skillfree 0.1.25 → 0.1.37

package/scripts/commands/models.js

@@ -10,27 +10,35 @@ async function listModels(type) {
     const t = m.type || 'chat'
     if (type && t !== type) continue
     if (!groups[t]) groups[t] = []
-    groups[t].push(m.id)
+    groups[t].push(m)
   }
 
   const typeLabel = {
-    chat:      '💬 对话',
-    image:     '🎨 图像',
-    tts:       '🔊 语音合成',
-    video:     '🎬 视频',
-    music:     '🎵 音乐',
-    ocr:       '📄 OCR',
-    embedding: '🔢 Embedding',
+    chat:   '💬 对话',
+    image:  '🎨 图像',
+    video:  '🎬 视频',
+    ocr:    '📄 OCR 解析',
+    search: '🌐 搜索',
   }
 
-  console.log('\n🦞 SkillFree 可用模型\n')
-  for (const [t, ids] of Object.entries(groups)) {
+  console.log('\n🦞 SkillFree 严选模型\n')
+  for (const [t, models] of Object.entries(groups)) {
     console.log(`${typeLabel[t] || t}`)
-    ids.forEach(id => console.log(`  ${id}`))
+    for (const m of models) {
+      let price = ''
+      if (m.tokenBased) {
+        price = `输入 ¥${((m.inputPrice || 0) / 100).toFixed(0)}/1M · 输出 ¥${((m.outputPrice || 0) / 100).toFixed(0)}/1M`
+      } else if (m.pricePerSec) {
+        price = `720p ¥${((m.pricePerSec['1280x720'] || 0) / 100).toFixed(1)}/s · 1080p ¥${((m.pricePerSec['1920x1080'] || 0) / 100).toFixed(1)}/s · 4K ¥${((m.pricePerSec['3840x2160'] || 0) / 100).toFixed(1)}/s`
+      } else if (m.estimatedCredits) {
+        price = `¥${(m.estimatedCredits / 100).toFixed(2)}/次`
+      }
+      const label = m.label ? `  ${m.label}` : ''
+      console.log(`  ${m.id}${label ? ' (' + m.label + ')' : ''}  ${price}`)
+    }
     console.log()
   }
-  console.log(`共 ${all.length} 个模型`)
-  console.log(`\n用法示例：skillfree pilot --type chat --model DeepSeek-V3.2-Fast --prompt "你好"`)
+  console.log(`\n用法：skillfree pilot --type chat --model claude-sonnet-4-6 --prompt "你好"`)
 }
 
 module.exports = { listModels }

package/scripts/commands/pilot.js

@@ -1,4 +1,4 @@
-const { post, postStream, get, request, getApiKey, BASE_URL, checkCredits } = require('../lib/client')
+const { post, postStream, request, getApiKey, BASE_URL, checkCredits } = require('../lib/client')
 const fs = require('fs')
 const path = require('path')
 
@@ -10,35 +10,23 @@ async function downloadAndSave(url, output) {
   console.log(`✅ 已保存到 ${output}`)
 }
 
-// ─── 工具：封装 PCM 为 WAV ─────────────────────────────────────────────────────
-function pcmToWav(pcmBytes, sampleRate = 24000, channels = 1, bitsPerSample = 16) {
-  const dataSize = pcmBytes.length
-  const wav = Buffer.alloc(44 + dataSize)
-  wav.write('RIFF', 0); wav.writeUInt32LE(36 + dataSize, 4); wav.write('WAVE', 8)
-  wav.write('fmt ', 12); wav.writeUInt32LE(16, 16); wav.writeUInt16LE(1, 20)
-  wav.writeUInt16LE(channels, 22); wav.writeUInt32LE(sampleRate, 24)
-  wav.writeUInt32LE(sampleRate * channels * bitsPerSample / 8, 28)
-  wav.writeUInt16LE(channels * bitsPerSample / 8, 32); wav.writeUInt16LE(bitsPerSample, 34)
-  wav.write('data', 36); wav.writeUInt32LE(dataSize, 40)
-  pcmBytes.copy(wav, 44)
-  return wav
-}
-
 async function pilot(flags) {
-  const type = flags.type || 'chat'
+  const type  = flags.type   || 'chat'
   const prompt = flags.prompt || flags.text || ''
   const output = flags.output || null
-  const model = flags.model || null
+  const model  = flags.model  || null
 
