@bolloon/bolloon-agent 0.1.30 → 0.1.32

package/src/web/api-config.html

@@ -20,9 +20,47 @@
       <a href="/" class="back-link">← 返回主页</a>
     </div>
 
-    <!-- 提供商列表 -->
-    <div class="provider-list" id="providerList">
-      <div class="loading-state">加载中...</div>
+    <!-- 标签切换 -->
+    <div class="api-tabs">
+      <button class="api-tab active" data-tab="llm" onclick="switchTab('llm')">
+        <span class="api-tab-icon">💬</span>
+        <span>LLM 对话</span>
+      </button>
+      <button class="api-tab" data-tab="video" onclick="switchTab('video')">
+        <span class="api-tab-icon">🎬</span>
+        <span>视频生成</span>
+      </button>
+      <button class="api-tab" data-tab="audio" onclick="switchTab('audio')">
+        <span class="api-tab-icon">🎵</span>
+        <span>音频生成</span>
+      </button>
+    </div>
+
+    <!-- LLM 面板 -->
+    <div class="api-panel" id="panel-llm" data-panel="llm">
+      <div class="provider-list" id="providerList">
+        <div class="loading-state">加载中...</div>
+      </div>
+    </div>
+
+    <!-- 视频生成面板 -->
+    <div class="api-panel" id="panel-video" data-panel="video" style="display: none;">
+      <div class="video-intro">
+        <p>视频生成走异步任务流（提交任务 → 轮询结果），与 LLM 实时对话不同。Seedance 由字节火山方舟 (ARK) 提供，支持文生视频 (t2v) 与图生视频 (i2v)；MiniMax Video 为同 tab 下的另一选项。</p>
+      </div>
+      <div class="provider-list" id="videoProviderList">
+        <div class="loading-state">加载中...</div>
+      </div>
+    </div>
+
+    <!-- 音频生成面板 -->
+    <div class="api-panel" id="panel-audio" data-panel="audio" style="display: none;">
+      <div class="video-intro">
+        <p>音频生成：TTS（文生语音）/ Music（文生音乐）。MiniMax 提供 speech-01 与 music-01 模型，TTS 默认走 OpenAI 兼容 <code>/audio/speech</code> 端点，Music 走 MiniMax 自有 <code>/music_generation</code> 端点。</p>
+      </div>
+      <div class="provider-list" id="audioProviderList">
+        <div class="loading-state">加载中...</div>
+      </div>
     </div>
 
     <!-- 配置弹窗 -->
@@ -43,7 +81,7 @@
           <div class="form-group">
             <label>API Key</label>
             <input type="password" id="apiKeyInput" placeholder="输入 API Key">
-            <p class="form-hint">已有 Key 会保留，输入新值以更新</p>
+            <p class="form-hint" id="apiKeyHint">已有 Key 会保留，输入新值以更新</p>
           </div>
 
           <div class="form-group">
@@ -52,11 +90,79 @@
             <p class="form-hint">留空使用默认地址</p>
           </div>
 
