@lightcone-ai/daemon 0.15.45 → 0.15.47

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lightcone-ai/daemon",
-  "version": "0.15.45",
+  "version": "0.15.47",
   "type": "module",
   "main": "src/index.js",
   "bin": {

package/src/chat-bridge.js

@@ -2,17 +2,9 @@
 import { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
 import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js';
 import { z } from 'zod';
-import { createReadStream, existsSync, mkdirSync, readFileSync, rmSync, writeFileSync } from 'fs';
+import { createReadStream, existsSync, mkdirSync, readFileSync, writeFileSync } from 'fs';
 import { createHash, randomUUID } from 'crypto';
 import path, { extname } from 'path';
-import { homedir } from 'os';
-import {
-  buildAssContent,
-  concatVideos,
-  muxAudioToVideo,
-  probeDurationMs,
-  transcodeForPlatform,
-} from './_vendor/video/composer/index.js';
 import { recordUrlNarration } from './_vendor/video/recorder/index.js';
 import {
   VIDEO_EXT,
@@ -22,6 +14,7 @@ import {
 import { runRecordUrlNarrationTool } from './record-url-narration-tool.js';
 import { runSubmitToLibraryTool } from './submit-to-library-tool.js';
 import { runRenderTextToImageTool } from './tools/render-text-to-image.js';
+import { runRenderHtmlToImageTool } from './tools/render-html-to-image.js';
 import { runSynthesisTtsTool } from './tools/synthesize-tts.js';
 import { runComposeVideoV2Tool } from './tools/compose-video-v2.js';
 import { runTakePageScreenshotTool } from './tools/take-page-screenshot.js';
@@ -74,18 +67,13 @@ function redactTokenPrefix(token) {
 // Current active workspaceId for memory isolation (defaults to spawn-time WORKSPACE_ID)
 let currentWorkspaceId = WORKSPACE_ID;
 
-const VOICEOVER_LOCAL_DIR = path.join(WORKSPACE_DIR, 'artifacts', 'audio');
-const VIDEO_COMPOSE_LOCAL_DIR = path.join(WORKSPACE_DIR, 'artifacts', 'video');
-const DEFAULT_OUTRO_PATH = path.join(homedir(), '.lightcone', 'assets', 'outros', 'default.mp4');
 // Temporary: block legacy video pipeline tools for a specific editor_in_chief agent.
 // Set via env so this doesn't need a code change when workspace/agent IDs rotate.
 // Remove entirely once the new atomic tool framework is stable and the legacy pipeline retires.
 const CVMAX_WORKSPACE_ID = process.env.BLOCKED_EDITOR_WORKSPACE_ID ?? '';
 const CVMAX_EDITOR_IN_CHIEF_AGENT_ID = process.env.BLOCKED_EDITOR_AGENT_ID ?? '';
 const CVMAX_EDITOR_BLOCKED_VIDEO_TOOLS = new Set([
-  'generate_voiceover',
   'record_url_narration',
-  'compose_video',
   'submit_to_library',
 ]);
 
@@ -131,27 +119,6 @@ function cvmaxEditorVideoToolError(toolName) {
   };
 }
 
-function normalizeVoiceFormat(value) {
-  const normalized = String(value ?? '').trim().toLowerCase();
-  if (!normalized) return 'mp3';
-  if (['mp3', 'wav', 'flac'].includes(normalized)) return normalized;
-  return 'mp3';
-}
-
-function inferAudioExtension(url, format = 'mp3') {
-  const normalizedFormat = normalizeVoiceFormat(format);
-  if (typeof url === 'string' && url.trim()) {
-    try {
-      const pathname = new URL(url).pathname;
-      const ext = extname(pathname).toLowerCase();
-      if (ext && ['.mp3', '.wav', '.flac'].includes(ext)) return ext;
-    } catch {
-      // noop
-    }
-  }
-  return `.${normalizedFormat}`;
-}
-
 function isInsideDir(filePath, dir) {
   const rel = path.relative(dir, filePath);
   return rel === '' || (!!rel && !rel.startsWith('..') && !path.isAbsolute(rel));
@@ -167,58 +134,6 @@ function resolveLocalWorkspaceFile(filePath) {
   throw new Error(`Local file must be inside the agent workspace or workspace shared artifacts/notes/tmp directories. Got: ${filePath}`);
 }
 
