Production quality of a 1m+ sub youtuber, which i know you'll be soon with videos like these. Really great vid! You explained everything and even the concept of the knowledge function. Really great! subbed :D
🎯 Key Takeaways for quick navigation: 00:00 🤖 *Custom GPTs and Assistance API are powerful AI features from OpenAI.* 00:28 🧩 *Custom GPTs are built on top of Chat GPT, making it easy for users with Chat GPT Plus accounts to access.* 00:42 🚀 *Custom GPTs are user-friendly and require minimal coding.* 00:56 💡 *Assistance API allows coding for highly flexible applications with custom logic and UI.* 01:24 🔄 *Maintenance is simple for GPTs, but Assistance API requires more responsibility.* 01:37 💰 *Business opportunities may favor Assistance API due to its monetization potential.* 02:31 🧠 *Both options offer knowledge and tool access.* 03:38 📜 *Context handling is automated in both GPTs and Assistance API.* 04:07 📱 *GPTs use Chat GPT interface, limiting UI customization.* 04:34 🛠 *Assistance API provides flexibility in creating custom UIs and features.* 05:14 📊 *Assistance API offers better visibility and analytics of user interactions.* 05:54 🖼 *Image generation and vision are currently available only in GPTs.* 06:09 📂 *Assistance API allows dynamic knowledge updates, not limited to uploaded files.* 07:17 🛠 *Maintenance is simpler for GPTs, while Assistance API requires more management.* 08:40 💸 *Future revenue-sharing potential with successful GPTs, but differentiation is a challenge.* 09:09 💼 *Assistance API offers more business opportunities with potential premium models.* 09:37 🤔 *Choosing between GPTs and Assistance API depends on specific use cases.*
Well done. I don't know what the word is, but I am the kind of learner that has to see all the details, and once that happens I feel like I become quite comfortable and I can really take decisive action. You have really provided this for me, so thank you. I think it would be really cool if you got into some more use cases of more advanced apps or businesses created using GPTs or API assistants. It's nice and all to make stickers and what not, but I would really feel like your grasp can paint a picture of what these things can really do for something far more complex and would have many moving parts to it. You mentioned different GPTs or assistance working with each other, yes I would really love to see what this stuff can really do. What do you think? 😅 Anyways, well done, keep going.
Thanks so much! Yeah with the GPT store coming next week I think I will make a video talking about some ways to make some advanced GPTs that stand out. Thanks for the ideas!
Thanks for the comparison, really useful info here. One note, if you're a free ChatGPT user, you CAN use custom GPTs developed by other people. So, technically, anyone with a ChatGPT free account can interact with your GPT, not only Plus users.
Thank you! They can't automatically turn into assistants, but probably a lot of the time building it will be spent on putting together the necessary prompts, documents that you want to use, etc. Then you can use the same prompts/docs in an assistant, so you can essentially benefit from doing the gpt first.
Great video, glad I found it and your channel! I started building with the older Completions API and was able to keep strict control of pricing. The Assistants API seems much more like a black box, since it automatically controls the whole context, where the token count can quickly add up. And unlike the Completions API, I don't believe the Assistants API returns prompt and completion token counts, which I use to get the exact usage costs my bots are generating. Hopefully they change that, because I have heard horror stories of single prompts costing multiple dollars, if you have a large context. So, for now, I am sticking with Custom GPTs which have quite a bit of functionality with Actions.
This is such a great and important point!! I've definitely spent more than I expected on some assistants queries and I worry about what it might cost if I used a huge knowledge base. Who knows what it will decide it needs to retrieve? I hope they figure out some way to make it more predictable. Thanks for watching!
Great job. Quick question with GPTs if you use your normal ChatGPT API for a third party app when users visit your platform are they accessing the GPT you created or ChatGPT and if not how can I use my GPT to build an app dedicated to my GPT?
If you build a third party app with the ChatGPT API, it is not a GPT. It will use the API key and cost you money in this case - but it is the only way to build something on a separate platform. As far as custom GPTs go - you cannot surface a custom GPT on a third party platform; it is only available inside of ChatGPT. Hope that helps!
