mirror of https://git.openapi.site/https://github.com/desirecore/market.git synced 2026-06-06 05:50:41 +08:00

Files

xyx 4f7037a6b6 fix: replace hardcoded ~/.desirecore paths with ${DESIRECORE_ROOT} variable (#16 )

## Summary

- 将所有技能文件中的硬编码 `~/.desirecore/` 和 `$HOME/.desirecore/` 路径替换为
`${DESIRECORE_ROOT}/` 变量
- 递增 manifest.json version 至 1.2.1

## Why

dev 模式下 `DESIRECORE_HOME=~/.desirecore-dev`，硬编码路径导致技能读取错误的端口文件和目录。主仓库的
`variable-substitutor.ts` 会在运行时将 `${DESIRECORE_ROOT}` 替换为实际根目录。

## Test plan

- [ ] `npm run dev` 启动后触发任意技能，确认端口路径解析为
`~/.desirecore-dev/agent-service.port`
- [ ] prod 模式确认路径为 `~/.desirecore/agent-service.port`

🤖 Generated with [Claude Code](https://claude.com/claude-code)

2026-05-29 15:36:19 +08:00

11 KiB

Raw Blame History

name, description, license, version, type, risk_level, status, disable-model-invocation, provider, tags, requires, metadata, market

name

description

license

version

type

risk_level

status

disable-model-invocation

provider

minimax-music-gen Skill

Mandatory Rules (violations will cause functionality to fail)

Must access agent-service over HTTPS — https://127.0.0.1:${PORT} with -k to skip certificate verification
Use Bash curl throughout — do not use the HttpRequest tool or Python
Do not use output_format: "url" — URL downloads will return empty files in scenarios such as Token Plan due to CDN authentication failures. Always use the default hex format; audio data is returned directly in the API response

Full Execution Flow

Prerequisites

The user has configured the MiniMax Provider (regular API or Token Plan) in Resource Manager → Compute and filled in the API Key
agent-service is running

Core Concepts

MiniMax Music Generation is a synchronous API (not an asynchronous task model); it returns audio data directly when called. Three modes are supported:

Mode	model	Description
Song generation	`music-2.6`	Provide prompt + lyrics to generate a song with vocals
Pure instrumental	`music-2.6`	Set `is_instrumental: true`; only a prompt is needed
Cover	`music-cover`	Provide a reference audio + prompt; rearrange based on the melodic skeleton

Lyrics Structure Tags

The lyrics field supports the following structure tags to organize song sections:

Tag	Meaning
`[verse]`	Verse
`[chorus]`	Chorus
`[bridge]`	Bridge
`[intro]`	Intro
`[outro]`	Outro
`[interlude]`	Interlude

Example lyrics format:

[verse]
夜晚的城市灯火阑珊
我独自走在回家的路上

[chorus]
这一刻时间仿佛停止
所有的喧嚣都已远去

Generate a Song (with Vocals)

Note: Do not pass the output_format parameter; use the default hex format.

PORT=$(cat ${DESIRECORE_ROOT}/agent-service.port)
curl -sk -X POST "https://127.0.0.1:${PORT}/api/media-proxy" \
  -H "Content-Type: application/json" \
  -d '{
    "provider": "minimax",
    "serviceType": "music_gen",
    "endpoint": "/music_generation",
    "body": {
      "model": "music-2.6",
      "prompt": "独立民谣,温暖,治愈,吉他伴奏",
      "lyrics": "[verse]\n歌词内容\n\n[chorus]\n副歌内容",
      "audio_setting": {
        "format": "mp3",
        "sample_rate": 44100,
        "bitrate": 256000
      }
    },
    "responseType": "json"
  }'

Generate Pure Instrumental

PORT=$(cat ${DESIRECORE_ROOT}/agent-service.port)
curl -sk -X POST "https://127.0.0.1:${PORT}/api/media-proxy" \
  -H "Content-Type: application/json" \
  -d '{
    "provider": "minimax",
    "serviceType": "music_gen",
    "endpoint": "/music_generation",
    "body": {
      "model": "music-2.6",
      "prompt": "电子音乐,氛围感,空灵,合成器铺底",
      "is_instrumental": true,
      "audio_setting": {
        "format": "mp3",
        "sample_rate": 44100,
        "bitrate": 256000
      }
    },
    "responseType": "json"
  }'

Response Handling and Saving

The API returns JSON; audio data is hex-encoded and stored in the data.data.audio.data field.

Response structure:

{
  "success": true,
  "data": {
    "data": {
      "audio": {
        "data": "hex编码的音频数据...",
        "status": 2
      }
    },
    "extra_info": {
      "music_duration": 180000,
      "music_sample_rate": 44100,
      "music_channel": 2,
      "bitrate": 256000,
      "music_size": 1234567
    },
    "base_resp": { "status_code": 0, "status_msg": "success" }
  },
  "statusCode": 200
}

Note: The status field means 1 = synthesizing (streaming scenario), 2 = synthesis complete. In non-streaming mode, the returned status is 2.