-function normalizeComposePath(filePath, label) {
-  const normalized = String(filePath ?? '').trim();
-  if (!normalized) throw new Error(`${label} is required.`);
-  return path.resolve(WORKSPACE_DIR, normalized);
-}
-
-function normalizeComposeTarget(value) {
-  const normalized = String(value ?? '').trim().toLowerCase();
-  if (!normalized) return 'short_video_cn';
-  if (['short_video_cn', 'douyin', 'xhs'].includes(normalized)) return normalized;
-  throw new Error(`Unsupported compose target: ${value}`);
-}
-
-function normalizeComposeAudioSegments(audioSegments) {
-  if (!Array.isArray(audioSegments) || audioSegments.length === 0) {
-    throw new Error('audio_segments must be a non-empty array.');
-  }
-
-  return audioSegments.map((segment, index) => {
-    if (!segment || typeof segment !== 'object' || Array.isArray(segment)) {
-      throw new Error(`audio_segments[${index}] must be an object.`);
-    }
-    const audioPath = normalizeComposePath(
-      segment.audio_path ?? segment.audioPath,
-      `audio_segments[${index}].audio_path`
-    );
-    const startMsRaw = segment.start_ms ?? segment.startMs;
-    const startMs = startMsRaw == null ? null : Number(startMsRaw);
-    if (startMsRaw != null && (!Number.isFinite(startMs) || startMs < 0)) {
-      throw new Error(`audio_segments[${index}].start_ms must be a non-negative number.`);
-    }
-    const phase = String(segment.phase ?? segment.phase_id ?? segment.phaseId ?? '').trim();
-    if (startMs == null && !phase) {
-      throw new Error(`audio_segments[${index}] requires start_ms, or provide phase with events_log.`);
-    }
-
-    const normalized = { audio_path: audioPath };
-    if (startMs != null) normalized.start_ms = Math.floor(startMs);
-    if (phase) normalized.phase = phase;
-    return {
-      ...normalized,
-    };
-  });
-}
-
-function cleanupLocalFiles(paths = []) {
-  for (const filePath of paths) {
-    if (!filePath) continue;
-    try { rmSync(filePath, { force: true }); } catch { /* noop */ }
-  }
-}
-
 const DEFAULT_TOOL_CLASSIFICATION = {
   check_messages: 'local',
   list_memory: 'local',
@@ -246,9 +161,7 @@ const DEFAULT_TOOL_CLASSIFICATION = {
   update_goal_field: 'mandatory',
   supersede_goal_field: 'mandatory',
   request_credential_auth: 'mandatory',
-  generate_voiceover: 'mandatory',
   record_url_narration: 'mandatory',
-  compose_video: 'mandatory',
   submit_to_library: 'mandatory',
   register_data_source: 'mandatory',
   bind_workspace_scenario: 'mandatory',
@@ -384,7 +297,6 @@ function inferToolForApi(method, apiPath, body) {
   if (method === 'POST' && cleanPath === '/goal-fields/update') return 'update_goal_field';
   if (method === 'POST' && cleanPath === '/goal-fields/supersede') return 'supersede_goal_field';
   if (method === 'POST' && cleanPath === '/credential-auth/request') return 'request_credential_auth';
-  if (method === 'POST' && cleanPath === '/tts/voiceover') return 'generate_voiceover';
   if (method === 'POST' && cleanPath === '/content-library/submit') return 'submit_to_library';
   if (method === 'POST' && cleanPath === '/api/data-sources') return 'register_data_source';
   if (method === 'POST' && cleanPath === '/orchestrate/decision') return 'write_governance_decision';
@@ -1421,6 +1333,19 @@ server.tool('render_text_to_image',
   async (args) => runRenderTextToImageTool(args)
 );
 
