Generative AI Use Cases (GenU)
A collection of business use cases for using generative AI safely
Main Features
Multiple Use Cases
Chat, RAG, text generation, and more, covering various business scenarios
Secure Implementation
Secure AI use base using Amazon Bedrock
Customizable
Add custom features with the use case builder
Use Cases
Chat
Large language models (LLM) can interact with each other in a chat format. Thanks to the platform that allows LLM to directly interact with each other, we can quickly respond to new use cases and detailed use cases. It is also effective as a verification environment for prompt engineering.
Use Case Builder
This feature allows you to create your own use cases by describing prompt templates in natural language. Since the use case screen is automatically generated from the prompt template, there is no need to change code or customize it.
RAG Chat
RAG is a technique that allows LLM to generate answers based on evidence by providing the latest information or domain knowledge that LLM is not good at. For example, if you pass in internal documents to LLM, you can automate internal inquiries. This repository uses Amazon Kendra or Knowledge Bases to get information.
Agent Chat
Agent is a technique that allows LLM to perform various tasks by connecting to API. This solution implements an Agent that searches for and answers necessary information using a search engine as a sample implementation.
Flow Chat
Amazon Bedrock Flows allows you to create workflows by connecting prompts, foundation models, and other AWS services. The Flow chat use case allows you to use a chat that selects and executes a pre-created flow.
Summarize
LLM is good at summarizing large amounts of text. Not only can it summarize text, but it can also retrieve necessary information in an interactive form by providing the text as context. For example, you can retrieve information such as "What are the conditions for XXX?" and "What is the amount for YYY?" by reading a contract.
Writing
LLM can assist with writing and proofreading articles. In addition to checking for typos and errors, it can suggest improvements in the flow and content of the article, and provide an objective view from a more neutral perspective. It is expected to improve the quality by checking points that the user may not have noticed themselves before showing it to others.
Translation
LLM that has learned multiple languages can also translate. In addition to just translating, it is possible to reflect various specified context information such as casualness and target layer into translation. For example, if you translate a contract, you can translate it into multiple languages.
Web Content Extraction
LLM can extract web content such as blogs and documents. It removes unnecessary information and formats it as a complete article. The extracted content can be used in other use cases such as summarizing and translating.
Image Generation
LLM can generate images based on text and images. It can help you visualize your ideas and improve your design work.
Video Generation
LLM can generate videos based on text. It can help you visualize your ideas and improve your design work.
Video Analysis
LLM can analyze videos by inputting not only text but also images. This feature allows you to analyze videos by inputting the image frames and text as input.
Diagram Generation
LLM can generate diagrams such as flowcharts, sequence diagrams, and mind maps from natural language descriptions, documents, and code. It can help you visualize complex relationships and understand them more efficiently.
