Usage of "Vector Embedding Value" node
What is the "Vector Embedding Value" node doing exactly? It transforms an embedding array (in what format by the way?) into a Firebase friendly "Vector" format, but the node itself returns an object? Furthermore I cannot see a "Vector" field type in the BuildShip database.
9 Replies
AI Support Bot Information
<@1227983829856157809> you can react on the relevant answer (message) with a ✅ in this thread when you think it has been solved, by the bot or by a human!
Anyone can react on this message with a ❌ if the GPT bot is unhelpful or hallucinating answers.
Please note: Team members will review and answer the questions on best efforts basis.
AI Response
The "Vector Embedding Value" node in BuildShip is designed to create a Firestore vector field value for storing embeddings, which can be used for vector similarity search in Firestore. The node takes an embedding vector, which is an array of numbers, and prepares it in a format suitable for Firestore's vector queries.
Here are the main points regarding the "Vector Embedding Value" node:
1. Input Format: The node expects an array of numbers representing the embedding vector. This array can be generated using tools like the "Generate Embeddings" node from the OpenAI integration, which can convert text into embedding vectors.
2. Purpose: The transformed "Vector" is intended for storage in a Firestore vector field, enabling advanced queries such as vector similarity searches.
3. Output: Yes, the node returns an object, but this object is meant to be stored in a Firestore field designated for vectors.
As for the BuildShip database, it is based on Firestore, which now supports vector fields. However, the BuildShip UI might not explicitly show a "Vector" field type, but these functionalities are supported via Firestore under the hood.
For more details, refer to the BuildShip documentation: Vector Embedding Value【source】.
BuildShip AI Support
So how can I store this vector on the database if the database does not support a vector?
Furthermore can I then use such an entry for the "OpenAI - Neighbour Search" node?
We are now tracking the opening neigh search node fix. Also, for the vector search on database will publish a template meanwhile, @Luis can share you a remix.
Hey @Brian, the "Vector Embedding Value" node will return a Firestore vector field for your embeddings. It's looks like a normal object when you preview it in your BuildShip database but under the hood it's typed as Firestore vector field. We have a recent tutorial that covers how to use this node alongside the BuildShip Vector Query node (Nearest neighbor search) for semantic search https://youtu.be/vBjxw5Qq_Kw?si=0d2dlk4exGt9srt8
BuildShip
YouTube
Build Smarter Search: AI-Powered Embedding Search with BuildShip & ...
Want to add powerful, natural language search to your app? In this video, we’ll show you how to build embedding search using BuildShip and Firestore databases—so you can match users to the right data easily.
🔍 We’ll cover:
✅ Setting up semantic search for natural language queries
✅ Matching patients with doctors based on symptoms
...
I am a bit lost here. This example is made for a very specific use case and I am not into this so much yet.
What I currently have:
A Workflow that expects several PDF links:
* Iterate over each PDF
* Extract text from PDF via "PDF To Text" node
* Split paragraphs and create a paragraph string array
* Loop over every paragraph and use "Generate Embeddings" OpenAI node
* Save to array variable
* Create JSON file
I attached this JSON file here.
I am currently saving this File on another CMS.
My goal was to load this JSON file's content into a Chatbot "session table" on the BuildShip database, so on every new user prompt the bot can do an embedding search.
Is this approach incorrect?
What part of the video is important for me to make this work?
@Brian so you're not immediately saving the embeddings to BuildShip database? Maybe we can schedule a short call for me to better understand your use case. You can pick at time here: https://cal.com/rodge-buildship/buildship-chat
@Luis That would be great. Can we also discuss this ticket of mine if possible:
https://discord.com/channels/853498675484819476/1356552625876631652/1356552625876631652
Not sure if you are the right person for this or if it would take too much time
Okay Gaurav would make a meeting over my other ticket. @Luis I will have to check today/tomorrow when I can make time for this one.
Sounds good @Brian