+// ── render_html_to_image ───────────────────────────────────────────────────────
+server.tool('render_html_to_image',
+  'Render a raw HTML string to a PNG image by navigating to it as a local file:// page. Unlike evaluate_script+document.write on about:blank, this preserves file:// origin so <img src="file:///..."> references load correctly. Returns the output image path.',
+  {
+    html: z.string().describe('Full HTML document to render (including <!doctype>, <html>, <head>, <body>).'),
+    output_path: z.string().optional().describe('Absolute path to save the PNG. Auto-generated in /tmp if omitted.'),
+    viewport_width: z.number().optional().describe('Viewport width in pixels. Default 1080.'),
+    viewport_height: z.number().optional().describe('Viewport height in pixels. Default 1920.'),
+    wait_until: z.enum(['load', 'networkidle', 'domcontentloaded']).optional().describe('Navigation wait condition. Default load.'),
+  },
+  async (args) => runRenderHtmlToImageTool(args)
+);
+
 // ── synthesize_tts ─────────────────────────────────────────────────────────────
 server.tool('synthesize_tts',
   'Convert text to speech using the workspace MiniMax TTS credential. Returns a local mp3 file path and duration. Use this to generate narration audio for individual video segments.',
@@ -1481,89 +1406,9 @@ server.tool('get_library_file',
   async (args) => runGetLibraryFileTool({ ...args, currentWorkspaceId, api, SERVER_URL, MACHINE_API_KEY, workspaceDir: WORKSPACE_DIR })
 );
 