Yep! I have a video about assistants API that should help you get started. Currently working on a video showing how to use files with it. Then I'll have another that covers how to deploy a chatbot to your site!
Great video on the differences! Any thoughts on the 40 messages/3hour limit? For example, If I share my custom GPT application with other Plus users, then if someone uses it 40 times in an hour, then I, as the app creator, would ALSO need to wait before prompting ChatGPT again? But if I write my own application, using assistant APIs/Secret-Key (MindStudio, AirTable), then I get around the limit but must pay for every use?
Thanks for watching! Everyone's usage is related to their account only - so others using your GPT won't count against your own cap (but you using your own GPT will). You are right that creating your own assistant would work around the rate limit, but then you'd have to pay the API token cost!
So what is the difference between feeding data to an assistant in say a json file vs using embeddings and a vector db? Like say, if I want to create a discord bot that responds to users based on knowledge... which way would be the right way to go about it? or what are the main differences between these approaches?
Embeddings and a vector DB were the only way to achieve this result prior to the Assistants API being announced. Now you have the option to take either approach. Assistants files are going to be a bit easier since you don't need to manage the vector DB, but that comes at the expense of flexibility and scale, since the amount of data and number of files is limited and you don't have control over how the data is retrieved. With Pinecone DB for example, you can combine keywords and attribute filtering along with the embedding search, so you can potentially get more accurate results. But then you need to write code to generate embeddings, store and retrieve them, etc. I'd probably recommend starting with the assistants files and if you are hitting limitations, explore a more hands-on RAG approach (I have a separate video on this btw). For a discord bot, I'd make the Assistant via the UI, upload files there, and use the ID in my code for creating threads and runs. If you need to refresh data regularly, you could write some code to store that data into files and add it to the assistant. Will be making a video on this soon!
@@VoloBuilds Thank you for your response! I did already use your assistant discord bot actually :) however, I had a few issues; Mainly, gpt4 being too expensive and 3.5 not being good enough since it has to go through all the json data after every query which makes it slow. It also gets confused if there's several files to pull data from. Would you consider perhaps creating a video on how to work with embeddings and a pinecone db? Ideally, I'd like to make a discord bot that would learn from the server chat data (I don't know if that falls in line with what you mentioned earlier), or at least one that works well with 3.5. I really appreciate your responses, I'm a big fan of your channel. Maybe you also want to consider having a link up to support it.
Thanks for the support and ideas :) I am about to dive into assistant api + files for my next video and I have heard some issues with it so I'll be sure to cover that. I do have a video talking about RAG conceptually that I made before the Assistants API was announced, but it's not a hands on tutorial. I may come back to that subject and do a tutorial on it at some point in the future! I think it does make sense to do a diy RAG setup in case you want to index/embed all your Discord messages. That would be hard to keep up to date with assistants files I think.
Great video! I was looking for these answers and it is super clear now. I am building a no code software with a specific use case. Using Chat Gpt was a must for me (though my own interface). This means I necessarily need to know how to code?
I haven't used many no-code platforms, but I expect you'd be able to wire them up to the Assistants API if you want your own interface but don't want to code!
Yes in the sense that you can upload your data or have it call a function to retrieve data when it thinks it is necessary, but it becomes difficult to configure and maintain so if you want to do true RAG I would use the assistants API
I have a test case for you. Do a medium dive on getting a GPT built with a GPT builder and share a simple action. Cut this is video # 1. Video # 2 do a simple dev chat with "Hello GPT4, what are the top five popular colors?" Cut video#2. See which video is more popular? You may find people that watch both. I will watch both. I have a 3rd case to use the Dev API, which is to write a Custom GPT where the end user is someone who does a git clone of your repo and you allow API key parameter/ variable so they get billed. Open source notoriety is better than pay. Especially if you can build a trusted dev team to grow the open source. Do a video on third. Cut video#3. I'll watch all three. Do a video on how the hell we keep up with this text sh*t? This AI is so full of rabbit holes. Down some of these holes were startups likely charging and were crushed with OpenAI over the last month. Someone needs to do a RUclips video on all freemiums that got crushed like this.
