AI Docs

A Gradle plugin that collects DOCUMENTATION.md artifacts from your project's dependencies and organizes them into an AI-navigable file structure — so AI coding assistants can deeply understand every library you use without blowing up their context window.

The Problem

When an AI agent works on a Java project with many dependencies, it either loads all documentation (too much context), loads nothing (works blind), or asks you every time. None of these scale.

The Solution

AI Docs introduces a simple convention:

1. Library authors publish a DOCUMENTATION.md alongside their jar (as a Maven artifact with classifier DOCUMENTATION)
2. Projects apply this plugin and run ./gradlew collectDocs
3. AI agents navigate the output efficiently: index → overview → specific lines

Three small reads instead of dumping everything into context.

Quick Start

1. Apply the Plugin

// settings.gradle
pluginManagement {
    repositories {
        mavenLocal()
        gradlePluginPortal()
    }
}

// build.gradle
plugins {
    id 'java'
    id 'one.jpro.aidocs' version '0.1.0-SNAPSHOT'
}

2. Collect Documentation

./gradlew collectDocs

This collects all dependency docs into build/ai-docs/ and installs a skill file at .claude/skills/docs/SKILL.md so AI agents automatically know the documentation is available.

3. Inspect the Output

build/ai-docs/
├── index.md                              # All libraries at a glance
├── one.jpro.platform/
│   └── jpro-routing-core/
│       ├── overview.md                   # Chapter titles + line ranges
│       └── DOCUMENTATION.md              # Full documentation

index.md — lightweight entry point:

# AI Documentation Index

## Available Libraries
- one.jpro.platform:jpro-routing-core:0.5.8 — [overview](one.jpro.platform/jpro-routing-core/overview.md)

overview.md — chapter structure with line ranges:

# jpro-routing-core (0.5.8)
Full documentation: DOCUMENTATION.md

## Chapters
- Overview (lines 1-25)
- Getting Started (lines 26-80)
- Filters API (lines 81-150)

An AI agent reads index.md to discover what's available, then reads overview.md for a specific library to see its structure, then reads only the relevant lines from DOCUMENTATION.md.

How AI Agents Navigate This

1. Read index.md — see all available libraries (always small and cheap)
2. Read overview.md for the relevant library — see chapter structure with line ranges
3. Read specific lines from DOCUMENTATION.md — load only the chapter needed

For Library Authors

To make your library's documentation available to AI agents, publish a DOCUMENTATION.md as a Maven artifact:

• Classifier: DOCUMENTATION
• Extension: md

The artifact coordinate looks like: com.example:my-lib:1.0.0:DOCUMENTATION@md

Libraries that don't publish this artifact are silently skipped — no errors, no warnings.

Try It

The example/ directory is a project with JPro dependencies. Set it up and let an AI agent build a full app:

./gradlew publishToMavenLocal   # from root
cd example
./gradlew collectDocs           # collect dependency docs

Then open the example/ directory in Claude Code and prompt:

Build an Expense Tracker web application using JavaFX and JPro.
Use jpro-routing for multi-page navigation and jpro-auth-routing for Google login.
Read the documentation in build/ai-docs/ to understand the frameworks.

The agent discovers the available libraries, learns their APIs from the collected docs, and generates the application — without any prior knowledge of these frameworks.

Building from Source

Requirements: Java 17+, Gradle 9.2+

# Build and run all tests
./gradlew build

# Publish to local Maven repository
./gradlew publishToMavenLocal

# Try the example project
cd example
./gradlew collectDocs
cat build/ai-docs/index.md

Project Structure

Module	Description
ai-docs-core	Reusable library for document collection and index/overview generation
ai-docs-gradle-plugin	Gradle plugin providing the collectDocs task
ai-docs-maven-plugin	Maven plugin (planned)
example/	Standalone project demonstrating real-world usage with JPro libraries

License

TBD