-// ── generate_voiceover ─────────────────────────────────────────────────────────
-server.tool('generate_voiceover',
-  'Generate a TTS voiceover using an authorized tts_provider credential and return a local audio file path.',
-  {
-    workspace_id: z.string().optional().describe('Target workspace id. Defaults to current workspace context.'),
-    text: z.string().describe('Text content to synthesize.'),
-    voice_preset: z.string().optional().describe('Platform-neutral voice preset id, e.g. "warm_female_zh_01".'),
-    speed: z.number().optional().describe('Speech speed. Typical range is 0.5 to 2.0.'),
-    format: z.enum(['mp3', 'wav', 'flac']).optional().describe('Audio format. Defaults to mp3.'),
-    credential_id: z.string().optional().describe('Optional explicit credential id. If omitted, uses latest granted tts_provider credential.'),
-  },
-  async ({ workspace_id, text, voice_preset, speed, format, credential_id }) => {
-    if (isBlockedCvmaxEditorVideoTool('generate_voiceover')) {
-      return cvmaxEditorVideoToolError('generate_voiceover');
-    }
-    const targetWorkspaceId = (workspace_id ?? currentWorkspaceId ?? WORKSPACE_ID ?? '').trim();
-    if (!targetWorkspaceId) {
-      return { isError: true, content: [{ type: 'text', text: 'workspace_id is required (no current workspace context).' }] };
-    }
-
-    const normalizedText = String(text ?? '').trim();
-    if (!normalizedText) {
-      return { isError: true, content: [{ type: 'text', text: 'text is required for generate_voiceover.' }] };
-    }
-
-    const normalizedSpeed = speed == null ? 1 : Number(speed);
-    if (!Number.isFinite(normalizedSpeed)) {
-      return { isError: true, content: [{ type: 'text', text: 'speed must be numeric.' }] };
-    }
-
-    const normalizedFormat = normalizeVoiceFormat(format);
-    const payload = {
-      workspace_id: targetWorkspaceId,
-      text: normalizedText,
-      speed: normalizedSpeed,
-      format: normalizedFormat,
-    };
-    if (voice_preset) payload.voice_preset = String(voice_preset).trim();
-    if (credential_id) payload.credential_id = String(credential_id).trim();
-
-    const data = await api('POST', '/tts/voiceover', payload);
-    const remoteAudioUrl = String(data.audio_url ?? '').trim();
-    if (!remoteAudioUrl) {
-      return { isError: true, content: [{ type: 'text', text: 'Voiceover API did not return audio_url.' }] };
-    }
-
-    const downloadRes = await fetch(remoteAudioUrl, {
-      method: 'GET',
-      headers: { 'Authorization': `Bearer ${MACHINE_API_KEY}` },
-    });
-    if (!downloadRes.ok) {
-      return {
-        isError: true,
-        content: [{ type: 'text', text: `Failed to download synthesized audio (${downloadRes.status})` }],
-      };
-    }
-
-    const fileBuffer = Buffer.from(await downloadRes.arrayBuffer());
-    mkdirSync(VOICEOVER_LOCAL_DIR, { recursive: true });
-    const audioExt = inferAudioExtension(remoteAudioUrl, data.format ?? normalizedFormat);
-    const localFileName = `voiceover-${Date.now()}-${randomUUID().slice(0, 8)}${audioExt}`;
-    const localAudioPath = path.join(VOICEOVER_LOCAL_DIR, localFileName);
-    writeFileSync(localAudioPath, fileBuffer);
-
-    return {
-      content: [{
-        type: 'text',
-        text:
-          `Voiceover generated.\n` +
-          `workspace_id=${data.workspace_id ?? targetWorkspaceId}\n` +
-          `local_audio_path=${localAudioPath}\n` +
-          `duration_ms=${data.duration_ms ?? 'unknown'}\n` +
-          `sample_rate=${data.sample_rate ?? 'unknown'}\n` +
-          `format=${data.format ?? normalizedFormat}\n` +
-          `size=${formatBytes(fileBuffer.length)}`,
-      }],
-    };
-  }
-);
-
 // ── record_url_narration ────────────────────────────────────────────────────────
 server.tool('record_url_narration',
-  'Record a silent video of a URL by orchestrating Xvfb + Chromium + ffmpeg, driven by a video plan. Outputs a silent mp4 plus an events.json timestamp log that compose_video can use to align audio segments.\n\nUse this as the canonical recording step for URL-narration videos. Falls back: if the page needs interactions outside the visual_action vocabulary (clicks, waits, OCR loops), use Monitor (Bash) with custom Playwright instead.\n\nRuntime requirements: this tool only works on a Linux daemon machine with Xvfb + ffmpeg (x11grab) + Chromium installed. macOS / Windows daemons will fail at startup.',
+  'Record a silent video of a URL by orchestrating Xvfb + Chromium + ffmpeg, driven by a video plan. Outputs a silent mp4 that can be passed to compose_video_v2 as a video-kind segment with an audio_path for narration.\n\nUse this as the canonical recording step for URL-narration videos. Falls back: if the page needs interactions outside the visual_action vocabulary (clicks, waits, OCR loops), use Monitor (Bash) with custom Playwright instead.\n\nRuntime requirements: this tool only works on a Linux daemon machine with Xvfb + ffmpeg (x11grab) + Chromium installed. macOS / Windows daemons will fail at startup.',
   {
     url: z.string().describe('Page URL to record'),
     plan: z.record(z.any()).describe('Must be the full output from detail_sections (not plan_video). detail_sections output includes detail_sections_version, sections[], audio metadata, and dwell_ms per phase.'),
@@ -1590,126 +1435,6 @@ server.tool('record_url_narration',
   }
 );
 