-  // ── 积分预检（< 100 积分时拦截并提示充值）────────────────────────────────────
+  // ── 余额预检（< 100 积分时拦截并提示充值）────────────────────────────────────
   await checkCredits()
 
   // ── CHAT ─────────────────────────────────────────────────────────────────────
   if (type === 'chat') {
+    const chatModel = model || 'claude-sonnet-4-6'
+
     if (!output) {
       // 流式输出
       const res = await postStream('/chat/completions', {
-        model: model || 'DeepSeek-V3.2-Fast',
+        model: chatModel,
         messages: [{ role: 'user', content: prompt }],
         stream: true,
       })
@@ -66,13 +54,14 @@ async function pilot(flags) {
       }
       return
     }
+
     const result = await post('/chat/completions', {
-      model: model || 'DeepSeek-V3.2-Fast',
+      model: chatModel,
       messages: [{ role: 'user', content: prompt }],
     })
     const text = result.choices?.[0]?.message?.content || JSON.stringify(result, null, 2)
-    if (output) { fs.writeFileSync(output, text); console.log(`✅ 已保存到 ${output}`) }
-    else console.log(text)
+    fs.writeFileSync(output, text)
+    console.log(`✅ 已保存到 ${output}`)
     return
   }
 
@@ -80,26 +69,6 @@ async function pilot(flags) {
   if (type === 'image') {
     const imageModel = model || 'gemini-3.1-flash-image-preview'
 
-    // qwen-image-edit-plus：图片编辑，走 /v1/images/edits，multipart/form-data
-    if (imageModel.startsWith('qwen-image')) {
-      if (!flags.file) throw new Error('qwen-image-edit 需要 --file 指定原始图片路径')
-      const imgBuf = fs.readFileSync(path.resolve(flags.file))
-      const form = new FormData()
-      form.append('model', imageModel)
-      form.append('prompt', prompt)
-      form.append('size', flags.size || '1024x1024')
-      form.append('image', new Blob([imgBuf], { type: 'image/png' }), path.basename(flags.file))
-      const res = await request('/images/edits', { method: 'POST', body: form })
-      const data = await res.json()
-      if (data.error) throw new Error(data.error.message || JSON.stringify(data.error))
-      const url = data.data?.[0]?.url
-      if (!url) throw new Error('未返回图像URL: ' + JSON.stringify(data).slice(0, 200))
-      if (output) await downloadAndSave(url, output)
-      else console.log('图像 URL:', url)
-      return
-    }
-
-    // doubao-seedream-5.0-lite / 其他标准图像模型
     const res = await post('/images/generations', {
       model: imageModel,
       prompt,
@@ -107,8 +76,10 @@ async function pilot(flags) {
       size: flags.size || '1024x1024',
     })
     if (res.error) throw new Error(res.error.message || JSON.stringify(res.error))
+
     const url = res.data?.[0]?.url || res.data?.[0]?.b64_json
     if (!url) throw new Error('未返回图像数据: ' + JSON.stringify(res).slice(0, 200))
+
     if (output) {
       if (url.startsWith('http')) await downloadAndSave(url, output)
       else { fs.writeFileSync(output, Buffer.from(url, 'base64')); console.log(`✅ 已保存到 ${output}`) }
@@ -118,156 +89,48 @@ async function pilot(flags) {
     return
   }
 
-  // ── TTS ───────────────────────────────────────────────────────────────────────
-  if (type === 'tts') {
-    const ttsModel = model || 'speech-2.6-hd'
-    const text = prompt
+  // ── VIDEO（Veo 异步轮询）─────────────────────────────────────────────────────
+  if (type === 'video') {
+    if (!prompt) throw new Error('--prompt 是必需的（视频描述）')
+    const videoModel = model || 'veo-3.1-fast-generate-preview'
+    const seconds    = flags.seconds || '8'
+    const size       = flags.size    || '1920x1080'
 