I'm not sure, but guessing some players who are already working on chatbots like Voiceflow will try to build on top of the assistants API but I haven't seen it yet. It's all so new! I'm actually exploring some possibilities of building something myself too
Is it not very relevant that ChatGPT and GPT Agents also have this static instruction set constantly there? While with the API assistant you can start with a blank page, thereby a benefit if your good at writing instructions, and bad if you cant write good instructions/prompts. Also data from GPT is used to train the model, while with the API assistant its not.
I really enjoyed this video. Very comprehensive and detailed explanations of the similarities, differences, use cases etc. Thank you for creating this. I have a better understanding of how they work.
i have been trying to replicate my custom GPT which was great at anwering questions from my knowdlege base (but cant share so its useless to me) with an assistant API which supposedly updloaded the same file, used the code interpreter but the answers are totally random and it keeps insisting the info is not in the file!! are you sure they have the same knowledge cabilities?
They should use the same infrastructure under the hood though I can't prove that :) it sounds like something might have gone wrong with your upload or maybe you're pointing at the wrong assistant id? I have a video about configuring the assistants API with your own data that could potentially help: ruclips.net/video/JzxUW0ZT4to/видео.html
have you actually tested it?? I have spent hours comparing the 2 based on the same knowledge text file, GPT gives perfect replies every single time, the assistant does not read the text file unless i explicitly ask it to do so repeatedly and even then it adds things like imaginary contact numbers @@VoloBuilds
Huh, that's weird - I am not sure why that would happen. When I tested it, things worked pretty well and consistent across the two. Maybe they changed something since I last played around with it? Maybe there is something buried in the prompt that is causing it to misbehave or instructions in the GPT that make it better? Not sure, just throwing ideas out there
Hey, I have seen someone saying it is not good to totally rely on openai API especially if you are an indie developer, it may stop working or shut down bcoz of the recent drama in openai leadership.....what do you think buddy?
This is a great observation and although the drama seems to have settled down, the API, models, and pricing could all change in the future so it's good to build in a way that is resilient to such changes. Thankfully there are competitors such as Anthropic that offer LLM APIs as well so I think we have options. I expect we will see some open source version of the assistants API at some point because the abstractions are really great and can be applied generally. The other major thing to consider and keep in mind is making sure to build stuff that goes beyond being a simple wrapper around OpenAI. I think a lot of startups that were simply feeding PDFs and files into ChatGPT are effectively out of business now with the assistants API so it's important to build stuff that solves specific problems for your target customers and does it in a way that OpenAI isn't likely to absorb into their core offering.
They actually make the core rules available for open and public use via the SRD which is what I uploaded! available here: dnd.wizards.com/resources/systems-reference-document
Hey, thanks for leaving this comment - I think a lot of people are in the same boat! I would love to make some content to help out people like you - but it's sometimes hard for me to think of what would be some good basic questions or subjects to cover. I'm curious, what are some of the initial questions you have or concepts you'd want to see explained?
@@VoloBuilds thanks for your interest in wanting to know how best to help me, that means a lot. Well, I don't know exactly what to ask, but I was thinking; if one could share a story of how he/she got started with AI, that would be a solid bedrock for folks like me who are so ignorant of the subject they don't even know the right questions to start with. You got a new follower. Thanks for the knowledge/information you've been sharing.
So there isn't exactly a "GPT API" - so you might be referring to 2 potential things. Assistants API, yes you need to pay for all the API calls. For Actions invoked by a custom GPT, it depends on what API you are calling (it won't be the OpenAI API) - so if you use something public, it could be free - or if you create your own custom actions, you'll need to find a way to host the API yourself so likely need to pay for hosting costs. I'll be making a video soon regarding deploying your own APIs for custom actions.
Just developing and running few tests with Assistant API cost $$$. Imagine if the app becomes popular… the cost is still prohibitive as many tokens are flying back and forth for simple use case. Better have strong business case.