-// ── compose_video ───────────────────────────────────────────────────────────────
-server.tool('compose_video',
-  'Compose a final short video by muxing audio onto a base video, optionally burning subtitles, concatenating an outro, and transcoding to platform spec.',
-  {
-    video_path: z.string().describe('Base silent video path. Relative paths resolve from the current workspace.'),
-    audio_segments: z.array(z.object({
-      audio_path: z.string().describe('Audio file path for one narration segment.'),
-      start_ms: z.union([z.number(), z.string()]).optional().describe('Segment start offset in milliseconds.'),
-      phase: z.string().optional().describe('Optional phase id. Used with events_log to derive start time.'),
-    })).describe('Ordered or unordered narration audio segments.'),
-    events_log: z.array(z.any()).optional().describe('Optional recorder event log. Used to resolve segment start time by phase.'),
-    subtitles: z.array(z.object({
-      text: z.string().describe('Subtitle text for this segment (the narration sentence).'),
-      start_ms: z.number().describe('Subtitle start time in milliseconds.'),
-      end_ms: z.number().describe('Subtitle end time in milliseconds.'),
-    })).optional().describe('Subtitle segments to burn into the video. Pass each phase sentence text (from detail_sections sections[].sentence) with cumulative start/end time derived from dwell_ms. Omit to produce no subtitles.'),
-    outro_path: z.string().optional().describe('Optional outro mp4 path. If omitted, uses ~/.lightcone/assets/outros/default.mp4 when present.'),
-    target: z.enum(['short_video_cn', 'douyin', 'xhs']).optional().describe('Transcode target profile. Defaults to short_video_cn.'),
-  },
-  async ({ video_path, audio_segments, events_log, subtitles, outro_path, target }) => {
-    if (isBlockedCvmaxEditorVideoTool('compose_video')) {
-      return cvmaxEditorVideoToolError('compose_video');
-    }
-    const composeInput = { video_path, audio_segments, events_log, subtitles, outro_path, target };
-    try {
-      const result = await runMandatoryLocalTool({
-        toolName: 'compose_video',
-        toolInput: composeInput,
-        executor: async (checkedInput) => {
-          const videoPath = normalizeComposePath(checkedInput.video_path, 'video_path');
-          const audioSegments = normalizeComposeAudioSegments(checkedInput.audio_segments);
-          const eventsLog = Array.isArray(checkedInput.events_log) ? checkedInput.events_log : [];
-          const targetProfile = normalizeComposeTarget(checkedInput.target);
-          const requestedOutroPath = String(checkedInput.outro_path ?? '').trim();
-
-          let resolvedOutroPath = null;
-          if (requestedOutroPath) {
-            resolvedOutroPath = path.resolve(WORKSPACE_DIR, requestedOutroPath);
-            if (!existsSync(resolvedOutroPath)) {
-              throw new Error(`outro_path not found: ${resolvedOutroPath}`);
-            }
-          } else if (existsSync(DEFAULT_OUTRO_PATH)) {
-            resolvedOutroPath = DEFAULT_OUTRO_PATH;
-          }
-
-          mkdirSync(VIDEO_COMPOSE_LOCAL_DIR, { recursive: true });
-          const runId = `${Date.now()}-${randomUUID().slice(0, 8)}`;
-          const muxedPath = path.join(VIDEO_COMPOSE_LOCAL_DIR, `compose-${runId}.muxed.mp4`);
-          const concatPath = path.join(VIDEO_COMPOSE_LOCAL_DIR, `compose-${runId}.concat.mp4`);
-          const finalPath = path.join(VIDEO_COMPOSE_LOCAL_DIR, `compose-${runId}.final.mp4`);
-          const assPath = path.join(VIDEO_COMPOSE_LOCAL_DIR, `compose-${runId}.ass`);
-          const intermediates = [muxedPath, concatPath, assPath];
-
-          const subtitleSegments = Array.isArray(checkedInput.subtitles) ? checkedInput.subtitles : [];
-          let subtitlesAssPath = null;
-          if (subtitleSegments.length > 0) {
-            const assContent = buildAssContent(subtitleSegments);
-            writeFileSync(assPath, assContent, 'utf8');
-            subtitlesAssPath = assPath;
-          }
-
-          try {
-            await muxAudioToVideo({
-              video_path: videoPath,
-              audio_segments: audioSegments,
-              events_log: eventsLog,
-              output: muxedPath,
-            });
-
-            let composedPath = muxedPath;
-            if (resolvedOutroPath) {
-              await concatVideos({
-                inputs: [muxedPath, resolvedOutroPath],
-                output: concatPath,
-              });
-              composedPath = concatPath;
-            }
-
-            await transcodeForPlatform({
-              input: composedPath,
-              output: finalPath,
-              target: targetProfile,
-              subtitlesAssPath,
-            });
-
-            const durationMs = await probeDurationMs(finalPath);
-            cleanupLocalFiles(intermediates);
-            return {
-              finalVideoPath: finalPath,
-              durationMs,
-              outroPath: resolvedOutroPath,
-              target: targetProfile,
-              subtitles: subtitleSegments.length > 0,
-            };
-          } catch (error) {
-            cleanupLocalFiles([...intermediates, finalPath]);
-            throw error;
-          }
-        },
-      });
-
-      const outroText = result.outroPath ? result.outroPath : 'skipped';
-      return {
-        content: [{
-          type: 'text',
-          text:
-            `Video composed.\n` +
-            `final_video_path=${result.finalVideoPath}\n` +
-            `duration_ms=${result.durationMs}\n` +
-            `target=${result.target}\n` +
-            `subtitles=${result.subtitles ? 'burned' : 'none'}\n` +
-            `outro=${outroText}`,
-        }],
-      };
-    } catch (error) {
-      return { isError: true, content: [{ type: 'text', text: `Error: ${error.message}` }] };
-    }
-  }
-);
-
 // ── submit_to_library ──────────────────────────────────────────────────────────
 server.tool('submit_to_library',
   '把已生成的视频成片归档进内容库（content_video_draft entry）。调用前 mp4 必须已经通过 write_workspace_file 落到 workspace 的 artifacts/ 路径。归档后内容库会出现一张新卡片，含视频预览 + 元数据 + 后续支持发布/回采链路。',

