Navigate to key moments👇 made via tubestamp.com 0:00 - Importance of setting up prompts in AI software 1:51 - Leveraging playground within OpenAI dashboard 3:01 - Understanding variables like Max tokens and temperature 4:57 - Structuring prompts for software outputs 6:54 - Importance of minimal data input for better performance 8:45 - Guidelines for structuring email response prompts 10:07 - Importance of specific prompts for consistent outputs Recap by TubeStamp ✏️
This is excellent, my guy. Can you make more videos like this in Playground and maybe even how you're implementing in code or optimizing token usage. 🙌
How complicated do you get? Do you ever process an input more than once....like processing with a JSON schema and output format to pull keywords (parts or actual words) and output that as a variables to use with a second run where you produce the natural language output? Is that overkill?
Yes, definitely, additional processes may be needed to ensure consistent outputs at scale. I talk more about this logic in this video: ua-cam.com/video/L_ENjBxRI_Q/v-deo.html
Can you explain what are the differences between "OpenAI's API" and "OpenAI's Assistant API", form my understanding it seems like Open AI is providing two types of API products. Not sure which one is more ideal to use for getting help in creating complex software for software engineers, basically making software and which API is better to use? I can't make up my mind which one I should go for. Thanks.
I would choose using the OpenAI API rather than the assistant. This is because the cost associated with using the assistant is very high, and for most use cases, one can accomplish the desired output with the OpenAI API.
Nice. Isn’t the max tokens the total tokens in both directions? So the Prompt + Response. Why would I ever limit this? I never want a response to get truncated so why bother lowering this limit?
Apologies, I'm not familiar with chat gpt API, I'm just an old school software engineer, but what's related to software and API here? You've basically shown how to use an application, and while it's using it sophisticatedly, I didn't get when you connect it to a real piece of software, or where is the API definition for hooking into it from a real piece of software?
The purpose of this video is to show you how to create prompts for the OpenAI API. If you wish to connect to software, I would suggest using VS Code and a backend like Google Cloud Functions.
I thought it was your system prompt where you are supposed to go into detail about what it is they model needs to do, and the user prompt is just where you hand in the input?
With the API, yes, but through a no-code way like Zapier, I believe no. But if you fine-tune your prompts well enough and provide previous context, the answers should be much more accurate!
Learned a little more, Cheers Corbin, I owe you a beer. Know a little secret now. hehe. Not mentioned in your vid. Thought i would expand on your code for fun. Added this at the end. " I want you to use this format response every time, but when I type in a different name I want you to change the email body name in that response. And change that new email response with a different name, into a different language. " The pieces are all coming together now.
Navigate to key moments👇
made via tubestamp.com
0:00 - Importance of setting up prompts in AI software
1:51 - Leveraging playground within OpenAI dashboard
3:01 - Understanding variables like Max tokens and temperature
4:57 - Structuring prompts for software outputs
6:54 - Importance of minimal data input for better performance
8:45 - Guidelines for structuring email response prompts
10:07 - Importance of specific prompts for consistent outputs
Recap by TubeStamp ✏️
I'd like to award you with a digital Labor Day medal for your hard work.
thank you so much Corbin for these videos and good luck on your software journeys!I wish you luck and prosperity
Hey! Is it free to use chatgpt playground?
This is excellent, my guy. Can you make more videos like this in Playground and maybe even how you're implementing in code or optimizing token usage. 🙌
Definitely! Check out a more recent video going over prompt engineering:
ua-cam.com/video/0tHH7EPBy9M/v-deo.html
Thank you! 🙏🏾
Very wise choice to wait to release the genie. But can’t wait for it, love your genuinely great content
How complicated do you get? Do you ever process an input more than once....like processing with a JSON schema and output format to pull keywords (parts or actual words) and output that as a variables to use with a second run where you produce the natural language output? Is that overkill?
Yes, definitely, additional processes may be needed to ensure consistent outputs at scale. I talk more about this logic in this video:
ua-cam.com/video/L_ENjBxRI_Q/v-deo.html
Love your content man. It’s very useful. Corbin we need a video on how to use OAuth in building GPT’s
Thanks! Check this video out; I show how to integrate OAuth with GPTs:
ua-cam.com/video/6HFp0ISO4XI/v-deo.html
Can you explain what are the differences between "OpenAI's API" and "OpenAI's Assistant API", form my understanding it seems like Open AI is providing two types of API products. Not sure which one is more ideal to use for getting help in creating complex software for software engineers, basically making software and which API is better to use? I can't make up my mind which one I should go for. Thanks.
I would choose using the OpenAI API rather than the assistant. This is because the cost associated with using the assistant is very high, and for most use cases, one can accomplish the desired output with the OpenAI API.
Nice. Isn’t the max tokens the total tokens in both directions? So the Prompt + Response. Why would I ever limit this? I never want a response to get truncated so why bother lowering this limit?
Apologies, I'm not familiar with chat gpt API, I'm just an old school software engineer, but what's related to software and API here? You've basically shown how to use an application, and while it's using it sophisticatedly, I didn't get when you connect it to a real piece of software, or where is the API definition for hooking into it from a real piece of software?
The purpose of this video is to show you how to create prompts for the OpenAI API. If you wish to connect to software, I would suggest using VS Code and a backend like Google Cloud Functions.
I thought it was your system prompt where you are supposed to go into detail about what it is they model needs to do, and the user prompt is just where you hand in the input?
Check out a more recent tutorial I did on Prompt Engineering:
ua-cam.com/video/0tHH7EPBy9M/v-deo.html
I'm fully subscribed to you Corbin. Thank you for this content.
Great video corbin
Thanks!
Thank you for this video. Really helped me out ^^
Can you set the temperature via zapier (assistant or assistant api)?
With the API, yes, but through a no-code way like Zapier, I believe no. But if you fine-tune your prompts well enough and provide previous context, the answers should be much more accurate!
Thanks buddy
Learned a little more, Cheers Corbin, I owe you a beer. Know a little secret now. hehe. Not mentioned in your vid. Thought i would expand on your code for fun. Added this at the end. " I want you to use this format response every time, but when I type in a different name I want you to change the email body name in that response. And change that new email response with a different name, into a different language. " The pieces are all coming together now.
🤝 🍻