This is such a good point. I might end up doing a video deep diving costs because it's so hard to predict right now - very much a black box and I'm seeing expenses rack up too.
You can have a custom GPT use Actions to send data elsewhere but you cannot let users access and chat with the GPT anywhere except the ChatGPT interface. You would need Assistants API to make use of OpenAI outside of ChatGPT - then you'll have a lot of flexibility!
Production quality of a 1m+ sub youtuber, which i know you'll be soon with videos like these. Really great vid! You explained everything and even the concept of the knowledge function. Really great! subbed :D
Wow, thank you so much!! 🤩
🎯 Key Takeaways for quick navigation:
00:00 🤖 *Custom GPTs and Assistance API are powerful AI features from OpenAI.*
00:28 🧩 *Custom GPTs are built on top of Chat GPT, making it easy for users with Chat GPT Plus accounts to access.*
00:42 🚀 *Custom GPTs are user-friendly and require minimal coding.*
00:56 💡 *Assistance API allows coding for highly flexible applications with custom logic and UI.*
01:24 🔄 *Maintenance is simple for GPTs, but Assistance API requires more responsibility.*
01:37 💰 *Business opportunities may favor Assistance API due to its monetization potential.*
02:31 🧠 *Both options offer knowledge and tool access.*
03:38 📜 *Context handling is automated in both GPTs and Assistance API.*
04:07 📱 *GPTs use Chat GPT interface, limiting UI customization.*
04:34 🛠 *Assistance API provides flexibility in creating custom UIs and features.*
05:14 📊 *Assistance API offers better visibility and analytics of user interactions.*
05:54 🖼 *Image generation and vision are currently available only in GPTs.*
06:09 📂 *Assistance API allows dynamic knowledge updates, not limited to uploaded files.*
07:17 🛠 *Maintenance is simpler for GPTs, while Assistance API requires more management.*
08:40 💸 *Future revenue-sharing potential with successful GPTs, but differentiation is a challenge.*
09:09 💼 *Assistance API offers more business opportunities with potential premium models.*
09:37 🤔 *Choosing between GPTs and Assistance API depends on specific use cases.*
What's exactly what I was lookin for. Thank you man!
Was hoping to find such a great comparison. Well done. Many thanks!
Thank you! Glad it was helpful! :)
Well done. I don't know what the word is, but I am the kind of learner that has to see all the details, and once that happens I feel like I become quite comfortable and I can really take decisive action. You have really provided this for me, so thank you. I think it would be really cool if you got into some more use cases of more advanced apps or businesses created using GPTs or API assistants. It's nice and all to make stickers and what not, but I would really feel like your grasp can paint a picture of what these things can really do for something far more complex and would have many moving parts to it. You mentioned different GPTs or assistance working with each other, yes I would really love to see what this stuff can really do. What do you think? 😅 Anyways, well done, keep going.
Thanks so much! Yeah with the GPT store coming next week I think I will make a video talking about some ways to make some advanced GPTs that stand out. Thanks for the ideas!
Thanks for the comparison, really useful info here.
One note, if you're a free ChatGPT user, you CAN use custom GPTs developed by other people. So, technically, anyone with a ChatGPT free account can interact with your GPT, not only Plus users.
Yep, you are right! When it was originally released (when I made the video) it was not the case, but they made it more broadly available now :)
Excellent video. Thank you. I was curious if you spent a lot of time building these custom gpts if you could later turn them into assistants
Thank you! They can't automatically turn into assistants, but probably a lot of the time building it will be spent on putting together the necessary prompts, documents that you want to use, etc. Then you can use the same prompts/docs in an assistant, so you can essentially benefit from doing the gpt first.
Copy that, great advice 😁. do you use a specific tool for saving all your prompts?
