MuleSoft Whisperer Connector

MuleSoft AI Chain (MAC) Project

MuleSoft Whisperer Connector

Breaking Changes in v0.4.0

Configuration Architecture Change

Version 0.4.0 introduces a breaking change to the configuration architecture. All applications using previous versions (v0.3.x or earlier) must update their XML configuration before upgrading.

The unified <whisperer:config> element has been removed and replaced with separate configuration elements:

• <whisperer:speech-to-text-config> for Speech-to-Text operations
• <whisperer:text-to-speech-config> for Text-to-Speech operations

This change enables better separation of concerns, future Speech-to-Speech (STS) support, improved type safety, and aligns with standard MuleSoft connector architecture patterns.

Migration Guide

Before (v0.3.x):

<whisperer:config name="Whisperer_Config" apiKey="${openai.api.key}" />

<flow name="transcribe-audio-flow">
    <whisperer:speech-to-text config-ref="Whisperer_Config">
        <whisperer:audio-file>#[payload]</whisperer:audio-file>
    </whisperer:speech-to-text>
</flow>

After (v0.4.0+):

<whisperer:speech-to-text-config name="STT_Config" apiKey="${openai.api.key}" />

<flow name="transcribe-audio-flow">
    <whisperer:speech-to-text config-ref="STT_Config">
        <whisperer:audio-file>#[payload]</whisperer:audio-file>
    </whisperer:speech-to-text>
</flow>

Migration steps:

1. Replace <whisperer:config> with <whisperer:speech-to-text-config> or <whisperer:text-to-speech-config>
2. Update all config-ref attributes to reference new config names
3. Test all flows before deploying to production

See CHANGELOG.md for complete details.

MAC Whisperer supports 2 main use cases,

• Speech-to-Text: Converts audio files (wav, mp3, etc.) into text
• Text-to-Speech: Converts text to audio files (wav, mp3, etc.)

Supported Audio Formats

Core Formats (Always Available)

• MP3 - Pure Java decoder (JLayer)
• WAV - No conversion needed

Extended Formats (Optional ByteDeco FFmpeg)

• M4A/AAC - Requires ByteDeco FFmpeg
• FLAC - Requires ByteDeco FFmpeg
• OGG - Requires ByteDeco FFmpeg
• WEBM - Requires ByteDeco FFmpeg

To enable extended format support, add the ByteDeco FFmpeg dependency to your Mule application:

<!-- Full platform support (~150MB) -->
<dependency>
    <groupId>org.bytedeco</groupId>
    <artifactId>ffmpeg-platform</artifactId>
    <version>6.1.1-1.5.10</version>
</dependency>

Or use a platform-specific variant for smaller size (~30MB):

<!-- Platform-specific (choose your target) -->
<dependency>
    <groupId>org.bytedeco</groupId>
    <artifactId>ffmpeg</artifactId>
    <version>6.1.1-1.5.10</version>
    <classifier>macosx-arm64</classifier> <!-- Mac M1/M2/M3 -->
</dependency>

Available platform classifiers:

• macosx-arm64 - Mac M1/M2/M3
• macosx-x86_64 - Intel Mac
• linux-x86_64 - Linux
• windows-x86_64 - Windows

Mule Application Configuration

Configure the mule-maven-plugin in your Mule application's pom.xml to include ByteDeco as a shared library:

<dependencies>
    <!-- Add ByteDeco FFmpeg dependency -->
    <dependency>
        <groupId>org.bytedeco</groupId>
        <artifactId>ffmpeg</artifactId>
        <version>6.1.1-1.5.10</version>
        <classifier>macosx-arm64</classifier> <!-- Choose your platform -->
    </dependency>
</dependencies>

<build>
    <plugins>
        <plugin>
            <groupId>org.mule.tools.maven</groupId>
            <artifactId>mule-maven-plugin</artifactId>
            <version>4.3.0</version>
            <extensions>true</extensions>
            <configuration>
                <sharedLibraries>
                    <!-- Configure as shared library (groupId and artifactId only) -->
                    <sharedLibrary>
                        <groupId>org.bytedeco</groupId>
                        <artifactId>ffmpeg</artifactId>
                    </sharedLibrary>
                    <sharedLibrary>
                        <groupId>org.bytedeco</groupId>
                        <artifactId>javacpp</artifactId>
                    </sharedLibrary>
                </sharedLibraries>
            </configuration>
        </plugin>
    </plugins>
</build>

Note: The sharedLibraries configuration uses only groupId and artifactId (no version or classifier). The plugin automatically matches against dependencies declared in the <dependencies> section.

For full platform support, use ffmpeg-platform artifact instead of the platform-specific ffmpeg artifact.

Why use sharedLibraries?

• Native libraries must be loaded at the application classloader level
• Prevents classloader conflicts between connectors
• Required for proper ByteDeco FFmpeg initialization
• See MuleSoft documentation for details

Requirements

Mule Runtime

Mulesoft Runtime >= 4.9.0

JDK

• The supported version for Java SDK is JDK 17.
• Compilation with Java SDK must be done with JDK 17.

Installation (using maven central dependency)

<dependency>
    <groupId>cloud.anypoint</groupId>
    <artifactId>mule-whisperer-connector</artifactId>
    <version>0.3.0</version>
    <classifier>mule-plugin</classifier>
</dependency>

Installation (building locally)

To use this connector, first build and install the connector into your local maven repository. Then add the following dependency to your application's pom.xml:

<dependency>
    <groupId>com.mulesoft.connectors</groupId>
    <artifactId>mule4-whisperer-connector</artifactId>
    <version>{version}</version>
    <classifier>mule-plugin</classifier>
</dependency>

Installation into private Anypoint Exchange

You can also make this connector available as an asset in your Anyooint Exchange.

This process will require you to build the connector as above, but additionally you will need to make some changes to the pom.xml. For this reason, we recommend you fork the repository.

Then, follow the MuleSoft documentation to modify and publish the asset.

Contribution

How to contribute

Documentation

• Check out the complete documentation in mac-project.ai

---

Stay tuned!

• 🌐 Website: mac-project.ai
• 📺 YouTube: @MuleSoft-MAC-Project
• 💼 LinkedIn: MAC Project Group