-    if (ttsModel === 'speech-2.8-hd' || ttsModel === 'minimax-clone-lastversion') {
-      // 走 /v1/responses，返回 hex 音频
-      const res = await request('/responses', {
-        method: 'POST',
-        body: JSON.stringify({
-          model: ttsModel,
-          input: text,
-          stream: false,
-          voice_setting: {
-            voice_id: flags.voice || 'female-shaonv',
-            speed: 1, vol: 1, pitch: 0, emotion: 'fluent',
-          },
-          audio_setting: { sample_rate: 32000, bitrate: 128000, format: 'mp3', channel: 1 },
-          output_format: 'hex',
-        }),
-      })
-      const data = await res.json()
-      if (!data.data?.audio) throw new Error(JSON.stringify(data).slice(0, 200))
-      const audioBuf = Buffer.from(data.data.audio, 'hex')
-      if (output) { fs.writeFileSync(output, audioBuf); console.log(`✅ 已保存到 ${output}（${audioBuf.length} bytes）`) }
-      else console.log(`✅ TTS 成功，时长约 ${(data.extra_info?.audio_length / 1000).toFixed(1)} 秒`)
+    console.log(`🎬 提交 Veo 视频任务（${videoModel}，${size}，${seconds}s）...`)
+    console.log('⏳ 视频生成中，预计需要 1-3 分钟...\n')
 
-    } else if (ttsModel === 'gemini-2.5-pro-preview-tts' || ttsModel === 'gemini-2.5-flash-preview-tts') {
-      // Gemini TTS：走 skillfree.tech 后端 /v1beta 路由
-      // 注意：用 fetch+BASE_URL 而非 request()，因为 request() 会自动加 /v1 前缀
-      const apiKey = getApiKey()
-      const res = await fetch(BASE_URL + '/v1beta/models/' + ttsModel + ':generateContent', {
-        method: 'POST',
-        headers: { 'Authorization': `Bearer ${apiKey}`, 'Content-Type': 'application/json' },
-        body: JSON.stringify({
-          contents: [{ parts: [{ text }] }],
-          generationConfig: {
-            responseModalities: ['AUDIO'],
-            speechConfig: {
-              voiceConfig: { prebuiltVoiceConfig: { voiceName: flags.voice || 'Kore' } }
-            }
-          }
-        }),
-      })
-      const data = await res.json()
-      if (data.error) throw new Error(data.error.message || JSON.stringify(data.error))
-      const pcm = Buffer.from(data.candidates[0].content.parts[0].inlineData.data, 'base64')
-      const ext = (output || '').endsWith('.mp3') ? 'mp3' : 'wav'
-      const finalBuf = ext === 'wav' ? pcmToWav(pcm) : pcm
-      if (output) { fs.writeFileSync(output, finalBuf); console.log(`✅ 已保存到 ${output}（${finalBuf.length} bytes）`) }
-      else console.log(`✅ Gemini TTS 成功，时长约 ${(pcm.length / (24000 * 2)).toFixed(1)} 秒`)
+    const apiKey = getApiKey()
+    const res = await fetch(`${BASE_URL}/v1/video/veo`, {
+      method: 'POST',
+      headers: { 'Authorization': `Bearer ${apiKey}`, 'Content-Type': 'application/json' },
+      body: JSON.stringify({ model: videoModel, input: prompt, seconds, size }),
+    })
 
