Generative AI and Similarity Search with Vertex AI Matching Engine

Given a talk at Google I/O Extended Bangkok 2023 and Google I/O Extended Vientiane 2023

This talk featured using PaLM API to embed the document and save the embeddings to Vertex AI Matching Engine. Langchain was also introduced to provide the interface in the demo.

The demo is based on https://github.com/GoogleCloudPlatform/generative-ai/tree/main

Transcript

1. Generative AI and Similarity Search with Vertex AI Matching Engine Kamolphan Liwprasert She/Her GDE Cloud / WTM Ambassador
2. Generative AI is a type of artificial intelligence (AI) that can create new content, such as text, images, or music. Generative AI Generative AI Vertex AI
3. LLM: Large Language Models LLM Example: ChatGPT, Bard (PaLM 2)
4. When all these begin? Language Model
5. PaLM 2 (Pathway Language Model) is a Google’s LLM. It comes with different model sizes and parameters for different tasks. cloud.google.com/vertex-ai/docs/generative-ai/learn/models PaLM API PaLM API Available models: • Text-bison • Textembedding-gecko • Chat-bison • Code-bison • Codechat-bison • Code-gecko
6. Getting started with Generative AI! Generative AI developers.generativeai.google/
7. MakerSuite MakerSuite developers.generativeai.google/
8. Vertex AI capabilities Vertex AI
9. Demo time!
10. Demo Scenario PDF files PaLM API embeddings Vector Store 🦜🔗
11. Embedding Google Developers Embedding Layer in a Deep Network An embedding is a relatively low-dimensional space into which you can translate high-dimensional vectors. Embeddings make it easier to do machine learning on large inputs like sparse vectors representing words
12. Similarity Search Flikr Similarity Search “Nearest Neighbor Search” : Compare embedding distance
13. Show me the code! Code
14. Code github.com/GoogleCloudPlatform/generative-ai/blob/main /language/use-cases/document-qa/question_answering_d ocuments_langchain_matching_engine.ipynb
15. Vector Store Vertex AI Matching Engine is a high-scale low latency vector database. These vector databases are commonly referred to as vector similarity-matching or an approximate nearest neighbor (ANN) service. Matching Engine provides the ability to search for semantically similar or semantically related items from its embeddings. Vertex AI Matching Engine Real world use cases such as: • Recommendation engines • Search engines • Ad targeting systems • Image classification or image search • Text classification • Question answering • Chatbots
16. Langchain LangChain is a framework (Python / JS library) for developing applications powered by large language models (LLMs). The main values of LangChain are: • Components: abstractions for working with language models, along with a collection of implementations for each abstraction. Components are modular and easy-to-use, whether you are using the rest of the LangChain framework or not • Off-the-shelf chains: a structured assembly of components for accomplishing specific higher-level tasks langchain.com
17. Code github.com/GoogleCloudPlatform/generative-ai/blob/main /language/examples/document-qa/question_answering_do cuments_langchain_matching_engine.ipynb
18. Architecture Image:github.com/GoogleCloudPlatform/generative-ai/blob/main/language/examples/document-qa/question_answering_document s_langchain_matching_engine.ipynb Store embeddings in Vertex AI Matching Engine (vector search)
19. Architecture Image:github.com/GoogleCloudPlatform/generative-ai/blob/main/language/examples/document-qa/question_answering_document s_langchain_matching_engine.ipynb RAG = Retrieval Augmented Generation
20. Vector Store Create Index * Take up to 1 hours Create Index Endpoint (deploy index) Add documents as embeddings Read PDF files chunks → embeddings Vertex AI Matching Engine
21. developers.generativeai.google Resource
22. github.com/GoogleCloudPlatform /generative-ai Resource
23. Generative AI Learning Path cloudskillsboost.google/journeys/118 Resource
24. Generative AI Learning Path cloudskillsboost.google/journeys/118 Resource
25. cloud.withgoogle.com/next Event Aug. 29-31, 2023
26. Q & A Kamolphan Liwprasert She/Her GDE Cloud / WTM Ambassador
27. Thank You Kamolphan Liwprasert She/Her GDE Cloud / WTM Ambassador