skillfree 0.1.3 → 0.1.5

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "skillfree",
-  "version": "0.1.3",
+  "version": "0.1.5",
   "description": "🦞 一个 API，满足所有龙虾技能需求",
   "main": "bin/skillfree.js",
   "bin": {

package/scripts/commands/pilot.js

@@ -1,100 +1,338 @@
-const { post, postStream, BASE_URL } = require('../lib/client')
+const { post, postStream, get, request, getApiKey, BASE_URL } = require('../lib/client')
 const fs = require('fs')
 const path = require('path')
 
+// ─── 工具：下载文件并保存 ──────────────────────────────────────────────────────
+async function downloadAndSave(url, output) {
+  const resp = await fetch(url)
+  if (!resp.ok) throw new Error(`下载失败: ${resp.status}`)
+  fs.writeFileSync(output, Buffer.from(await resp.arrayBuffer()))
+  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 prompt = flags.prompt || flags.text || ''
   const output = flags.output || null
+  const model = flags.model || null
 
-  // 构造请求体
-  const body = { type, inputs: {} }
-
-  switch (type) {
-    case 'chat':
-      body.model = flags.model
-      body.inputs.messages = [{ role: 'user', content: prompt }]
-      break
-    case 'image':
-      body.model = flags.model
-      body.inputs.prompt = prompt
-      if (flags.size) body.inputs.size = flags.size
-      break
-    case 'tts':
-      body.model = flags.model
-      body.inputs.text  = prompt
-      body.inputs.input = prompt
-      body.inputs.voice = flags.voice || 'nova'
-      break
-    case 'stt':
-      if (flags.file) {
-        body.inputs.audio_data = fs.readFileSync(path.resolve(flags.file)).toString('base64')
-        body.inputs.filename   = path.basename(flags.file)
+  // ── CHAT ─────────────────────────────────────────────────────────────────────
+  if (type === 'chat') {
+    if (!output) {
+      // 流式输出
+      const res = await postStream('/chat/completions', {
+        model: model || 'DeepSeek-V3.2-Fast',
+        messages: [{ role: 'user', content: prompt }],
+        stream: true,
+      })
+      const reader = res.body.getReader()
+      const decoder = new TextDecoder()
+      let buf = ''
+      process.stdout.write('\n')
+      while (true) {
+        const { done, value } = await reader.read()
+        if (done) break
+        buf += decoder.decode(value, { stream: true })
+        const lines = buf.split('\n')
+        buf = lines.pop()
+        for (const line of lines) {
+          if (line.startsWith('data: ')) {
+            const chunk = line.slice(6)
+            if (chunk === '[DONE]') { process.stdout.write('\n'); return }
+            try {
+              const d = JSON.parse(chunk)
+              const text = d.choices?.[0]?.delta?.content
+              if (text) process.stdout.write(text)
+            } catch {}
+          }
+        }
       }
-      break
-    case 'video':
-      body.model = flags.model
-      body.inputs.prompt   = prompt
-      body.inputs.duration = flags.duration ? parseInt(flags.duration) : 5
-      break
-    case 'music':
-      body.model = flags.model
-      body.inputs.prompt = prompt
-      break
-    default:
-      body.inputs.prompt = prompt
+      return
+    }
+    const result = await post('/chat/completions', {
+      model: model || 'DeepSeek-V3.2-Fast',
+      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)
+    return
   }
 
-  // 流式输出（chat 默认流式）
-  if (type === 'chat' && !output) {
-    const res = await postStream('/chat/completions', {
-      model: flags.model || 'DeepSeek-V3.2-Fast',
-      messages: [{ role: 'user', content: prompt }],
-      stream: true,
+  // ── IMAGE ─────────────────────────────────────────────────────────────────────
+  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,
+      n: 1,
+      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}`) }
+    } else {
+      console.log('图像 URL:', url)
+    }
+    return
+  }
+
+  // ── TTS ───────────────────────────────────────────────────────────────────────
+  if (type === 'tts') {
+    const ttsModel = model || 'speech-2.6-hd'
+    const text = prompt
+
+    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`)
 