package/src/tools/render-html-to-image.js

@@ -0,0 +1,58 @@
+import { mkdirSync, writeFileSync } from 'fs';
+import { randomUUID } from 'crypto';
+import path from 'path';
+import os from 'os';
+
+function toolText(text) {
+  return { content: [{ type: 'text', text }] };
+}
+
+function toolError(text) {
+  return { isError: true, content: [{ type: 'text', text }] };
+}
+
+async function launchBrowser() {
+  let playwright;
+  try {
+    playwright = await import('playwright');
+  } catch {
+    throw new Error('playwright_import_failed: run npm install playwright');
+  }
+  return playwright.chromium.launch({ headless: true });
+}
+
+export async function runRenderHtmlToImageTool({ html, output_path, viewport_width, viewport_height, wait_until }) {
+  const normalizedHtml = String(html ?? '').trim();
+  if (!normalizedHtml) return toolError('html is required and must not be empty.');
+
+  const vw = Number(viewport_width ?? 1080);
+  const vh = Number(viewport_height ?? 1920);
+  const waitUntil = String(wait_until ?? 'load');
+
+  const tmpDir = path.join(os.tmpdir(), 'lightcone-html-render');
+  mkdirSync(tmpDir, { recursive: true });
+
+  const htmlFile = path.join(tmpDir, `render-${randomUUID().slice(0, 8)}.html`);
+  writeFileSync(htmlFile, normalizedHtml, 'utf-8');
+
+  const outPath = output_path
+    ? String(output_path)
+    : path.join(tmpDir, `render-${randomUUID().slice(0, 8)}.png`);
+
+  const browser = await launchBrowser();
+  try {
+    const page = await browser.newPage();
+    await page.setViewportSize({ width: vw, height: vh });
+    // Navigate to file:// so local file:// image src attributes load correctly
+    await page.goto(`file://${htmlFile}`, { waitUntil, timeout: 30000 });
+    await page.screenshot({ path: outPath });
+    await page.close();
+    return toolText([
+      'render_html_to_image completed.',
+      `path=${outPath}`,
+      `viewport=${vw}x${vh}`,
+    ].join('\n'));
+  } finally {
+    await browser.close();
+  }
+}