal4016]

Thank you for putting this out so fast

[@ralfelfving]

Thanks for taking the time to watch and comment!

[@i_forget]

Can different users have different threads in this app?

[@ralfelfving]

No, this is a really simple example that doesn't manage users or thread management. I'm working on a tutorial that will cover that.

[@i_forget]

@@ralfelfving I’m looking forward to it!

[@sasukeuchiha-ck4hy]

@@ralfelfving I would also love to see that!

[@hugojohansson3693]

can you use assistent in swiftui?

[@ralfelfving]

Yes, you can.

[@hugojohansson3693]

how? without node sdk? @@ralfelfving

[@davida.m.4978]

and how can I use this in postman?

[@ralfelfving]

Go to OpenAIs documentation for Assistants API, copy the Curl code, ask ChatGPT to translate it to something you can use with Postman and provide the steps you need to take in there.

[@sasukeuchiha-ck4hy]

@@ralfelfving What Curl code in the Docs do you mean?

[@RobBrulinski]

thx!

[@ralfelfving]

You're welcome, thanks for stopping by!

[@PIPSAI]

Loved your tutorial, you deserve 100% retention rate. Do you mind if I make the same content but in python?

[@ralfelfving]

Thank you for the kind words. I dont have monopoly on making content, as a matter of fact I encourage you to create any content that can help others get started and learn. Thanks for taking the time to do so!

[@PIPSAI]

@@ralfelfving Thank you, I was trying to think of a less complex project to present the api and I found that yours was simple yet comprehensive enough to cover everything.

[@ralfelfving]

@PIPSAI sometimes you have to start simple, and go from there. All the best with creating your tutorial!

[@santiagotorres9365]

Hey @@PIPSAI do you mind sharing your python one when its done? Thanks!!

[@PIPSAI]

@@santiagotorres9365 Yup I will. I just had a bunch of homework today so I will do it early tomorrow.

[@jackculpan]

🔥🔥🔥

[@ralfelfving]

♥️♥️♥️

[@박민성-u8c]

Hello, I enjoyed your KZbin video. I'm watching in Korea. Can you also include information related to Python? I'd like to see a code that combines assistants api and function calling! Of course, node.js is also good.. However, async is too difficult for beginners. Thank you for sharing the good video quickly. I also recommended this video to my friends.

[@ralfelfving]

Thank you for watching, and the kind comments. I'll keep producing some OpenAI API content, let's see if it is Assistants + Functions. But it'll likely be in Node. I'm trying to learn more Node, and there's already a healthy Python click in the OpenAI ecosystem -- so many others will create tutorials with that language :) Keep up learning and building!

[@ai-yp]

@@ralfelfvingI agree that less people will have Node tutorials , it`s a good niche and also, the python equivalent I have seen so far, was not running properly yet ;-) But plenty more Python versions will pop. Only a matter of time with OpenAI`s API stuttering most of today. You were quick on the trigger, kudos!

[@ralfelfving]

Be fast, be first :)

[@holdthetruthhostage]

I can't wait for this as I plan to use it for Unreal & Unity

[@GrygD-d6w]

AI npcs?

[@holdthetruthhostage]

@@GrygD-d6w I need it to control game engines as in piloted and just do what I say

[@victorzhang8770]

Thanks for the tutorial! Did you try the file functionality for the assistant API? Could that be a knowledge base for the questions?

[@ralfelfving]

It could. There's a quick code example in the link provided for this tutorial: kzbin.info/www/bejne/jJeudWiEgZmVq7s

[@HarpaAI]