Save the hex Audio Data to media-store

Extract the hex string from the data.data.audio.data field of the response JSON, convert it to binary, and upload:

PORT=$(cat ${DESIRECORE_ROOT}/agent-service.port)
# Save the API response to a temporary file (avoid letting large hex data overflow shell variables)
# Assume the curl output of the previous step has been saved to /tmp/minimax-music-resp.json

# Extract hex data and convert to binary (pure Bash, no Python dependency)
jq -r '.data.data.audio.data' /tmp/minimax-music-resp.json | xxd -r -p > /tmp/minimax-music.mp3

# Verify the file is valid (greater than 1KB and in audio format)
FILE_SIZE=$(stat -f%z /tmp/minimax-music.mp3 2>/dev/null || stat -c%s /tmp/minimax-music.mp3 2>/dev/null)
if [ "$FILE_SIZE" -lt 1024 ]; then
  echo "ERROR: 音频文件异常（${FILE_SIZE} 字节），可能生成失败"
  exit 1
fi

# Upload to media-store
curl -sk -X POST "https://127.0.0.1:${PORT}/api/media/upload" \
  -F "file=@/tmp/minimax-music.mp3;type=audio/mpeg"

Extract the mediaId field from the upload response JSON.

Display the Result

In the reply, use a dc-media protocol reference (the frontend will automatically detect the audio extension and render a player):

![音乐生成结果](dc-media://这里替换为mediaId)

Parameter Descriptions

Parameter	Description	Required	Default
model	Model name	Yes	"music-2.6"
prompt	Music style/mood description	Optional when lyrics are present; required for pure instrumental/cover	—
lyrics	Lyrics (structure tags supported)	Required when not in pure instrumental mode	—
is_instrumental	Whether to generate pure instrumental	No	false
lyrics_optimizer	Auto-generate lyrics from the prompt	No	false
audio_setting.format	Audio format: mp3/wav/pcm	No	"mp3"
audio_setting.sample_rate	Sample rate: 16000/24000/32000/44100	No	32000
audio_setting.bitrate	Bitrate: 32000/64000/128000/256000	No	128000

Tips for Writing Prompts

The prompt is used to describe the music's style, mood, and instrumentation; commas are recommended to separate keywords:

Style: 独立民谣, 电子舞曲, 古典钢琴, 摇滚, R&B, 爵士, 嘻哈
Mood: 温暖, 忧郁, 欢快, 史诗感, 空灵, 治愈
Instruments: 吉他伴奏, 钢琴独奏, 弦乐铺底, 合成器, 鼓点强劲
Structure: 渐进式编曲, 开场留白渐入高潮, 轻柔开头爆发副歌

Example: "独立民谣,温暖治愈,木吉他为主,轻柔的鼓点,渐进式编曲"

Auto-generated Lyrics Mode

If the user only describes the desired music style without providing lyrics, set lyrics_optimizer: true and the model will auto-generate lyrics from the prompt:

PORT=$(cat ${DESIRECORE_ROOT}/agent-service.port)
curl -sk -X POST "https://127.0.0.1:${PORT}/api/media-proxy" \
  -H "Content-Type: application/json" \
  -d '{
    "provider": "minimax",
    "serviceType": "music_gen",
    "endpoint": "/music_generation",
    "body": {
      "model": "music-2.6",
      "prompt": "一首关于夏日海边回忆的歌,独立民谣,温暖,吉他",
      "lyrics_optimizer": true,
      "audio_setting": {
        "format": "mp3",
        "sample_rate": 44100,
        "bitrate": 256000
      }
    },
    "responseType": "json"
  }'

Error Handling

base_resp.status_code: 1002: rate limit reached, retry later
base_resp.status_code: 1004: API Key authentication failed
base_resp.status_code: 1008: insufficient balance
base_resp.status_code: 1026: content sensitive, modify the lyrics or prompt and retry
base_resp.status_code: 2013: parameter error, check required fields
success: false + error: "未找到匹配的供应商": No enabled MiniMax provider with music_gen service found

Notes

The prompt length limit is 1-2000 characters; the lyrics length limit is 1-3500 characters
Token Plan users: all plans use music-2.6 for free (100 tracks/day, each track ≤5 minutes)
Unless the user specifies otherwise, default to music-2.6 + mp3 format + 44100 sample rate
If the user only gives a theme without lyrics, use lyrics_optimizer: true to auto-generate lyrics
If the user requests pure music/accompaniment, set is_instrumental: true
Music generation takes a relatively long time (typically 30-90 seconds); please be patient
The hex data volume is large (several MB); always use a temporary file as intermediary, do not store it in shell variables

11 KiB Raw Blame History