Building AI and ML applications with Xano using Embeddings
ฝัง
- เผยแพร่เมื่อ 5 ส.ค. 2024
- Vector Embeddings in Xano, powered by pgVector, allow you to store embeddings inside of your Xano database and use them in queries. Vector embeddings are a crucial element in using AI and ML models to determine complex data relationships and semantics. In this video, we'll go over how vectors work and show you how to use them in Xano.
Documentation: docs.xano.com/database/field-...
00:00 Intro
01:03 The Vector Database Field
03:53 Generating Embeddings in a Function Stack
05:30 Indexing a Vector Field
06:44 Using Embeddings in a Function Stack
13:48 A Visual Vectors Walkthrough
16:01 Outro
Xano - The Fastest No Code Backend Dev Platform
go.xano.co/3UZPXxH
Twitter
/ nocodebackend
Subscribe to Xano's Channel for Weekly New Content
/ nocodebackend - วิทยาศาสตร์และเทคโนโลยี
This is amazing. Next Step would be how to create a chat memory like inhouse or connecting with REDIS.
This is amazing, well done Xano team
That’s amazing! I’ve been using Pinecone with Xano, but the process wasn’t great. Having this natively integrated in Xano is truly a game changer 👏🏼
Vectors + add-ons are like a super power, makes it so easy to create queries and return related data.
Was just thinking the same, I am using Pinecone and thinking of moving all to XANO sicne it is my DB. I noticed though in Pinecone we process in chunks, and in this example, it is one full document. Did you apply same process? How is in terms of performance? Ideally I would like to keep all in XANO.
@@thejorgemarques unfortunately, I didn’t get the chance to try it fully yet, so I’m not certain how it will perform. But I will test it out very soon..!
we're in the future ))
Great tutorial Chris, I was able to follow along and get this working quite easily. I think I might have spotted one mistake though. In the video you say that we should sort the results of the embedding search in ascending order. This would mean we get results from least to most relevant. I tested this out and found that ascending order gives me irrelevant results and descending order (highest similarity value to lowest) works as expected. Maybe you could update the docs to clarify this to people?
Hey Greg you are correct, sorting the inner product with an ascending sort will return the most dissimilar results. The reason Chris mentioned to use the ascending sort is because this is a requirement for the query to utilise the vector index (the index wont be utilised with a descending sort), he has just applied the wrong eval in this scenario. In order to query using the index + return the most similar you can apply the negative inner product with an ascending sort.
im struggling with the first step generating the embeddings. How can i simulate a test in the db trigger function stack to see what I am returning to work in the function stack.
Hey, Daniel! If your trigger is set to run on updates, you can just edit literally any field of a record and then refresh your table view, and you should see the updated embeddings.
@@nocodebackend Hmm.. I have update and I see in the trigger request history that its working. When I create a blank new record I also see the embeddings column get updated with all 00000s.
But when I try to generate an embedding for a field that is already updated nothing happens in my table.
So that is why I want to see whats happening in the function stack. Im not sure if I have an error with my external api request to open AI or if I have my edit record in embeddings setup incorrectly.
@@Danielatblackpeak I see what you mean, thank you. The best way to know for sure is to use Run & Debug to simulate a trigger. Here's a quick Loom going over how that works. If you need more help here, feel free to open up a support chat with us by clicking the option in the lower-left corner anywhere in Xano.
www.loom.com/share/9ec8fc6dc57649c293cfb3452f882fe4?sid=4e2c81be-0339-4f04-94df-49c2e952f1ed
@@nocodebackend Incredibly helpful. Thank you.
How can we make APIs that perform CRUD opration, and how can we use this API in my realworld App and web, I mean How can we add this api in firebase with webflow and flutterflow too
Hey there! Check-out some of our other videos - we break down exactly how to do this. We also have some videos that show you how to connect to webflow and flutterflow specifically! Reach out to support or on our community if you are running into a specific issue.
@@LachlanXano Hey, bor my database will be firebase, so can we inegrate our APIs in firebase.
And if I integrate my APIs in firebase, so we need add realtime database
Hi Chris, thank you for the knowledge you provide, the community would appreciate it very much if you make a tutorial on how we can resell Gemini tokens now that this AI is booming, and as Xano can help us manage that process in our app, thanks in advance, successes friend
Hey there, we do have video showing you how you can issue credits to users - it was a stripe workshop hosted by Liz you can find it here on youtube! I like the idea of a video showing you how to track the cost of api charges per user and deduct these from a users credits / apply a markup. I will see what I can arrange in the coming weeks!
🎉
I'd like to request a tutorial on Vectorshift and Xano, focusing on how to integrate these platforms effectively for my development projects. Cheers
Hey there, what exactly are you trying to do with vectorshift? You can do most of this natively in Xano, is there a particular item within vectorshift that you were trying to use?
@@LachlanXano Thank you for your response. I am aiming to build a chatbot using Retrieval Augmented Generation (RAG) and was considering using VectorShift for its easy no code pipeline builder. However, if most of this functionality can be achieved natively in Xano, I would love to know how to set it up.
@@Anunaka We will release a tutorial on how to do this very soon!
@@LachlanXano Does this mean, I could replace most of my Flowise stack with XANO or 100% even? Thanks.
@@LachlanXano Preparing the popcorn :D.