-    } else {
-      // 标准 OpenAI 兼容（speech-2.6-hd 等）
-      const res = await request('/audio/speech', {
-        method: 'POST',
-        body: JSON.stringify({
-          model: ttsModel,
-          input: text,
-          voice: flags.voice || 'female-shaonv',
-        }),
-      })
-      if (res.status !== 200) {
-        const err = await res.json()
-        throw new Error(err.error?.message || JSON.stringify(err))
-      }
-      const buf = Buffer.from(await res.arrayBuffer())
-      if (output) { fs.writeFileSync(output, buf); console.log(`✅ 已保存到 ${output}（${buf.length} bytes）`) }
-      else console.log(`✅ TTS 成功，${buf.length} bytes`)
+    if (!res.ok) {
+      const err = await res.json().catch(() => ({}))
+      throw new Error(err.error || `HTTP ${res.status}`)
     }
-    return
-  }
-
-  // ── MUSIC ─────────────────────────────────────────────────────────────────────
-  if (type === 'music') {
-    const musicModel = model || 'chirp-v5'
-
-    if (musicModel === 'chirp-v5') {
-      // Suno 异步接口（路由在 /suno/，不在 /v1/ 下，需用 BASE_URL 直接请求）
-      const apiKey = getApiKey()
-      const res = await fetch(`${BASE_URL}/suno/submit/music`, {
-        method: 'POST',
-        headers: { 'Content-Type': 'application/json', 'Accept': 'application/json', 'Authorization': `Bearer ${apiKey}` },
-        body: JSON.stringify({
-          gpt_description_prompt: prompt,
-          make_instrumental: true,
-          mv: 'chirp-v5',
-          notify_hook: '',
-        }),
-      })
-      const submitData = await res.json()
-      if (submitData.code !== 'success') throw new Error(submitData.message || JSON.stringify(submitData))
-      const taskId = submitData.data
-      console.log(`🎵 Suno 任务已提交，task_id: ${taskId}，等待生成（约 60-90 秒）...`)
 
-      // 轮询结果
-      for (let i = 0; i < 15; i++) {
-        await new Promise(r => setTimeout(r, 10000))
-        const poll = await fetch(`${BASE_URL}/suno/fetch/${taskId}`, {
-          headers: { 'Authorization': `Bearer ${apiKey}` }
-        })
-        const result = await poll.json()
-        if (result.code !== 'success') throw new Error(result.message || JSON.stringify(result))
-        const status = result.data?.status
-        const progress = result.data?.progress || '0%'
-        process.stdout.write(`\r状态: ${status} 进度: ${progress}   `)
-        if (status === 'SUCCESS') {
-          const songs = result.data?.data || []
-          process.stdout.write('\n')
-          console.log(`✅ 生成成功！共 ${songs.length} 首`)
-          songs.forEach((s, i) => {
-            console.log(`  歌曲${i+1}: ${s.title} (${Math.floor(s.duration/60)}m${Math.floor(s.duration%60)}s)`)
-            console.log(`  音频: ${s.audio_url}`)
-          })
-          // 如果指定 output，下载第一首
-          if (output && songs[0]?.audio_url) await downloadAndSave(songs[0].audio_url, output)
-          return
-        }
-        if (status === 'FAILED') throw new Error('Suno 任务失败: ' + JSON.stringify(result.data))
-      }
-      throw new Error(`Suno 生成超时（150s），任务仍在后台运行\n📋 稍后手动查询：skillfree task suno:${taskId}`)
+    const data = await res.json()
 
+    // 服务端已轮询完成，返回 video_base64 或 url
+    if (data.video_base64) {
+      const videoBytes = Buffer.from(data.video_base64, 'base64')
+      const savePath = output || `./veo_${Date.now()}.mp4`
+      fs.writeFileSync(savePath, videoBytes)
+      console.log(`✅ 视频已保存到 ${savePath}（${(videoBytes.length / 1024 / 1024).toFixed(1)} MB）`)
+    } else if (data.url) {
+      if (output) await downloadAndSave(data.url, output)
+      else console.log('🔗 视频链接:', data.url)
     } else {
-      // music-2.5（MiniMax），走 /v1/responses，需要 lyrics
-      const res = await request('/responses', {
-        method: 'POST',
-        body: JSON.stringify({
-          model: 'music-2.5',
-          input: prompt,
-          lyrics: flags.lyrics || `[verse]\n${prompt}`,
-          audio_setting: { sample_rate: 44100, bitrate: 256000, format: 'mp3' },
-          output_format: 'url',
-          stream: false,
-        }),
-      })
-      const data = await res.json()
-      if (data.error) throw new Error(data.error.message || JSON.stringify(data.error))
-      const audioUrl = data.output?.[0]?.content?.[0]?.audio
-      if (!audioUrl) throw new Error('未返回音频: ' + JSON.stringify(data).slice(0, 200))
-      const duration = ((data.extra_info?.music_duration || 0) / 1000).toFixed(1)
-      console.log(`✅ music-2.5 生成成功！时长约 ${duration} 秒`)
-      console.log('音频 URL:', audioUrl)
-      if (output) await downloadAndSave(audioUrl, output)
+      throw new Error('未返回视频数据: ' + JSON.stringify(data).slice(0, 200))
     }
+    return
   }
 
