Building Intelligent Search Apps with Firestore Vector Search and App Hosting

Building Intelligent Search Apps with Firestore Vector Search and App Hosting

Transcript

1. Building Intelligent Search Apps with Firestore Vector Search and App Hosting Hassan Bahati Software Engineer
2. Storage Authentication Realtime Database Hosting Machine Learning Cloud Firestore Build Build with Firebase
3. Crashlytics Performance Monitoring Test Lab App Distribution Release &Monitor Release & Monitor with Firebase
4. Analytics Remote Config A/B Testing In App Messaging Google Admob Cloud Messaging Engage Engage with Firebase
5. Vector Search? Vector search is a technique that uses vector representations of data to perform more accurate and context-aware searches compared to traditional keyword-based methods. Vector search represents your data as arrays of numbers and performs searches based on similarity. Find items that are semantically similar to a given query.
6. Vector Search avails new experiences of how users can interact with your data unlike traditional keyword based searching.
7. Use Cases E-commerce - Product search and recommendations based on user behavior and preferences Media : - Searching and categorizing multimedia content. Customer Support - Intelligent FAQs and support ticket routing based on query analysis.
8. How it works? • Data is turned in vector embeddings. • By embedding data points and calculating the similarity between user queries in natural language and these vectors, it becomes possible to find or recommend items that match the user’s intent, based on the data points. • Your app can now understand and make symatic comparisons. • When you make a vector search, items semantically similar to your query are returned.
9. Vector Search with Firestore Extension automatically computes vector embeddings making it easy to perform vector search for mobile or web apps.
10. Indexing data with vectors • Prepare the data. • Decide what data points you want to index. Eg product descriptions.
11. Querying Data with Firestore Vector Search
12. Query Techniques // Use cosine similarity or other distance measures to find the most relevant vectors. Example - Semantic search: Finding documents similar in meaning to a query. - Recommendations: Suggesting similar items based on user preferences.
13. Pre-requisites // Set up a Firestore project in the Firebase Console. // Enable necessary APIs and install Firebase SDKs in your project. // Set up Firestore Vector Search Extension // Choose an environment eg Nodejs, Python
14. Building a grocery app 1. Create descriptions for vectorization 2. Embed the descriptions 3. Register data to Firestore 4. Create indexes for vector search 5. Perform vector search using natural language
15. Firebase App Hosting Firebase App Hosting is a service that allows you to deploy and host dynamic web applications (Next.js, Angular, etc.) with GitHub integration and seamless connections to other Firebase products like Authentication, Cloud Firestore, and AI tools.
16. Hand-on session 🚀🚀 Deep dive into a hand-on second half where we shall build a grocery app with an intelligent search functionality.
17. Resources 1. Search with vector embeddings | Firestore | Firebase 2. Vector Search with Firestore | Firebase Extensions Hub