-          <div class="form-row">
+          <div class="form-group">
+            <label>模型</label>
+            <input type="text" id="modelInput" placeholder="如 gpt-4">
+            <p class="form-hint" id="modelHint"></p>
+          </div>
+
+          <!-- 视频模型专用字段 -->
+          <div class="video-only-fields" style="display: none;">
+            <div class="form-row">
+              <div class="form-group">
+                <label>分辨率</label>
+                <select id="resolutionInput">
+                  <option value="480p">480p</option>
+                  <option value="720p">720p</option>
+                  <option value="1080p">1080p</option>
+                </select>
+              </div>
+              <div class="form-group">
+                <label>宽高比</label>
+                <select id="ratioInput">
+                  <option value="16:9">16:9</option>
+                  <option value="9:16">9:16</option>
+                  <option value="1:1">1:1</option>
+                  <option value="4:3">4:3</option>
+                  <option value="3:4">3:4</option>
+                  <option value="21:9">21:9</option>
+                </select>
+              </div>
+            </div>
+            <div class="form-group">
+              <label>默认时长（秒）</label>
+              <input type="number" id="durationInput" min="3" max="12" step="1" value="5">
+              <p class="form-hint">Seedance 支持 3-12 秒</p>
+            </div>
+          </div>
+
+          <!-- 音频模型专用字段 -->
+          <div class="audio-only-fields" style="display: none;">
             <div class="form-group">
-              <label>模型</label>
-              <input type="text" id="modelInput" placeholder="如 gpt-4">
+              <label>音色 / 风格</label>
+              <input type="text" id="voiceInput" placeholder="male-qn-jingying / female-shaonv ...">
+              <p class="form-hint" id="voiceHint"></p>
+            </div>
+            <div class="form-row">
+              <div class="form-group">
+                <label>语速</label>
+                <input type="number" id="speedInput" min="0.5" max="2.0" step="0.1" value="1.0">
+              </div>
+              <div class="form-group">
+                <label>输出格式</label>
+                <select id="formatInput">
+                  <option value="mp3">mp3</option>
+                  <option value="pcm">pcm</option>
+                  <option value="wav">wav</option>
+                </select>
+              </div>
+            </div>
+            <div class="form-group" id="audioModeGroup" style="display: none;">
+              <label>模式</label>
+              <select id="modeInput">
+                <option value="instrumental">纯音乐</option>
+                <option value="lyrics">带歌词</option>
+              </select>
+            </div>
+            <div class="form-group" id="audioDurationGroup" style="display: none;">
+              <label>默认时长（秒）</label>
+              <input type="number" id="audioDurationInput" min="5" max="120" step="1" value="30">
+              <p class="form-hint">music-01 支持 5-120 秒</p>
             </div>
+          </div>
+
+          <!-- LLM 专用字段 -->
+          <div class="llm-only-fields">
             <div class="form-group form-group-small">
               <label>温度</label>
               <input type="number" id="temperatureInput" value="0.7" min="0" max="2" step="0.1">
@@ -80,49 +186,115 @@
   </div>
 
   <script>
-    // 当前配置数据
-    let configData = null;
+    // 当前 tab
+    let currentTab = 'llm';
+
+    // 缓存三份配置
+    let llmConfigData = null;
+    let videoConfigData = null;
+    let audioConfigData = null;
     let currentProvider = null;
+    let currentProviderType = 'llm'; // 'llm' | 'video' | 'audio'
+
+    // ==================== Tab 切换 ====================
+    function switchTab(tab) {
+      currentTab = tab;
+      document.querySelectorAll('.api-tab').forEach(btn => {
+        btn.classList.toggle('active', btn.dataset.tab === tab);
+      });
+      document.querySelectorAll('.api-panel').forEach(panel => {
+        panel.style.display = panel.dataset.panel === tab ? 'block' : 'none';
+      });
+      // 切换 tab 时更新计数 badge
+      updateCountBadge();
+    }
 
-    // 加载配置
-    async function loadConfig() {
+    function updateCountBadge() {
+      const data = currentTab === 'llm' ? llmConfigData :
+                   currentTab === 'video' ? videoConfigData : audioConfigData;
+      if (!data) {
+        document.getElementById('configCount').textContent = '加载中...';
+        return;
+      }
+      let configured = 0, total = 0;
+      for (const key in data.providers) {
+        total++;
+        const p = data.providers[key];
+        if (currentTab === 'llm') {
+          if (p.enabled && p.apiKey) configured++;
+        } else {
+          const info = data.providerInfo[key] || {};
+          if (p.enabled && (!info.requiresApiKey || p.apiKey)) configured++;
+        }
+      }
+      const labelMap = { llm: '', video: '视频', audio: '音频' };
+      const label = labelMap[currentTab] || '';
+      document.getElementById('configCount').textContent = configured + '/' + total + ' ' + label + '已配置';
+    }
+
+    // ==================== 加载 ====================
+    async function loadAll() {
+      await Promise.all([loadLLMConfig(), loadVideoConfig(), loadAudioConfig()]);
+    }
+
+    async function loadLLMConfig() {
       try {
         const resp = await fetch('/api/llm-config');
-        configData = await resp.json();
-        renderProviders();
+        llmConfigData = await resp.json();
+        renderProviders('llm');
+        if (currentTab === 'llm') updateCountBadge();
       } catch (err) {
         document.getElementById('providerList').innerHTML =
           '<div class="error-state">加载失败: ' + err.message + '</div>';
       }
     }
 