   // ── OCR ───────────────────────────────────────────────────────────────────────
   if (type === 'ocr') {
     let input = prompt
-    // 如果是本地文件路径，读取为 base64
     if (flags.file && fs.existsSync(path.resolve(flags.file))) {
       input = fs.readFileSync(path.resolve(flags.file)).toString('base64')
     }
@@ -293,147 +156,7 @@ async function pilot(flags) {
     return
   }
 
-  // ── STT ───────────────────────────────────────────────────────────────────────
-  if (type === 'stt') {
-    if (!flags.file) throw new Error('--file 是必需的（音频文件路径）')
-    const audioBase64 = fs.readFileSync(path.resolve(flags.file)).toString('base64')
-    const result = await post('/v1/audio/transcriptions', {
-      model: 'whisper-1',
-      file: audioBase64,
-      filename: path.basename(flags.file),
-    })
-    const text = result.text || JSON.stringify(result, null, 2)
-    if (output) { fs.writeFileSync(output, text); console.log(`✅ 已保存到 ${output}`) }
-    else console.log(text)
-    return
-  }
-
-  // ── EMBEDDING ─────────────────────────────────────────────────────────────────
-  if (type === 'embedding') {
-    // doubao-embedding-vision：走 /v1/responses（多模态），input 为数组
-    // 简单文本 embedding 也可用
-    const inputData = flags.file
-      ? [{ type: 'image_url', image_url: { url: 'data:image/png;base64,' + fs.readFileSync(path.resolve(flags.file)).toString('base64') } }]
-      : [{ type: 'text', text: prompt }]
-    const res = await request('/embeddings', {
-      method: 'POST',
-      body: JSON.stringify({
-        model: model || 'doubao-embedding-vision-251215',
-        input: inputData,
-        encoding_format: 'float',
-        dimensions: 1024,
-        sparse_embedding: { type: 'disabled' },
-      }),
-    })
-    const data = await res.json()
-    if (data.error) throw new Error(data.error.message || JSON.stringify(data.error))
-    const embedding = data.data?.[0]?.embedding
-    if (!embedding) throw new Error('未返回向量: ' + JSON.stringify(data).slice(0, 200))
-    console.log(`✅ Embedding 成功！维度: ${embedding.length}`)
-    if (output) {
-      fs.writeFileSync(output, JSON.stringify(embedding))
-      console.log(`已保存到 ${output}`)
-    } else {
-      console.log('前5维:', embedding.slice(0, 5))
-    }
-    return
-  }
-
-  // ── VIDEO ─────────────────────────────────────────────────────────────────────
-  if (type === 'video') {
-    if (!prompt) throw new Error('--prompt 是必需的（视频描述）')
-    const videoModel = model || 'kling-v2-6-text2video'
-
-    // 1. 提交任务
-    process.stdout.write(`🎬 提交视频任务（${videoModel}）...`)
-    const res = await request('/video/generations', {
-      method: 'POST',
-      body: JSON.stringify({ model: videoModel, prompt }),
-    })
-    const submitData = await res.json()
-    if (!submitData.task_id) throw new Error('提交失败: ' + JSON.stringify(submitData).slice(0, 200))
-
-    const taskId = submitData.task_id
-    console.log(` ✅\n📋 task_id: ${taskId}`)
-    console.log('⏳ 视频生成中，预计需要 1-3 分钟...\n')
-
-    // 2. 轮询状态（最多等 5 分钟）
-    const maxWait = 300000
-    const interval = 5000
-    const start = Date.now()
-
-    while (Date.now() - start < maxWait) {
-      await new Promise(r => setTimeout(r, interval))
-      const pollRes = await request(`/tasks/${taskId}`, { method: 'GET' })
-      const task = await pollRes.json()
