Feature Map

This page is the AI4J feature map. It does not replace each topic page; it only tells you:

What capabilities exist today.
What problems each capability is suitable for solving.
Which page you should start reading from.
Which capabilities are the stable mainline and which are better suited for advanced exploration.

Maturity markers

Marker	Meaning
`stable`	Recommended as the everyday integration mainline; docs and API semantics are relatively stable
`advanced`	The capability already forms a coherent system, but is better used once you have a clear engineering goal
`preview`	Has an implementation and doc entry point, but the interface, behavior, or best practices may still change
`experimental`	More exploratory or tied to a specific integration; confirm the source code, examples, and limits before use

Getting-started paths

Capability	Status	Module	When to use it	Start here
First Java call	`stable`	`ai4j`	First integration; you want the shortest path to get one model request through	Quickstart for Java
Plain Java quickstart	`stable`	`ai4j`	You want to verify dependencies, configuration, and one model call first	Quickstart for Java
Spring Boot quickstart	`stable`	`ai4j-spring-boot-starter`	You already have a Spring Boot project and want to wire in via configuration and beans	Quickstart for Spring Boot
Chat call semantics	`stable`	`ai4j`	The first call works; you want to understand the message model and call details	Model Access / Chat
First tool call	`stable`	`ai4j`	You want the model to call local functions or tools	First Tool Call
Path selection	`stable`	docs	You're not sure whether to go SDK, Spring, Agent, or FlowGram	Choose Your Path
Documentation map	`stable`	docs	You want to confirm the canonical mainline and where the old paths lead	Documentation Map

Adopt by module

AI4J's module relationships are stacked upward from a foundation, not a platform you must adopt wholesale. You can pick the minimal module for your current goal.

Current goal	Minimal module to adopt	Dependencies	Suitable scenarios
Only do model calls, tools, RAG, MCP	`ai4j`	No internal AI4J dependencies	Get AI capabilities running in a plain Java project first
Wire a Spring Boot app into AI	`ai4j-spring-boot-starter`	Depends on `ai4j`	You need configuration properties, auto-configuration, and bean extensions
Embed an Agent runtime	`ai4j-agent`	Depends on `ai4j`	You need memory, state, workflow, tracing, or team orchestration
Build a local Coding Agent runtime	`ai4j-coding`	Depends on `ai4j`, `ai4j-agent`	You need workspace tools, sessions, the outer loop, and compaction
Provide a CLI / TUI / ACP entry point	`ai4j-cli`	Depends on `ai4j`, `ai4j-coding`	You need a terminal product shell and a local session entry point
Connect a FlowGram backend	`ai4j-flowgram-spring-boot-starter`	Depends on `ai4j-agent`, `ai4j-spring-boot-starter`	You need visual workflows, task API, and a trace bridge
Run the FlowGram demo	`ai4j-flowgram-demo`	Depends on the FlowGram starter	You need a sample backend to validate the integration
Unify versions	`ai4j-bom`	Manages versions across multiple artifacts	Reduce version drift when pulling in multiple AI4J modules

The rule is simple: pull in the minimal module that solves your current problem first; only stack on the next layer when the need naturally rises.

Core SDK

Capability	Status	Module	What it solves	Further reading
Model Access	`stable`	`ai4j`	Unified model integration mainline	Overview
Chat	`stable`	`ai4j`	Conversational model calls	Chat
Responses	`stable`	`ai4j`	Unified call for the Responses style	Responses
Streaming	`stable`	`ai4j`	Streamed output, incremental results, and frontend display	Streaming
Multimodal	`advanced`	`ai4j`	Multimodal input/output such as text and images	Multimodal
Tools / Function Call	`stable`	`ai4j`	Local function declaration, execution, and safety boundaries	Tools
Skills	`advanced`	`ai4j`	Let the model read instructions, templates, and workflow assets on demand	Skills

