Did anyone succeed using the prototyper to setup a semantic search on an indexed vector search?
The AI keeps failing setting it up.
I don’t code and I only use the prototyper, so the help documents is not useful.
We wont be able to just give you code to make it happen as we do not know your project or what language you are coding in. Also since you say you dont know code, then it will be harder to explain from the ground up.
Since you are only using the code assist in your project and dont know code you will be reliant on that. However, here are some videos that may help.
These videos cover the native support for vector embeddings recently added to Firestore, which allows you to perform semantic searches directly on your database documents.
- Introducing Firestore vector search support
- A quick 49-second overview of how Firestore now supports semantically similar item searches to power chatbots and recommendation engines.
- Firestore Vector Embeddings - Release Notes Deep Dive
- A detailed 7-minute technical deep dive into how to use Firestore’s native vector embedding support. This is excellent for understanding the “under the hood” mechanics of the new feature.
- Vector Semantic Search | Hands On Tutorial
- A comprehensive 26-minute “hands-on” guide that walks you through the actual implementation of vector similarity search using Firestore’s capabilities.
Advanced Integration & “Data Connect”
Firebase has introduced Firebase Data Connect, which uses a PostgreSQL backend to handle advanced vector search more robustly.
- Firebase Data Connect Deep Dive
- This 48-minute session explores Firebase Data Connect, which integrates SQL, GraphQL, and AI. It’s the “new way” to handle complex relational data and vector searches in the Firebase ecosystem.
- How to Build a RAG App: Using Firebase Genkit and Vector Search
- Learn how to build Retrieval-Augmented Generation (RAG) applications—which is exactly what you need for an AI companion that remembers past conversations—using Firebase Genkit.
Conceptual Overview
If you want a quick refresher in the future on the logic (turning text into coordinates), these are the gold standard for “non-coder” explanations:
1 Like
Thank you for trying to help.
I don’t use data connect, I just use firestore which is basically hosting JSONs.
I now just want to chat with these without looping over them one by one with AI, which is what I am currently doing (and this costs a lot of time and money).
I installed the vector search on the wanted collection, and it created an index.
But now the AI has no clue how to create a vectorized search query to search in that index. And I cannot help it, despite giving it full up-to-date explanations with the help of GPT.