-    // 渲染提供商列表
-    function renderProviders() {
-      const list = document.getElementById('providerList');
-      const providers = configData.providers;
-      const info = configData.providerInfo;
+    async function loadVideoConfig() {
+      try {
+        const resp = await fetch('/api/video-config');
+        videoConfigData = await resp.json();
+        renderProviders('video');
+        if (currentTab === 'video') updateCountBadge();
+      } catch (err) {
+        document.getElementById('videoProviderList').innerHTML =
+          '<div class="error-state">加载失败: ' + err.message + '</div>';
+      }
+    }
 
-      // 计算已配置数量
-      let configured = 0;
-      let total = 0;
-      for (const key in providers) {
-        total++;
-        if (providers[key].enabled && providers[key].apiKey) {
-          configured++;
-        }
+    async function loadAudioConfig() {
+      try {
+        const resp = await fetch('/api/audio-config');
+        audioConfigData = await resp.json();
+        renderProviders('audio');
+        if (currentTab === 'audio') updateCountBadge();
+      } catch (err) {
+        document.getElementById('audioProviderList').innerHTML =
+          '<div class="error-state">加载失败: ' + err.message + '</div>';
       }
-      document.getElementById('configCount').textContent = configured + '/' + total + ' 已配置';
+    }
+
+    // ==================== 渲染 ====================
+    function renderProviders(type) {
+      const data = type === 'llm' ? llmConfigData :
+                   type === 'video' ? videoConfigData : audioConfigData;
+      const listEl = document.getElementById(
+        type === 'llm' ? 'providerList' :
+        type === 'video' ? 'videoProviderList' : 'audioProviderList'
+      );
+      if (!data) return;
+
+      const providers = data.providers;
+      const info = data.providerInfo;
+      const isActive = (key) => data.activeProvider === key;
 
-      // 生成卡片
       let html = '';
       for (const key in providers) {
         const p = providers[key];
         const i = info[key] || { name: key, description: '供应商', requiresApiKey: true };
-        const isActive = configData.activeProvider === key;
+        const active = isActive(key);
 
-        // 状态判断
         let statusText, statusClass;
-        if (isActive) {
+        if (active) {
           statusText = '使用中';
           statusClass = 'status-accent';
         } else if (p.enabled && p.apiKey) {
@@ -136,16 +308,16 @@
           statusClass = 'status-muted';
         }
 
-        // 配置状态
         const isConfigured = p.enabled && p.apiKey;
 
         html += `
-          <div class="provider-card ${isActive ? 'active' : ''}" onclick="openModal('${key}')">
+          <div class="provider-card ${active ? 'active' : ''}" onclick="openModal('${type}', '${key}')">
             <div class="provider-header">
               <div class="provider-icon">${i.name.charAt(0).toUpperCase()}</div>
               <div class="provider-info">
                 <h3 class="provider-name">${i.name}</h3>
                 <p class="provider-desc">${i.description}</p>
+                ${i.docs ? '<a class="provider-docs" href="' + i.docs + '" target="_blank" onclick="event.stopPropagation()">📖 文档</a>' : ''}
               </div>
               <div class="status-badge ${statusClass}">
                 <span class="status-dot"></span>
@@ -163,35 +335,75 @@
           </div>
         `;
       }
-      list.innerHTML = html;
+      listEl.innerHTML = html;
     }
 