-    const reader = res.body.getReader()
-    const decoder = new TextDecoder()
-    let buf = ''
-    process.stdout.write('\n')
-    while (true) {
-      const { done, value } = await reader.read()
-      if (done) break
-      buf += decoder.decode(value, { stream: true })
-      const lines = buf.split('\n')
-      buf = lines.pop()
-      for (const line of lines) {
-        if (line.startsWith('data: ')) {
-          const chunk = line.slice(6)
-          if (chunk === '[DONE]') { process.stdout.write('\n'); return }
-          try {
-            const d = JSON.parse(chunk)
-            const text = d.choices?.[0]?.delta?.content
-            if (text) process.stdout.write(text)
-          } catch {}
+    } else if (ttsModel === 'gemini-2.5-pro-preview-tts' || ttsModel === 'gemini-2.5-flash-preview-tts') {
+      // Gemini TTS：走 /v1beta/models/:id:generateContent，认证无 Bearer
+      const apiKey = getApiKey()
+      const res = await fetch(BASE_URL.replace('skillfree.tech', 'dmxapi.cn') + '/v1beta/models/' + ttsModel + ':generateContent', {
+        method: 'POST',
+        headers: { 'Authorization': 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`)
+
+    } 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`)
+    }
+    return
+  }
+
+  // ── MUSIC ─────────────────────────────────────────────────────────────────────
+  if (type === 'music') {
+    const musicModel = model || 'chirp-v5'
+
+    if (musicModel === 'chirp-v5') {
+      // Suno 异步接口
+      const res = await request('/suno/submit/music', {
+        method: 'POST',
+        headers: { 'Accept': 'application/json' },
+        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 request('/suno/fetch/' + taskId, { method: 'GET' })
+        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: ${s.audio_url}`)
+          })
+          // 如果指定 output，下载第一首
+          if (output && songs[0]?.audio_url) await downloadAndSave(songs[0].audio_url, output)
+          return
         }
+        if (status === 'FAILED') throw new Error('Suno 任务失败: ' + JSON.stringify(result.data))
       }
+      throw new Error('Suno 生成超时（150s），请手动查询 task_id: ' + taskId)
+
+    } else {
+      // music-2.5（MiniMax），走 /v1/responses，需要 lyrics
+      const res = await request('/responses', {
+        method: 'POST',
+        body: JSON.stringify({
+          model: 'music-2.5',
+          input: prompt,
+          lyrics: flags.lyrics || `[verse]\n${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)
     }
+  }
+
+  // ── 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')
+    }
+    const res = await request('/responses', {
+      method: 'POST',
+      body: JSON.stringify({
+        model: model || 'hehe-tywd',
+        input,
+        parse_mode: 'auto',
+        dpi: 144,
+        page_start: 0,
+        page_count: 1000,
+        markdown_details: 1,
+        page_details: 0,
+        table_flavor: 'html',
+      }),
+    })
+    const data = await res.json()
+    if (data.code !== 200) throw new Error(data.message || JSON.stringify(data).slice(0, 200))
+    const markdown = data.result?.markdown || ''
+    if (output) { fs.writeFileSync(output, markdown); console.log(`✅ OCR 结果已保存到 ${output}`) }
+    else console.log(markdown)
     return
   }
 
-  // 非流式
-  const result = await post('/pilot', body)
+  // ── 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
+  }
 
-  // 如果有媒体 URL 且指定了输出文件，下载保存
-  const mediaUrl = result.image_url || result.video_url || result.audio_url
-    || result.data?.[0]?.url || result.data?.[0]
-  if (output && mediaUrl) {
-    const resp = await fetch(typeof mediaUrl === 'string' ? mediaUrl : mediaUrl.url)
-    fs.writeFileSync(output, Buffer.from(await resp.arrayBuffer()))
-    console.log(`✅ 已保存到 ${output}`)
+  // ── 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
   }
 
-  // 文本输出
-  const text = result.choices?.[0]?.message?.content
-    || result.text || result.transcript || JSON.stringify(result, null, 2)
-  console.log(text)
+  throw new Error(`不支持的类型: ${type}，可选: chat | image | tts | stt | music | ocr | video | embedding`)
 }
 
 module.exports = { pilot }