RAG, retrieval, and MCP

Capability	Status	Module	What it solves	Further reading
Search & RAG	`advanced`	`ai4j`	Retrieve from external knowledge, augment answers, and preserve citation leads	Overview
Ingestion Pipeline	`advanced`	`ai4j`	Document ingestion, splitting, and pre-indexing	Ingestion Pipeline
Hybrid Retrieval	`advanced`	`ai4j`	Combine keyword, vector, and other recall strategies	Hybrid Retrieval
Rerank	`advanced`	`ai4j`	Rerank candidate results to improve retrieval quality	Rerank
MCP	`advanced`	`ai4j`	Integrate external tools, services, and capability gateways via the protocol	MCP Overview
MCP Client Integration	`advanced`	`ai4j`	Connect to and consume MCP capabilities on the client side	Client Integration

App integration and upper runtimes

Capability	Status	Module	Suitable scenarios	Start here
Spring Boot Starter	`stable`	`ai4j-spring-boot-starter`	Configuration-driven Spring integration, auto-configuration, and bean extensions	Spring Boot Overview
Agent Runtime	`preview`	`ai4j-agent`	You need memory, state, tool registry, workflow, or team orchestration	Agent Overview
Agent Quickstart	`preview`	`ai4j-agent`	You want to run a minimal Agent first	Agent Quickstart
Agent Teams	`preview`	`ai4j-agent`	Multi-agent collaboration and division-of-labor orchestration	Agent Teams
Coding Agent	`preview`	`ai4j-coding`, `ai4j-cli`	Repository-scoped task execution, workspace tools, and CLI/TUI	Coding Agent Overview
Coding Agent Quickstart	`preview`	`ai4j-coding`, `ai4j-cli`	You want to try the local Coding Agent product entry point	Coding Agent Quickstart
FlowGram	`preview`	`ai4j-flowgram-spring-boot-starter`	Visual-workflow platform backend, node execution, and trace bridge	FlowGram Overview
FlowGram Quickstart	`preview`	`ai4j-flowgram-demo`	You want to run the FlowGram demo or starter integration	FlowGram Quickstart
Solutions	`advanced`	multiple	Reuse combined solutions by business scenario	Solutions Overview

Production readiness and maintenance

Capability	Status	Module	What it solves	Further reading
Version Compatibility	`stable`	docs	Capability boundaries for Java, Maven, modules, and providers	Version Compatibility
Release and Artifacts	`stable`	docs	Maven artifacts, BOM, and module pull order	Release and Artifacts
Security	`stable`	docs	Security boundaries for secrets, Tools, MCP, RAG, Agent, and FlowGram	Security Overview
Production Checklist	`stable`	docs	Pre-launch checks for configuration, permissions, observability, and regression	Production Checklist
Migration	`stable`	docs	Mental-model migration for old paths, old examples, and old APIs	Migration Guide
Troubleshooting	`stable`	docs	Troubleshooting entry points for providers, Tools, MCP, RAG, Agent, and FlowGram	Troubleshooting
Comparison	`stable`	docs	Selection boundaries vs. Spring AI, LangChain4j, AgentScope Java, and Pi Agent	Comparison

Integrations without a dedicated page yet

Some ecosystem integrations or platform connections may not yet have a stable topic page. If the docs mention Dify, Coze, n8n, AgentFlow, or other external platforms, read by capability category first:

What you want to connect	Which mainline to read first
External tool or service gateway	MCP
Local Java function or business service	Tools
Structured prompts, process instructions, and reusable task assets	Skills
Knowledge base, retrieval augmentation, or document Q&A	Search & RAG
Visual-workflow backend	FlowGram

Integration maturity

These integrations should not be packaged as fully stable capabilities on the entry page. Once the corresponding topic pages are filled in, add deep links from here.

Maturity markers​

Getting-started paths​

Adopt by module​

Core SDK​

RAG, retrieval, and MCP​

App integration and upper runtimes​

Production readiness and maintenance​

Integrations without a dedicated page yet​

Recommended reading order​