Microsoft’s TypeChat

Microsoft has been pacing fast enough in the AI sector by launching a variety of products for users.

Microsoft recently launched an open-source library called TypeChat on its GitHub blog. TypeChat uses TypeScript and generative AI to bridge natural language, application schema, and APIs.

Microsoft mentions: “how do we augment traditional UI with natural language interfaces?How do we use AI to take a user request and turn it into something our apps can operate on?And how do we make sure our apps are safe, and doing work that developers and users alike can trust?

“Today we're releasing TypeChat, an experimental library that aims to answer these questions. It uses the type definitions in your codebase to retrieve structured AI responses that are type-safe,” it adds.

This new platformwas created to make it easier for developers to build natural language interfaces to large language models (LLMs).

One of the biggest problems that developers have with integrating large language models (LLMs) into software is that LLMs rarely give text that machines can read.

Even when the user asks the model to provide structured data, LLMs fall short. However, Microsoft has found that these generative AI algorithms can translate user queries into JSON format, which works well for machines.

In simpler terms, LLMs are really good at generating text that humans can understand. But they're not as good at generating text that machines can understand.

This is because LLMs are trained on large datasets of text that is written by humans. This text is full of human-specific quirks and conventions that machines don't understand.

JSON is a different kind of text. It's a format that's designed to be machine-readable. JSON text is made up of objects, which are like lists of key-value pairs.

Each key-value pair is a pair of words, where the first word is the key and the second word is the value.

Microsoft states: “The current wave of LLMs default to conversational natural language —languages that humans communicate in like English. Parsing natural language is an extremely difficult task, no matter howmuch you pamper a prompt with rules like "respond in the form a bulleted list". Natural language might have structure, but it's hard for typical software to reconstruct it from raw text.

“Surprisingly, we can ask LLMs to respond in the form of JSON, and they generally respond with something sensible,” it adds.

That is why TypeChat is a helpful tool that can help developers to be integrated in apps. TypeChat works by combining a human prompt with a "response schema."

The prompt is a natural language question or instruction, and the response schema is a description of the expected response from the LLM.

TypeChat uses TypeScript, a programming language that is well-suited for working with structured data, to make it easy to define the prompts and response schemas.

How to use TypeChat?

Developers can start with TypeChat by installing it using the following command: “npm install typechat”

Once TypeChat is installed, you can start using it in your application. To do this, you need to create a TypeScript file that defines the prompts and response schemas for your NLI. 

In conclusion, TypeChat is a powerful tool that can be used to simplify the process of building NLIs to LLMs. TypeChat is still under development but it can be quite beneficial.

For more details visit following sites:

TypeChat: https://microsoft.github.io/TypeChat/