-    // 打开弹窗
-    function openModal(providerKey) {
+    // ==================== 弹窗 ====================
+    function openModal(type, providerKey) {
+      currentProviderType = type;
       currentProvider = providerKey;
-      const p = configData.providers[providerKey];
-      const i = configData.providerInfo[providerKey] || { name: providerKey, description: '供应商' };
+      const data = type === 'llm' ? llmConfigData :
+                   type === 'video' ? videoConfigData : audioConfigData;
+      const p = data.providers[providerKey];
+      const i = data.providerInfo[providerKey] || { name: providerKey, description: '供应商' };
 
       document.getElementById('modalIcon').textContent = i.name.charAt(0).toUpperCase();
       document.getElementById('modalTitle').textContent = '配置 ' + i.name;
       document.getElementById('modalSubtitle').textContent = i.description;
 
       document.getElementById('apiKeyInput').value = '';
+      document.getElementById('apiKeyHint').textContent = p.apiKey
+        ? '当前已配置 (***' + (p.apiKey.slice(-4) || '') + ')，输入新值以更新'
+        : '输入 API Key';
       document.getElementById('baseUrlInput').value = p.baseUrl || '';
       document.getElementById('modelInput').value = p.model || '';
-      document.getElementById('temperatureInput').value = p.temperature || 0.7;
+
+      // 视频/音频/LLM 专用字段显隐
+      const isVideo = type === 'video';
+      const isAudio = type === 'audio';
+      document.querySelector('.video-only-fields').style.display = isVideo ? 'block' : 'none';
+      document.querySelector('.audio-only-fields').style.display = isAudio ? 'block' : 'none';
+      document.querySelector('.llm-only-fields').style.display = (!isVideo && !isAudio) ? 'block' : 'none';
+
+      if (isVideo) {
+        document.getElementById('resolutionInput').value = p.resolution || '720p';
+        document.getElementById('ratioInput').value = p.ratio || '16:9';
+        document.getElementById('durationInput').value = p.duration || 5;
+        document.getElementById('modelHint').textContent = '示例: doubao-seedance-1-0-lite-t2v-250428 (文生视频) 或 ...-i2v-... (图生视频)';
+      } else if (isAudio) {
+        const isMusic = (i.kind === 'music');
+        // TTS 字段
+        document.getElementById('voiceInput').parentElement.style.display = isMusic ? 'none' : 'block';
+        document.getElementById('speedInput').parentElement.parentElement.style.display = isMusic ? 'none' : 'flex';
+        document.getElementById('modeInput').parentElement.style.display = isMusic ? 'block' : 'none';
+        document.getElementById('audioDurationInput').parentElement.style.display = isMusic ? 'block' : 'none';
+
+        if (!isMusic) {
+          document.getElementById('voiceInput').value = p.voice || 'male-qn-jingying';
+          document.getElementById('voiceHint').textContent = '常用音色: male-qn-jingying, female-shaonv, female-yujie, presenter_male, presenter_female';
+          document.getElementById('speedInput').value = p.speed || 1.0;
+          document.getElementById('formatInput').value = p.format || 'mp3';
+        } else {
+          document.getElementById('modeInput').value = p.mode || 'instrumental';
+          document.getElementById('audioDurationInput').value = p.duration || 30;
+        }
+        document.getElementById('modelHint').textContent = isMusic
+          ? '示例: music-01'
+          : '示例: speech-01 (TTS) / asr-01 (语音转写)';
+      } else {
+        document.getElementById('temperatureInput').value = p.temperature || 0.7;
+      }
 
       document.getElementById('testResult').style.display = 'none';
       document.getElementById('configModal').style.display = 'flex';
     }
 
-    // 关闭弹窗
     function closeModal() {
       document.getElementById('configModal').style.display = 'none';
       currentProvider = null;
     }
 
-    // 测试连接
+    // ==================== 测试连接 ====================
     async function testConnection() {
       const btn = document.getElementById('testBtn');
       const result = document.getElementById('testResult');
@@ -200,8 +412,12 @@
       btn.innerHTML = '<span class="spinner"></span> 测试中...';
       result.style.display = 'none';
 
+      const endpoint = currentProviderType === 'llm' ? '/api/llm-test' :
+                       currentProviderType === 'video' ? '/api/video-test' :
+                       '/api/audio-test';
+
       try {
-        const resp = await fetch('/api/llm-test', {
+        const resp = await fetch(endpoint, {
           method: 'POST',
           headers: { 'Content-Type': 'application/json' },
           body: JSON.stringify({ provider: currentProvider })
@@ -222,44 +438,71 @@
       btn.innerHTML = '⚡ 测试连接';
     }
 