-      const status = task.status
-
-      process.stdout.write(`\r状态: ${status}  已等待: ${Math.round((Date.now() - start) / 1000)}s   `)
-
-      if (status === 'completed') {
-        console.log('\n')
-        const videoUrl = task.result_url || task.output_url
-        if (!videoUrl) throw new Error('任务完成但未返回视频 URL: ' + JSON.stringify(task))
-
-        if (output) {
-          await downloadAndSave(videoUrl, output)
-        } else {
-          console.log(`✅ 视频生成完成！`)
-          console.log(`🔗 下载链接：${videoUrl}`)
-          console.log(`\n💡 加 --output ./video.mp4 可自动下载到本地`)
-        }
-        return
-      }
-
-      if (status === 'failed') {
-        throw new Error('视频生成失败: ' + (task.error || JSON.stringify(task)))
-      }
-    }
-
-    // 超时，给出手动查询命令
-    console.log(`\n⏰ 等待超时（5分钟），任务仍在后台运行`)
-    console.log(`📋 稍后可手动查询进度：`)
-    console.log(`   skillfree task ${taskId}`)
-    return
-  }
-
-  // ── SEARCH ────────────────────────────────────────────────────────────────────
-  if (type === 'search') {
-    if (!prompt) throw new Error('--prompt 是必需的（搜索词）')
-    const searchModel = model || 'tavily-search'
-    const res = await request('/search', {
-      method: 'POST',
-      body: JSON.stringify({
-        model: searchModel,
-        query: prompt,
-        include_answer: true,
-        max_results: flags.maxResults || 5,
-      }),
-    })
-    const data = await res.json()
-    if (data.error) throw new Error(data.error.message || JSON.stringify(data.error))
-
-    // 输出格式化结果
-    if (data.answer) {
-      console.log('\n📝 摘要答案：')
-      console.log(data.answer)
-    }
-    if (data.results?.length) {
-      console.log(`\n🔗 搜索结果（共 ${data.results.length} 条）：`)
-      data.results.forEach((r, i) => {
-        console.log(`\n${i+1}. ${r.title}`)
-        console.log(`   ${r.url}`)
-        if (r.content) console.log(`   ${r.content.slice(0, 150)}...`)
-      })
-    }
-    if (output) {
-      fs.writeFileSync(output, JSON.stringify(data, null, 2))
-      console.log(`\n✅ 完整结果已保存到 ${output}`)
-    }
-    return
-  }
-
-  throw new Error(`不支持的类型: ${type}，可选: chat | image | tts | stt | music | ocr | video | embedding | search`)
+  throw new Error(`不支持的类型: ${type}\n可选: chat | image | video | ocr`)
 }
 
 module.exports = { pilot }

package/scripts/commands/video.js

@@ -1,97 +1,21 @@
-const { run } = require('./run')
-
 /**
- * Video generation command
- * @param {object} params - Video parameters
- * @param {string} params.model - Model in "vendor/model" format
- * @param {string} params.prompt - Video generation prompt
- * @param {string} [params.output] - Output file path
- * @returns {Promise<object>} Video generation result
+ * Video generation via Veo（异步，服务端轮询完成后返回）
+ * @param {object} params
+ * @param {string} params.model   - veo-3.1-fast-generate-preview | veo-3.1-generate-preview
+ * @param {string} params.prompt  - 视频描述文本
+ * @param {string} [params.seconds] - 时长：4 / 6 / 8（默认 8）
+ * @param {string} [params.size]    - 分辨率：1280x720 / 1920x1080 / 3840x2160（默认 1920x1080）
+ * @param {string} [params.output]  - 本地保存路径（.mp4）
  */
 async function video(params) {
-  if (!params.prompt) {
-    throw new Error('--prompt is required for video generation')
-  }
+  if (!params.prompt) throw new Error('--prompt is required for video generation')
 
