Introduction¶

In the DeepSeek Harness (DSH) workflow, having an agent directly produce audio content often requires bypassing the plugin system or calling external APIs. As a DSH plugin, yinchao-ai-music integrates the capabilities of the YinChao Open Platform, enabling an agent to directly generate playable songs, instrumental music, or background music from text descriptions or lyrics. It encapsulates audio generation capabilities as a standard DSH plugin, making it convenient to invoke in multi-turn conversations or complex workflows.

Plugin Overview¶

This is an AI music generation Agent Skill maintained by yinhcao. It supports generating complete songs, instrumental music (BGM), or continuing existing music based on text descriptions, lyrics, or reference audio. The plugin runs as a DeepSeek Harness component and calls the YinChao Open Platform API through Harness credentials to complete the text-to-audio generation process.

Core Features¶

The plugin provides the following audio generation capabilities:
1. Text-to-song generation: Generate a complete song based on theme, genre, mood, instruments, and vocal requirements.
2. Instrumental music generation: Generate instrumental music or BGM without vocals based on style, mood, theme, or usage scenario.
3. Lyrics-to-song composition: Compose and perform existing lyrics, preserving the original text by default.
4. Reference audio creation: Create a new song based on MP3, WAV, public audio URLs, or YinChao song IDs.
5. Music continuation: Continue creating and extending music from the end of a song or a specified timestamp.
6. Long-running task handling: Retain the task ID during long-running tasks to avoid duplicate submissions after interruption.
7. On-demand content generation: Generate song titles and lyrics separately only when explicitly requested by the user.

Installation and Enablement¶

To install the plugin, use the officially provided command; do not manually assemble the repository address. Run the following command to add the plugin to the specified DSH profile:

dsh plugin --profile web add github:yinhcao/yinchao-ai-music-skill

After installation, configure the YinChao Open Platform API Key in Harness.

Configuring the API Key¶

Before use, you must create an API Key on the YinChao Open Platform. The DSH Plugin recommends configuring it with Harness credentials. Edit the configuration file (located at ~/.dsh/.credentials.yaml, or $DSH_HOME/.credentials.yaml):

version: 1

refs:
  YINCHAO_API_KEY: "你的 API Key"

After making changes, set the file permissions to ensure only the current user can read it:

chmod 600 ~/.dsh/.credentials.yaml

Typical Usage¶

An agent can invoke the plugin’s features using natural language instructions. Here are some typical input examples:

写一首温暖的中文民谣，木吉他和轻柔男声，主题是多年未见的朋友。
把下面这段歌词谱成一首完整歌曲，保持歌词原文不变：……
做一段轻快的原声吉他 BGM，温暖、松弛，适合咖啡馆使用，不要人声。
参考 /path/to/reference.mp3 的律动和氛围创作一首新歌，但不要复刻旋律。
把 /path/to/song.mp3 从 60 秒开始继续扩写，接下来唱这段歌词：……

Applicable Scenarios and Notes¶

This plugin is suitable for scenarios where an agent needs audio generation capabilities, such as content creation assistants or multimedia interaction agents.

Security and licensing: The code is released under MIT License; platform services, generated content, and brand identity are subject to the terms displayed by the YinChao Open Platform at the time. Do not send secrets to chats, Issues, logs, or commit records.
Permissions: Use only audio content you have the rights to use, and review the privacy policy and terms of service.
Transmission: When using local reference audio or continuation audio, the files will be uploaded to the YinChao Open Platform for this creation task.

Conclusion¶

With yinchao-ai-music, developers can directly bring AI music generation capabilities into the DSH ecosystem. For more details, see the plugin catalog or the source code repository.

yinchao-ai-music: AI Music Generation Agent Skill