@tandryukha/aidemo 0.17.0 → 0.17.1

package/README.md

@@ -164,7 +164,9 @@ Voice/captions need `OPENAI_API_KEY` in `.env` — **or** `AIDEMO_TTS_PROVIDER=l
 (no key, offline), **or** `OPENAI_BASE_URL` at a local server. See
 [docs/LOCAL_MODELS.md](docs/LOCAL_MODELS.md). No Playwright browser download is
 needed — the engine uses your system Chrome (`channel: "chrome"`; see
-[Setup](#setup) to use a different browser).
+[Setup](#setup) to use a different browser). The one Playwright download is
+its ~2 MB ffmpeg helper, which records the take: `aidemo doctor` checks for it
+and prints the exact install command (`npx playwright install ffmpeg`).
 
 [![Quickstart output — the bundled fixture rendered end-to-end](docs/quickstart.gif)](https://github.com/tandryukha/aidemo/releases/download/v0.3.0/quickstart-demo.mp4)
 
@@ -251,7 +253,9 @@ Codex).
 
 ## Setup
 
-Prereqs: **Node 20+**, **Google Chrome**, **ffmpeg + ffprobe** on `PATH`.
+Prereqs: **Node 20+**, **Google Chrome**, **ffmpeg + ffprobe** on `PATH`, and
+Playwright's ~2 MB ffmpeg helper that records the take (one per Playwright
+version; `aidemo doctor` prints the exact install command).
 Developed and tested on macOS; Linux works for the default (Playwright) capture
 and `--capture obs`. Run `aidemo doctor` to check your setup.
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tandryukha/aidemo",
-  "version": "0.17.0",
+  "version": "0.17.1",
   "license": "MIT",
   "description": "AI Demo Engine — turn a storyboard into a narrated, captioned product demo video",
   "type": "module",

package/src/browser.ts

@@ -8,10 +8,72 @@
  * the options are exactly the engine's historical `{ channel: "chrome", ... }`.
  */
 import { chromium, type Browser, type BrowserContext, type LaunchOptions } from "playwright";
+import { readFileSync } from "node:fs";
+import { createRequire } from "node:module";
+import { homedir } from "node:os";
+import { dirname, isAbsolute, join, resolve } from "node:path";
 import { BrowserSetupError, chromeLaunchOptions, type ChromeLaunchTarget } from "./config.js";
 
 type PersistentOptions = NonNullable<Parameters<typeof chromium.launchPersistentContext>[1]>;
 