@@BrianFirth Generally it's just a part of my codebase so committed to github repos
Great video, glad I found it and your channel! I started building with the older Completions API and was able to keep strict control of pricing. The Assistants API seems much more like a black box, since it automatically controls the whole context, where the token count can quickly add up. And unlike the Completions API, I don't believe the Assistants API returns prompt and completion token counts, which I use to get the exact usage costs my bots are generating. Hopefully they change that, because I have heard horror stories of single prompts costing multiple dollars, if you have a large context. So, for now, I am sticking with Custom GPTs which have quite a bit of functionality with Actions.
This is such a great and important point!! I've definitely spent more than I expected on some assistants queries and I worry about what it might cost if I used a huge knowledge base. Who knows what it will decide it needs to retrieve? I hope they figure out some way to make it more predictable. Thanks for watching!
Great Video! this really helped me figure out the next steps for my project! i was hitting a brick wall
Awesome!! Glad it helped you!!
Is there a GitHub repo where an assistent is there with an upload button to update knowledge and than chat about it?
I don't know of one, but that seems like a very useful thing to create. Let me know if you find something like that!
Great job. Quick question with GPTs if you use your normal ChatGPT API for a third party app when users visit your platform are they accessing the GPT you created or ChatGPT and if not how can I use my GPT to build an app dedicated to my GPT?
If you build a third party app with the ChatGPT API, it is not a GPT. It will use the API key and cost you money in this case - but it is the only way to build something on a separate platform. As far as custom GPTs go - you cannot surface a custom GPT on a third party platform; it is only available inside of ChatGPT. Hope that helps!
Can you make a Video how to write a Chatbot with the assitent API? @11:00
Yep! I have a video about assistants API that should help you get started. Currently working on a video showing how to use files with it. Then I'll have another that covers how to deploy a chatbot to your site!
What a great video, comprehensive and well structured 👍
Thank you so much!! :) Glad it was helpful!
Great video on the differences! Any thoughts on the 40 messages/3hour limit?
For example, If I share my custom GPT application with other Plus users, then if someone uses it 40 times in an hour, then I, as the app creator, would ALSO need to wait before prompting ChatGPT again?
But if I write my own application, using assistant APIs/Secret-Key (MindStudio, AirTable), then I get around the limit but must pay for every use?
Thanks for watching! Everyone's usage is related to their account only - so others using your GPT won't count against your own cap (but you using your own GPT will). You are right that creating your own assistant would work around the rate limit, but then you'd have to pay the API token cost!
dude, what a great video, already subscribed. Super clear, thanks for the info!!!!
Thank you so much!! 🤩
You answered my questions! Thank you!
That's awesome :) happy to help!
Great video. Thanks!
Glad you liked it!
So what is the difference between feeding data to an assistant in say a json file vs using embeddings and a vector db? Like say, if I want to create a discord bot that responds to users based on knowledge... which way would be the right way to go about it? or what are the main differences between these approaches?
Embeddings and a vector DB were the only way to achieve this result prior to the Assistants API being announced. Now you have the option to take either approach.
Assistants files are going to be a bit easier since you don't need to manage the vector DB, but that comes at the expense of flexibility and scale, since the amount of data and number of files is limited and you don't have control over how the data is retrieved. With Pinecone DB for example, you can combine keywords and attribute filtering along with the embedding search, so you can potentially get more accurate results. But then you need to write code to generate embeddings, store and retrieve them, etc.
I'd probably recommend starting with the assistants files and if you are hitting limitations, explore a more hands-on RAG approach (I have a separate video on this btw). For a discord bot, I'd make the Assistant via the UI, upload files there, and use the ID in my code for creating threads and runs. If you need to refresh data regularly, you could write some code to store that data into files and add it to the assistant. Will be making a video on this soon!
@@VoloBuilds Thank you for your response! I did already use your assistant discord bot actually :) however, I had a few issues; Mainly, gpt4 being too expensive and 3.5 not being good enough since it has to go through all the json data after every query which makes it slow. It also gets confused if there's several files to pull data from.
Would you consider perhaps creating a video on how to work with embeddings and a pinecone db? Ideally, I'd like to make a discord bot that would learn from the server chat data (I don't know if that falls in line with what you mentioned earlier), or at least one that works well with 3.5.
