watching you do it makes me want to build AI automation, even though I have ZERO coding knowledge. The future is about optimizing your life's processes--especially at work. This is so cool!
I've been looking for this kind of information for a month now, and I spend at least an hour a day. And the universe heard me. Many thanks for this video.
I’ve been working with RAG and using Airtable for a while, and although your system is impecable for a small number of queries, if you ramp up the number of records, the cost will not be efficient. A better way would be building simple node.js script and hosting it on Heroku or AWS, then write an Airtable automation that is triggered when the record goes into the view you mentioned, which would run a simple script that triggers the script in node.js to push, get and have both of your db in sync, while respecting your criteria.
The channel is about low code or no code automation. For personal usecase maybe we can deal with a small script but when it's complex and changes based on requirement then this is best.
there are some valid points here but its all comes down to the details, this video is more for concept that production as well. thanks for your thoughts.
I literally just draw up a diagram to build a “knowledge base” like this last week. I need to work faster and start releasing my own videos haha. Good stuff
@Stephen - actually a video walk through of a user asking your NCA Toolkit CustomGPT would help some folks. This is where a person can figure out the many specific use cases for your super useful tools!
I understand that the limitatations of the data input is the issue of LLM models here. But Claude just came out with Claude Projects recently with custom prompts and pdf file, image or even videos as inputs. How does your system offer more benefits than Claude projects? Thanks for replying.
Thanks, Stephen. Just random questions popped up. how do make sure the search terms are exactly matching against the vector database? how many embeddings are perfectly matched? how much accuracy is maintained based on matched and unmatched percentage of user query to vector database?
Holy shit!! 🤯🤯🤯 One question Stephen. Why the content overlap in the "chunks"? Is it so the "embeddings" encryptions contain "linked" content with the previous / next number, so the LLM understands the content is connected somehow? And if "yes", how did you figure that out?
Yes, i figured it out because i tried to add a larger file and it was too big. So i wondered what to do and wondered if there a strategy for breaking them up.
The value added is that the ai systems that you have implemented into your business will now have respond with more concise responses and another benifical part is that it will be self learning.
I provide the templates you see in the video, and of course we provide support as well. I don't have what I'd call formal training on this specific subject yet. But that will develop.
This is great content and shows how versatile Make is. But to be honest, this is not the best tool for building a rag with a vector store and embeddings. I would consider n8n or better yet Flowise. Best tool for each job. 😊
Is Slack the best interface for this sort of thing I would love to build out a similar system for a client, but I’m not sure I wanna onboard them to Slack. Would be cool if there was an extension in Airtable…
I think it works well but i could see other things work well too, like sms, voice, email event interactive avatars, telegram and whatapps could even work too
You said "Explain the API" when talking about the "hidden API". You should clarify that process might be illegal in countries like America. Even if you can see the API calls, you would have to reverse engineer them to get the data you want and this definitely is not a beginner process especially if you don't know anything about coding.
watching you do it makes me want to build AI automation, even though I have ZERO coding knowledge. The future is about optimizing your life's processes--especially at work. This is so cool!
I've been looking for this kind of information for a month now, and I spend at least an hour a day. And the universe heard me. Many thanks for this video.
💪💪
@@StephenGPope WHEN YOU WILL PUBLISH THIS AS A TEAMPLATE FOR A KNOWLEDGE BASE?
I’ve been working with RAG and using Airtable for a while, and although your system is impecable for a small number of queries, if you ramp up the number of records, the cost will not be efficient.
A better way would be building simple node.js script and hosting it on Heroku or AWS, then write an Airtable automation that is triggered when the record goes into the view you mentioned, which would run a simple script that triggers the script in node.js to push, get and have both of your db in sync, while respecting your criteria.
The channel is about low code or no code automation.
For personal usecase maybe we can deal with a small script but when it's complex and changes based on requirement then this is best.
What a great answer, if you can help us with a video tutorial, I would be very grateful. Thanks.
there are some valid points here but its all comes down to the details, this video is more for concept that production as well. thanks for your thoughts.
I literally just draw up a diagram to build a “knowledge base” like this last week. I need to work faster and start releasing my own videos haha.
Good stuff
Speed is the key to some of this stuff 👌
amazing ideas!
Thank you Stephen!
Another fantastic model! Thx 🙏
This is exciting. Analyzing everything now! Thank you. Watching for the classroom!
Awesome! Traveling today so a little later!
Awesome demonstration 😍, I use Flowise and Make for those use cases, but it was very interesting to see your process!
Nice flowise looks cool
This is awesome. THANK YOU
amazing video!
Thanks Stephen
🙏🙏
What app are you using to make your mouse arrow pointer? Thank you for your response!
So much value in this video!!
Stephen, love the content. What's the live screen recording annotation tool you're using? Thanks.
@Stephen - actually a video walk through of a user asking your NCA Toolkit CustomGPT would help some folks. This is where a person can figure out the many specific use cases for your super useful tools!
Hello, great videos. What app are you using to make this kind of video? Camtasia?
I understand that the limitatations of the data input is the issue of LLM models here. But Claude just came out with Claude Projects recently with custom prompts and pdf file, image or even videos as inputs. How does your system offer more benefits than Claude projects? Thanks for replying.
But you can’t make an API call to Claude projects.
Automation.
A few others have mentioned, projects are cool but dont work with automation or apis
Great automation! Really looking forward to creating this one. Is there any way to get in contact with you directly?
Yea right here or skool.com/no-code-architects
Thanks, Stephen. Just random questions popped up. how do make sure the search terms are exactly matching against the vector database? how many embeddings are perfectly matched? how much accuracy is maintained based on matched and unmatched percentage of user query to vector database?
Holy shit!! 🤯🤯🤯
One question Stephen. Why the content overlap in the "chunks"? Is it so the "embeddings" encryptions contain "linked" content with the previous / next number, so the LLM understands the content is connected somehow?
And if "yes", how did you figure that out?
Yes, i figured it out because i tried to add a larger file and it was too big. So i wondered what to do and wondered if there a strategy for breaking them up.
Incredible.
Great video. Not enough on this topic
I dont understand the value added here
Its a system for creating customized knowledge basis for companies.
The value added is that the ai systems that you have implemented into your business will now have respond with more concise responses and another benifical part is that it will be self learning.
Feels like a lot of the automations I have build following your tutorials were just the prepwork.
Thats a good way to look at it!
Do you teach how to put this together in the community?
I provide the templates you see in the video, and of course we provide support as well. I don't have what I'd call formal training on this specific subject yet. But that will develop.
He uses the term "rational database". Does he really mean "relational database?" or is there some other kind of database that is rational?
Just an error on my part
This is great content and shows how versatile Make is. But to be honest, this is not the best tool for building a rag with a vector store and embeddings. I would consider n8n or better yet Flowise. Best tool for each job. 😊
Is Slack the best interface for this sort of thing I would love to build out a similar system for a client, but I’m not sure I wanna onboard them to Slack.
Would be cool if there was an extension in Airtable…
I think it works well but i could see other things work well too, like sms, voice, email event interactive avatars, telegram and whatapps could even work too
You said "Explain the API" when talking about the "hidden API". You should clarify that process might be illegal in countries like America. Even if you can see the API calls, you would have to reverse engineer them to get the data you want and this definitely is not a beginner process especially if you don't know anything about coding.