Documentation

IntentSharp applies a natural-language intent to a .NET value and returns a typed, verified result from OpenAI, Anthropic or any Microsoft.Extensions.AI IChatClient. Intents, result shapes and rules can live in versioned JSON or YAML files that reload while the application runs.

These pages are written from the code. Each one describes what the library does today, with the names of the types and methods involved.

Start here

Quick start: one console program that builds a client, makes a typed call, filters a collection, adds a rule of your own and runs an intent from a file.
Architecture overview: the packages, the component map and the pipeline in one picture.
Request pipeline: every step of a call, from the prompt to the result.
Verification: what makes a result succeed or fail.

The three packages

Package	What it adds	Read
`IntentSharp`	The client, the request pipeline, the OpenAI and Anthropic providers, collection transforms, the database runtime, memory, cache and rate limiting.	Overview, Public API
`IntentSharp.Catalog`	Intents as data: definitions in JSON or YAML files, rules, examples, hot reload and audit tags.	Intent catalog
`IntentSharp.Extensions.AI`	Runs the client on any `IChatClient`, and exposes catalog intents as `AIFunction` tools for Agent Framework agents.	Microsoft.Extensions.AI and Agent Framework

By topic

Prompting and schemas: the prompt layout, the JSON envelope and strict schemas.
Providers: OpenAI, Anthropic, fallback, and adding your own.
Collection transforms: map, filter, sort, group, reduce and map-values.
Database runtime: natural-language queries behind a read-only, allowlisted guard.
Memory, cache and rate limiting.
Licensing enforcement: what the library does with a license key, and without one.

Background

Glossary: the terms used in the code and these pages.
Licensing background: the license for users, and how to buy.
Changelog: what changed in each release.

Buying and licensing