I really appreciate your responses, I'm a big fan of your channel. Maybe you also want to consider having a link up to support it.
Thanks for the support and ideas :) I am about to dive into assistant api + files for my next video and I have heard some issues with it so I'll be sure to cover that. I do have a video talking about RAG conceptually that I made before the Assistants API was announced, but it's not a hands on tutorial. I may come back to that subject and do a tutorial on it at some point in the future!
I think it does make sense to do a diy RAG setup in case you want to index/embed all your Discord messages. That would be hard to keep up to date with assistants files I think.
Thank you for creating this!
My pleasure! Thanks for watching! 😄
Great video! I was looking for these answers and it is super clear now. I am building a no code software with a specific use case. Using Chat Gpt was a must for me (though my own interface). This means I necessarily need to know how to code?
I haven't used many no-code platforms, but I expect you'd be able to wire them up to the Assistants API if you want your own interface but don't want to code!
Is possible to go with RAG in GPTs?
Yes in the sense that you can upload your data or have it call a function to retrieve data when it thinks it is necessary, but it becomes difficult to configure and maintain so if you want to do true RAG I would use the assistants API
I have a test case for you. Do a medium dive on getting a GPT built with a GPT builder and share a simple action. Cut this is video # 1. Video # 2 do a simple dev chat with "Hello GPT4, what are the top five popular colors?" Cut video#2. See which video is more popular? You may find people that watch both. I will watch both. I have a 3rd case to use the Dev API, which is to write a Custom GPT where the end user is someone who does a git clone of your repo and you allow API key parameter/ variable so they get billed. Open source notoriety is better than pay. Especially if you can build a trusted dev team to grow the open source. Do a video on third. Cut video#3. I'll watch all three. Do a video on how the hell we keep up with this text sh*t? This AI is so full of rabbit holes. Down some of these holes were startups likely charging and were crushed with OpenAI over the last month. Someone needs to do a RUclips video on all freemiums that got crushed like this.
Thanks for all the great ideas!!
Any no code front-end tools that may work well for Assistants API use cases?
I'm not sure, but guessing some players who are already working on chatbots like Voiceflow will try to build on top of the assistants API but I haven't seen it yet. It's all so new! I'm actually exploring some possibilities of building something myself too
Great video! Thank you. Very clear and concise.
Glad it was helpful!
Thx for a great explanation. You really broke it down and showed the strengths and cool ways we can use each tool. Now I kinda get it. ^_^
Glad it was helpful! Thanks for watching!
great video
Thanks for watching!!
Is it not very relevant that ChatGPT and GPT Agents also have this static instruction set constantly there? While with the API assistant you can start with a blank page, thereby a benefit if your good at writing instructions, and bad if you cant write good instructions/prompts. Also data from GPT is used to train the model, while with the API assistant its not.
I really enjoyed this video. Very comprehensive and detailed explanations of the similarities, differences, use cases etc.
Thank you for creating this. I have a better understanding of how they work.
Thanks for watching!! Glad it helped you 😄
i have been trying to replicate my custom GPT which was great at anwering questions from my knowdlege base (but cant share so its useless to me) with an assistant API which supposedly updloaded the same file, used the code interpreter but the answers are totally random and it keeps insisting the info is not in the file!! are you sure they have the same knowledge cabilities?
They should use the same infrastructure under the hood though I can't prove that :) it sounds like something might have gone wrong with your upload or maybe you're pointing at the wrong assistant id? I have a video about configuring the assistants API with your own data that could potentially help: ruclips.net/video/JzxUW0ZT4to/видео.html
have you actually tested it?? I have spent hours comparing the 2 based on the same knowledge text file, GPT gives perfect replies every single time, the assistant does not read the text file unless i explicitly ask it to do so repeatedly and even then it adds things like imaginary contact numbers @@VoloBuilds
Huh, that's weird - I am not sure why that would happen. When I tested it, things worked pretty well and consistent across the two. Maybe they changed something since I last played around with it? Maybe there is something buried in the prompt that is causing it to misbehave or instructions in the GPT that make it better? Not sure, just throwing ideas out there
Hey, I have seen someone saying it is not good to totally rely on openai API especially if you are an indie developer, it may stop working or shut down bcoz of the recent drama in openai leadership.....what do you think buddy?