🎯 Key Takeaways for quick navigation: 00:00 🚀 *Introduction to OpenAI's Assistants API* - Introduction to OpenAI's new Assistants API and its key features, - Explanation of the purpose of the tutorial, including the use of threads and the Code Interpreter, - Mention of the availability of resources and documentation. 01:25 🛠️ *Setting up the Environment* - Setting up the development environment using VS Code and creating necessary files, - Mention of creating an environment variable for the OpenAI API key. 04:06 📦 *Importing Dependencies and Creating Functions* - Importing necessary dependencies and explaining the importance of updating the OpenAI npm package, - Introduction to the main functions used in the tutorial. 05:54 🤖 *Creating the Assistant and Defining Its Purpose* - Creating an OpenAI assistant with a specific name and purpose, - Explaining the use of tools and the choice of the GPT-4 1106 model. 08:06 🤝 *Interacting with the Assistant* - Starting the interaction with the assistant, including greeting the user, - Introduction to the concept of threads and state management for ongoing conversations. 10:00 🔄 *Implementing a Polling Mechanism* - Explaining the need for a polling mechanism to check the status of the assistant's tasks, - Describing the process of checking for completion and the need for potential improvements. 11:50 📄 *Retrieving and Displaying Messages* - Retrieving messages from the assistant's thread and focusing on the most recent answer, - Displaying the assistant's response to the user's queries. 13:36 🛑 *Terminating or Continuing the Interaction* - Giving the user the option to terminate the interaction or continue asking more questions, - Explaining the continuation of context in the conversation. 16:24 🧐 *Conclusion and Future Plans* - Summarizing the tutorial and its purpose, - Encouraging users to subscribe for future content and improvements in code and use cases. Made with HARPA AI

[@messengercreator]

and I'll challenge u AI CHAT DEEPAI and OPENAI only create whole game who gonna win

[@PizzaLord]

This is not a math tutor. Math tutors help the students work out the answers. A maths tutor would never just give an answer immediately to a math question because then the student would never learn. Also a math tutor follows a syllabus and does not move on with the syllabus to the next part until the student has demonstrated that they have learnt something. This is just a glorified calculator.

[@ralfelfving]

Good point. I could have taken two seconds longer to name my tutorial file more accurately. But because this is a coding tutorial where the aim is to teach people some basics of how to work with a particular API, I didn't bother, because not naming it correctly is at the end completely irrelevant in that context.

[@PizzaLord]

@@ralfelfving You are missing my point. The very frustrations and limitations that are lot of people are feeling (and also why AI feels over hyped right now) is because you have to lead the AI by the nose by either asking it questions or giving instructions. When teaching some basics of how to work with a particular API examples should have a use case that makes people go "wow" just at the use case. This use case is, sadly like most AI use cases, useless.

[@ralfelfving]

@PizzaLord I disagree with your bar for what makes a useful tutorial. I will continue to sometimes use simplistic use cases knowing that a) most people will learn what the tutorial intended to teach them, b) that they can apply that on whatever "wow" use case they have in their head.

[@brytonkalyi277]

°•°• I believe we are meant to be like Jesus in our hearts and not in our flesh. But be careful of AI, for it is just our flesh and that is it. It knows only things of the flesh (our fleshly desires) and cannot comprehend things of the spirit such as peace of heart (which comes from obeying God's Word). Whereas we are a spirit and we have a soul but live in the body (in the flesh). When you go to bed it is your flesh that sleeps but your spirit never sleeps (otherwise you have died physically) that is why you have dreams. More so, true love that endures and last is a thing of the heart (when I say 'heart', I mean 'spirit'). But fake love, pretentious love, love with expectations, love for classic reasons, love for material reasons and love for selfish reasons that is a thing of our flesh. In the beginning God said let us make man in our own image, according to our likeness. Take note, God is Spirit and God is Love. As Love He is the source of it. We also know that God is Omnipotent, for He creates out of nothing and He has no beginning and has no end. That means, our love is but a shadow of God's Love. True love looks around to see who is in need of your help, your smile, your possessions, your money, your strength, your quality time. Love forgives and forgets. Love wants for others what it wants for itself. Take note, love works in conjunction with other spiritual forces such as faith and patience. We should let the Word of God be the standard of our lives not AI. If not, God will let us face AI on our own and it will cast the truth down to the ground, it will be the cause of so much destruction like never seen before, it will deceive many and take many captive in order to enslave them into worshipping it and abiding in lawlessness. We can only destroy ourselves but with God all things are possible. God knows us better because He is our Creater and He knows our beginning and our end. Our prove text is taken from the book of John 5:31-44, 2 Thessalonians 2:1-12, Daniel 7-9, Revelation 13-15, Matthew 24-25 and Luke 21. Let us watch and pray... God bless you as you share this message to others.

[@oOempulseOo]

You should take your meds

[@JoJoAcrylicArtwork]

Another excellent tutorial/guide thanks Ralf!

[@ralfelfving]

Thanks. But I made a much better tutorial later. Check this for better logic: kzbin.info/www/bejne/eKCcg5aXqpyaobs

[@JoJoAcrylicArtwork]

@@ralfelfving yes just watching both thise now. Always helps to have a visual example to run through in addition to the docs.