-  const [vendor] = params.model.split('/')
-  const inputs = {}
-
-  if (vendor === 'vertex') {
-    // Vertex/Veo uses instances array format
-    inputs.instances = [{ prompt: params.prompt }]
-    inputs.parameters = {}
-  } else if (vendor === 'mm') {
-    // MM video models: t2v (text-to-video), i2v (image-to-video)
-    inputs.prompt = params.prompt
-    if (params.size) {
-      // Convert "1280x720" to "1280*720" if needed
-      inputs.size = params.size.replace('x', '*')
-    }
-    if (params.duration) {
-      inputs.duration = parseInt(params.duration)
-    }
-    if (params.image) {
-      // i2v mode: image-to-video
-      inputs.image_url = params.image
-    }
-  } else {
-    // MiniMax and others use 'prompt'
-    inputs.prompt = params.prompt
+  return {
+    model:   params.model   || 'veo-3.1-fast-generate-preview',
+    input:   params.prompt,
+    seconds: params.seconds || '8',
+    size:    params.size    || '1920x1080',
   }
-
-  return run({ model: params.model, inputs, output: params.output })
-}
-
-/**
- * Multimodal understanding command (video/image/audio analysis)
- * @param {object} params - Multimodal parameters
- * @param {string} params.model - Model in "vendor/model" format (e.g., mm/qwen3-vl-plus)
- * @param {string} params.prompt - Text prompt/question about the media
- * @param {string} [params.video] - Video URL to analyze
- * @param {string} [params.image] - Image URL to analyze
- * @param {string} [params.audio] - Audio URL to analyze/transcribe
- * @returns {Promise<object>} Multimodal analysis result
- */
-async function multimodal(params) {
-  if (!params.prompt) {
-    throw new Error('--prompt is required for multimodal')
-  }
-  if (!params.video && !params.image && !params.audio) {
-    throw new Error('At least one of --video, --image, or --audio is required')
-  }
-
-  const [vendor] = params.model.split('/')
-  const inputs = {}
-
-  if (vendor === 'mm') {
-    // MM multimodal models use messages format
-    const content = []
-    if (params.video) {
-      content.push({ video: params.video })
-      if (params.fps) {
-        content[content.length - 1].fps = parseInt(params.fps)
-      }
-    }
-    if (params.image) {
-      content.push({ image: params.image })
-    }
-    if (params.audio) {
-      content.push({ audio: params.audio })
-    }
-    content.push({ text: params.prompt })
-
-    inputs.input = {
-      messages: [{ role: 'user', content }]
-    }
-  } else {
-    // Generic format
-    inputs.prompt = params.prompt
-    if (params.video) inputs.video_url = params.video
-    if (params.image) inputs.image_url = params.image
-    if (params.audio) inputs.audio_url = params.audio
-  }
-
-  return run({ model: params.model, inputs })
 }
 
-module.exports = { video, multimodal }
+module.exports = { video }

package/scripts/commands/music.js

@@ -1,28 +0,0 @@
-const { run } = require('./run')
-
-/**
- * Music generation command
- * @param {object} params - Music parameters
- * @param {string} params.model - Model in "vendor/model" format
- * @param {string} params.prompt - Music generation prompt
- * @param {number} [params.duration] - Duration in seconds
- * @param {string} [params.output] - Output file path
- * @returns {Promise<object>} Music generation result
- */
-async function music(params) {
-  if (!params.prompt) {
-    throw new Error('--prompt is required for music generation')
-  }
-
-  const inputs = {
-    prompt: params.prompt,
-  }
-
-  if (params.duration) {
-    inputs.duration = parseInt(params.duration)
-  }
-
-  return run({ model: params.model, inputs, output: params.output })
-}
-
-module.exports = { music }