This is a great observation and although the drama seems to have settled down, the API, models, and pricing could all change in the future so it's good to build in a way that is resilient to such changes. Thankfully there are competitors such as Anthropic that offer LLM APIs as well so I think we have options.
I expect we will see some open source version of the assistants API at some point because the abstractions are really great and can be applied generally.
The other major thing to consider and keep in mind is making sure to build stuff that goes beyond being a simple wrapper around OpenAI. I think a lot of startups that were simply feeding PDFs and files into ChatGPT are effectively out of business now with the assistants API so it's important to build stuff that solves specific problems for your target customers and does it in a way that OpenAI isn't likely to absorb into their core offering.
@@VoloBuilds thank you for the valuable points...
When uploading the DnD rule set, isn't there a copyright issue?
They actually make the core rules available for open and public use via the SRD which is what I uploaded! available here: dnd.wizards.com/resources/systems-reference-document
I'm interested in AI, but I'm completely new to it, everything sounds like a new language, dont know where to start from.
Hey, thanks for leaving this comment - I think a lot of people are in the same boat! I would love to make some content to help out people like you - but it's sometimes hard for me to think of what would be some good basic questions or subjects to cover. I'm curious, what are some of the initial questions you have or concepts you'd want to see explained?
@@VoloBuilds thanks for your interest in wanting to know how best to help me, that means a lot.
Well, I don't know exactly what to ask, but I was thinking; if one could share a story of how he/she got started with AI, that would be a solid bedrock for folks like me who are so ignorant of the subject they don't even know the right questions to start with.
You got a new follower.
Thanks for the knowledge/information you've been sharing.
Very valuable vid, Thanks !
Glad you found it helpful! 😄
For the usage of gpt API, do I need to pay for the usage of the api calls?
So there isn't exactly a "GPT API" - so you might be referring to 2 potential things. Assistants API, yes you need to pay for all the API calls. For Actions invoked by a custom GPT, it depends on what API you are calling (it won't be the OpenAI API) - so if you use something public, it could be free - or if you create your own custom actions, you'll need to find a way to host the API yourself so likely need to pay for hosting costs. I'll be making a video soon regarding deploying your own APIs for custom actions.
Just developing and running few tests with Assistant API cost $$$. Imagine if the app becomes popular… the cost is still prohibitive as many tokens are flying back and forth for simple use case. Better have strong business case.
This is such a good point. I might end up doing a video deep diving costs because it's so hard to predict right now - very much a black box and I'm seeing expenses rack up too.
Enteresan. Bu aksani nerede duysam taniyorum.🙂
Super useful vid. Thank you!!
Thanks for watching! :)
Nice video, thank you.
Thanks for watching!!
is it possible to integrate a Custom GPTs to anywhere ?
You can have a custom GPT use Actions to send data elsewhere but you cannot let users access and chat with the GPT anywhere except the ChatGPT interface. You would need Assistants API to make use of OpenAI outside of ChatGPT - then you'll have a lot of flexibility!
@@VoloBuilds Can i create same thing as custom gpts with assistant api
There are differences as described in this video, but at a high level, you can use knowledge + actions and build a chat bot using the assistants API
@@VoloBuilds thanks
thx this video explains it very well
Thanks for watching! :)
u are awesome dude!
🥰 thank you bro
No problem 😎 thanks for watching!
great content! thanks. but the hands, the hands.... :)
I lost all the love in the world to all the tragedies when the guest actor
:huh cat meme:
Skip the cheesy stock footage clips which add zero value to your video
Thanks for the input; I added them to enhance the visual storytelling of the video which helps some folks retain what is being said
I'm with volo
I'm also with Volo. Do your thing bro.... Very useful video imo.
why u mad bro
are you dancing or trying to explain a thing?
😂
cringe!!!