-    // 保存配置
+    // ==================== 保存 ====================
     document.getElementById('configForm').onsubmit = async function(e) {
       e.preventDefault();
 
+      const data = currentProviderType === 'llm' ? llmConfigData : videoConfigData;
+      const existing = data.providers[currentProvider];
+      const apiKeyVal = document.getElementById('apiKeyInput').value;
+
       const updateData = {
         enabled: true,
-        apiKey: document.getElementById('apiKeyInput').value || configData.providers[currentProvider].apiKey || '',
-        baseUrl: document.getElementById('baseUrlInput').value || configData.providers[currentProvider].baseUrl || '',
-        model: document.getElementById('modelInput').value || configData.providers[currentProvider].model || '',
-        temperature: parseFloat(document.getElementById('temperatureInput').value) || 0.7
+        apiKey: apiKeyVal || existing.apiKey || '',
+        baseUrl: document.getElementById('baseUrlInput').value || existing.baseUrl || '',
+        model: document.getElementById('modelInput').value || existing.model || ''
       };
 
+      if (currentProviderType === 'video') {
+        updateData.resolution = document.getElementById('resolutionInput').value;
+        updateData.ratio = document.getElementById('ratioInput').value;
+        updateData.duration = parseInt(document.getElementById('durationInput').value) || 5;
+      } else if (currentProviderType === 'audio') {
+        const info = audioConfigData.providerInfo[currentProvider] || {};
+        const isMusic = info.kind === 'music';
+        if (isMusic) {
+          updateData.mode = document.getElementById('modeInput').value;
+          updateData.duration = parseInt(document.getElementById('audioDurationInput').value) || 30;
+        } else {
+          updateData.voice = document.getElementById('voiceInput').value;
+          updateData.speed = parseFloat(document.getElementById('speedInput').value) || 1.0;
+          updateData.format = document.getElementById('formatInput').value;
+        }
+      } else {
+        updateData.temperature = parseFloat(document.getElementById('temperatureInput').value) || 0.7;
+      }
+
+      const endpoint = currentProviderType === 'llm' ? '/api/llm-config' :
+                       currentProviderType === 'video' ? '/api/video-config' :
+                       '/api/audio-config';
+
       try {
-        await fetch('/api/llm-config', {
+        await fetch(endpoint, {
           method: 'POST',
           headers: { 'Content-Type': 'application/json' },
           body: JSON.stringify({ provider: currentProvider, config: updateData })
         });
 
-        // 显示保存成功
         const saveBtn = this.querySelector('.btn-save');
         saveBtn.textContent = '✓ 已保存';
         setTimeout(() => {
           closeModal();
-          loadConfig();
+          if (currentProviderType === 'llm') loadLLMConfig();
+          else if (currentProviderType === 'video') loadVideoConfig();
+          else loadAudioConfig();
         }, 500);
       } catch (err) {
         alert('保存失败: ' + err.message);
       }
     };
 
-    // 点击遮罩关闭弹窗
+    // ==================== 关闭弹窗 ====================
     document.getElementById('configModal').onclick = function(e) {
       if (e.target === this) closeModal();
     };
 
     // 启动
-    loadConfig();
+    loadAll();
   </script>
 </body>
-</html>
+</html>

package/src/web/server.ts

@@ -18,6 +18,8 @@ import { documentReader } from '../documents/reader.js';
 import { initMinimax, getMinimax } from '../constraints/index.js';
 import { createAgentSession, type AgentSession, type StreamCallback, type StreamEvent } from '../agents/pi-sdk.js';
 import { llmConfigStore, type ModelProvider, PROVIDER_INFO } from '../llm/config-store.js';
+import { videoConfigStore, type VideoProvider } from '../llm/video-config-store.js';
+import { audioConfigStore, type AudioProvider } from '../llm/audio-config-store.js';
 import { irohTransport } from '../network/iroh-transport.js';
 import { createAgentDelegateApp } from './agent-delegate-server.js';
 import { createIrohDelegateTransport } from './iroh-delegate-transport.js';
@@ -2276,6 +2278,130 @@ app.get('/channels', async (_req, res) => {
     }
   });
 
