Creating Embeddings With Closed-Source Model in Weaviate | Vector Databases for Beginners | Part 8

Key Takeaways

Now I wanted to go through a different example where we use OpenAI. This is also one where you can use cohhere. Um and this is where we're going to send our data through their API since it's a closed source model in order to generate the embeddings. So um I did the same stuff up here. We have the same random but not the same random a different random data set of 100 objects. Um, and now we need to connect to the cloud again here. So, let me go back over here, click the connect button for the right sandbox and copy our URL and also copy our API key. And I've set my OpenAI API key as an environment variable in Google Collab because that's one I don't want you to see. So, [laughter] um, but you can use it in Collab like this or use, um, any other library in Python. So, now we need to connect to the cloud again. So, I'm going to copy this and paste it here. And there's actually one other thing we need to include here, which is a header. So, if we go over to Weeve's documentation and model provider integrations, you'll see all of our different model providers that you can use for vector embeddings slash um generative modules. So, OpenAI has text embeddings. That's the one we want to use. We'll talk about generative modules in the third webinar. You should also join that one. Um, and if we scroll down, we'll see here this will be our header that we need to include in the connect. So, I'm just going to copy this here. Go back over here. Paste it in there. And I named this slightly differently. So, let's change that. Um, forgot to run this, but here's our different data set of 100 objects. And hopefully when I run this, we're going to get a is ready. Nice. True. Awesome. So, [sighs] as you can see, I updated the headers. It's the same collection. Super updatable. Very nice. And now we're going to create a new collection. This one's going to be called OpenAI vectors collection. Um, again, doing the same thing if it exists. Delete it. Now, our vectorzer config is going to be different again. So, we're going to do the same thing. Pull up my notes over here. You can find this in the documentation. Um yeah. Okay. Vectorzer config is going to be WC dot configure dot vectorzer dot text to vec open ai parenthesis. And that's it. So hopefully we run this and again we've defined properties. So we have a text and a title and basically behind the scenes what we will do is any data object that you import it will automatically vectorize it using this config. So we run this. We're going to create the collection with those settings and then we can insert our objects again here. This time we're not including the vector property because we is going to do that automatically. So let's do that. It's taking a little bit longer because again it has to vectorize it. And then if we go to our cloud again and then if we go to the explorer tool again and choose our sandbox. Oh, we have a new collection. If we click on that we are seeing data. Awesome. Okay. And we can do the same little searchy thing. This time instead of near vector, we're going to use near text because we have specified a vector. So we can just actually add the text directly. We don't have to vectorize it beforehand. And we get results. and different results cuz it's a different data set. But that's why