moqui-langchain4j

Moqui tool component that integrates LangChain4j for LLM/AI access via ec.getTool("AI", ...).

What this component does

Registers AiToolFactory as a Moqui ToolFactory using the standard component plugin pattern — no framework changes required. Once installed, any Groovy service or script can call ec.getTool("AI", AiToolFactory.class) to send chat messages to an LLM provider and get back plain text or structured JSON output.

Prerequisites

• Moqui Framework 4.0+
• Java 21
• An API key from OpenAI or a compatible provider

Installation

Clone or copy this repository into your Moqui runtime component directory:

$ cd runtime/component
$ git clone https://github.com/patelanil/moqui-langchain4j.git

Moqui picks up the component automatically on the next startup. No additional registration is needed; MoquiConf.xml in this component registers the tool factory and is merged at runtime.

Configuration

Properties are resolved in this order: JVM system property → environment variable → default.

Property	Default	Description
ai_provider	openai	Provider to use. Currently supported: openai.
ai_model	gpt-4o-mini	Model name passed to the provider.
ai_api_key	(none)	Required. API key for the provider.
ai_base_url	(none)	Override the provider base URL (see Multiple Providers below).
ai_timeout	60	Request timeout in seconds.

Via environment variable (recommended for secrets):

export AI_API_KEY=sk-...
export AI_MODEL=gpt-4o

Via MoquiConf.xml override in your runtime conf file:

<default-property name="ai_provider" value="openai"/>
<default-property name="ai_model" value="gpt-4o"/>
<default-property name="ai_api_key" value="${AI_API_KEY}"/>
<default-property name="ai_timeout" value="30"/>

Multiple providers

Setting ai_base_url redirects the OpenAI-compatible HTTP client to any endpoint that speaks the OpenAI API, including:

Provider	Example base URL
Ollama	http://localhost:11434/v1
Groq	https://api.groq.com/openai/v1
Together AI	https://api.together.xyz/v1
Azure OpenAI	https://<your-resource>.openai.azure.com/openai/deployments/<deployment>

Set ai_provider=openai with the appropriate ai_base_url and ai_api_key for the target service.

Usage

Free-form text generation:

import org.moqui.ai.AiToolFactory

def ai = ec.getTool("AI", AiToolFactory.class)

def messages = [
    [role: "system",  content: "You are a helpful assistant."],
    [role: "user",    content: "Summarise this order in one sentence: ${orderText}"]
]
String summary = ai.generate(messages)

Structured output (returns a Map matching the schema):

import org.moqui.ai.AiToolFactory

def ai = ec.getTool("AI", AiToolFactory.class)

def messages = [
    [role: "user", content: "Extract the customer name and total from: ${invoiceText}"]
]
def schema = [
    customerName: [type: "string"],
    total:        [type: "number"]
]
Map result = ai.generateStructured(messages, schema)
// result.customerName, result.total

Schema supports types: string, number, integer, boolean, array, object.

Running the smoke test

Requires ai_api_key to be configured (see Configuration above).

export AI_API_KEY=sk-...
./gradlew :runtime:component:moqui-langchain4j:testAi

The test boots the full Moqui ECF, calls both generate() and generateStructured() against the live provider, and asserts non-empty responses.

Adding new providers

To support a new provider, add a case to the switch (provider) block in src/main/groovy/org/moqui/ai/AiToolFactory.groovy. Build a ChatModel using the appropriate LangChain4j module (e.g. langchain4j-anthropic, langchain4j-google-ai-gemini) and assign it to this.chatModel. Add the new module to build.gradle dependencies and set ai_provider to your new case value.