+  // ==================== 视频生成配置 (Seedance 等) ====================
+
+  // 获取视频生成配置
+  app.get('/api/video-config', async (req, res) => {
+    try {
+      const config = await videoConfigStore.getConfig();
+      const providerInfo = videoConfigStore.getAllProviderInfo();
+
+      // 脱敏：不返回 apiKey 明文
+      const masked = Object.fromEntries(
+        Object.entries(config.providers).map(([key, val]) => [
+          key,
+          { ...val, apiKey: val.apiKey ? '***' + val.apiKey.slice(-4) : '' }
+        ])
+      );
+
+      res.json({
+        activeProvider: config.activeProvider,
+        providers: masked,
+        providerInfo
+      });
+    } catch (err: any) {
+      res.status(500).json({ error: err.message });
+    }
+  });
+
+  // 更新视频供应商配置
+  app.post('/api/video-config', async (req, res) => {
+    try {
+      const { provider, config } = req.body;
+
+      if (!provider || !config) {
+        return res.status(400).json({ error: 'provider and config required' });
+      }
+
+      // 如果前端发的是掩码（***xxx），从当前配置里取真实 key
+      const currentConfig = await videoConfigStore.getProvider(provider as VideoProvider);
+      if (currentConfig && config.apiKey && config.apiKey.startsWith('***')) {
+        config.apiKey = currentConfig.apiKey;
+      }
+
+      await videoConfigStore.updateProvider(provider as VideoProvider, config);
+      res.json({ ok: true });
+    } catch (err: any) {
+      res.status(500).json({ error: err.message });
+    }
+  });
+
+  // 测试视频供应商连接
+  app.post('/api/video-test', async (req, res) => {
+    try {
+      const { provider } = req.body;
+
+      if (!provider) {
+        return res.status(400).json({ error: 'provider required' });
+      }
+
+      const result = await videoConfigStore.testProvider(provider as VideoProvider);
+      res.json(result);
+    } catch (err: any) {
+      res.status(500).json({ error: err.message });
+    }
+  });
+
+  // ==================== 音频生成配置 (TTS / Music) ====================
+
+  // 获取音频配置
+  app.get('/api/audio-config', async (req, res) => {
+    try {
+      const config = await audioConfigStore.getConfig();
+      const providerInfo = audioConfigStore.getAllProviderInfo();
+
+      const masked = Object.fromEntries(
+        Object.entries(config.providers).map(([key, val]) => [
+          key,
+          { ...val, apiKey: val.apiKey ? '***' + val.apiKey.slice(-4) : '' }
+        ])
+      );
+
+      res.json({
+        activeProvider: config.activeProvider,
+        providers: masked,
+        providerInfo
+      });
+    } catch (err: any) {
+      res.status(500).json({ error: err.message });
+    }
+  });
+
+  // 更新音频供应商配置
+  app.post('/api/audio-config', async (req, res) => {
+    try {
+      const { provider, config } = req.body;
+      if (!provider || !config) {
+        return res.status(400).json({ error: 'provider and config required' });
+      }
+
+      // 掩码回写真实 key
+      const currentConfig = await audioConfigStore.getProvider(provider as AudioProvider);
+      if (currentConfig && config.apiKey && config.apiKey.startsWith('***')) {
+        config.apiKey = currentConfig.apiKey;
+      }
+
+      await audioConfigStore.updateProvider(provider as AudioProvider, config);
+      res.json({ ok: true });
+    } catch (err: any) {
+      res.status(500).json({ error: err.message });
+    }
+  });
+
+  // 测试音频供应商连接
+  app.post('/api/audio-test', async (req, res) => {
+    try {
+      const { provider } = req.body;
+      if (!provider) {
+        return res.status(400).json({ error: 'provider required' });
+      }
+      const result = await audioConfigStore.testProvider(provider as AudioProvider);
+      res.json(result);
+    } catch (err: any) {
+      res.status(500).json({ error: err.message });
+    }
+  });
+
   // 统一 AI 解析入口：CLI / 接收方节点 调这里完成 LLM + judgment + harness
   // 入参: { text, mimeType, fileName, fromNodeId, source }
   // 出参: { summary, qualityScore, judgmentId?, gateArtifact? }