+/** Executable name of Playwright's ffmpeg helper, per platform. */
+const FFMPEG_HELPER_EXE: Partial<Record<NodeJS.Platform, string>> = {
+  linux: "ffmpeg-linux",
+  darwin: "ffmpeg-mac",
+  win32: "ffmpeg-win64.exe",
+};
+
+/**
+ * Playwright's own ffmpeg helper: the encoder recordVideo writes every take
+ * with. It is neither the system ffmpeg (compose) nor a browser — a ~2 MB
+ * download per Playwright ffmpeg *revision*, and no npm install fetches it
+ * (playwright has no install script). So a fresh npx install, or a Playwright
+ * minor that bumps the revision (1.64: 1011 → 1013, with VP9 encoding), leaves
+ * it missing and the take fails at launch. Mirrors playwright-core's
+ * documented lookup (PLAYWRIGHT_BROWSERS_PATH, else the per-OS
+ * ms-playwright cache dir). `install` is the exact command for THIS engine's
+ * Playwright. null when the location can't be determined.
+ */
+export function recorderFfmpeg(): { path: string; revision: string; install: string } | null {
+  try {
+    const exe = FFMPEG_HELPER_EXE[process.platform];
+    if (!exe) return null;
+    const req = createRequire(import.meta.url);
+    const playwrightPkg = req.resolve("playwright/package.json");
+    // playwright-core is playwright's dependency: resolve it from there.
+    const corePkg = createRequire(playwrightPkg).resolve("playwright-core/package.json");
+    const coreDir = dirname(corePkg);
+    const browsers = JSON.parse(readFileSync(join(coreDir, "browsers.json"), "utf8")) as {
+      browsers: Array<{ name: string; revision: string }>;
+    };
+    const revision = browsers.browsers.find((b) => b.name === "ffmpeg")?.revision;
+    if (!revision) return null;
+    const env = process.env.PLAYWRIGHT_BROWSERS_PATH;
+    let registry =
+      env === "0"
+        ? join(coreDir, ".local-browsers")
+        : env || join(defaultPlaywrightCacheDir(), "ms-playwright");
+    if (!isAbsolute(registry)) registry = resolve(process.env.INIT_CWD || process.cwd(), registry);
+    const cli = join(dirname(playwrightPkg), "cli.js");
+    return {
+      path: join(registry, `ffmpeg-${revision}`, exe),
+      revision,
+      install: `node "${cli}" install ffmpeg`,
+    };
+  } catch {
+    return null;
+  }
+}
+
+function defaultPlaywrightCacheDir(): string {
+  if (process.platform === "darwin") return join(homedir(), "Library", "Caches");
+  if (process.platform === "win32") return process.env.LOCALAPPDATA || join(homedir(), "AppData", "Local");
+  return process.env.XDG_CACHE_HOME || join(homedir(), ".cache");
+}
+
+/** A launch that failed on Playwright's missing ffmpeg helper, not on the browser. */
+const FFMPEG_HELPER_RE = /playwright install ffmpeg|[\\/]ffmpeg-\d+[\\/]ffmpeg/i;
+
 /** Human-readable description of the configured browser, for logs/doctor. */
 export function describeChromeTarget(target: ChromeLaunchTarget = chromeLaunchOptions()): string {
   if ("executablePath" in target) return `executable ${target.executablePath} (AIDEMO_CHROME_EXECUTABLE)`;
@@ -31,6 +93,20 @@ const MISSING_RE =
 export function explainLaunchError(err: unknown, target: ChromeLaunchTarget): Error {
   const e = err instanceof Error ? err : new Error(String(err));
   if (!MISSING_RE.test(e.message)) return e;
+  if (FFMPEG_HELPER_RE.test(e.message)) {
+    // The browser is fine — Playwright's recording helper is what's missing.
+    // Blaming Chrome here sent users to reinstall a browser that works.
+    const helper = recorderFfmpeg();
+    const out = new BrowserSetupError(
+      `Playwright's ffmpeg helper${helper ? ` (revision ${helper.revision})` : ""} is missing — ` +
+        `the ~2 MB encoder Playwright records the take with (not the system ffmpeg, not a ` +
+        `browser; no npm install fetches it). Install it once per Playwright version: ` +
+        `${helper?.install ?? "npx playwright install ffmpeg"}` +
+        `\nOriginal error: ${e.message}`
+    );
+    (out as Error & { cause?: unknown }).cause = e;
+    return out;
+  }
   let hint: string;
   if ("executablePath" in target) {
     hint =

package/src/distribute.ts

@@ -25,7 +25,7 @@ import {
   chromeLaunchOptions,
   type ChromeLaunchTarget,
 } from "./config.js";
-import { describeChromeTarget, probeConfiguredBrowser } from "./browser.js";
+import { describeChromeTarget, probeConfiguredBrowser, recorderFfmpeg } from "./browser.js";
 import { localTtsStatus } from "./voice-local.js";
 import { Project } from "./project.js";
 import { STARTER_BRIEF, STARTER_STORYBOARD, briefFromInspect, storyboardFromInspect } from "./starter.js";
@@ -659,6 +659,12 @@ export interface DoctorReport {
     };
   };
   playwright: boolean;
+  /**
+   * Playwright's ffmpeg helper for THIS engine's Playwright — the encoder the
+   * built-in capture records every take with (see recorderFfmpeg). null when
+   * its location can't be determined.
+   */
+  recorderFfmpeg: { found: boolean; path: string; revision: string; install: string } | null;
   skill: {
     target: string;
     installed: { version: string; stableAvailable: string | null } | null;
@@ -694,6 +700,7 @@ export async function doctorReport(dir: string | undefined): Promise<DoctorRepor
   const localTts = ttsProvider() === "local" ? await localTtsStatus() : null;
   const chrome = await probeChrome();
   const browser = await probeBrowserTarget(chrome);
+  const helper = recorderFfmpeg();
   const report: DoctorReport = {
     engineVersion: latest,
     node: process.version,
@@ -725,6 +732,7 @@ export async function doctorReport(dir: string | undefined): Promise<DoctorRepor
         : {}),
     },
     playwright: playwrightOk,
+    recorderFfmpeg: helper ? { found: await exists(helper.path), ...helper } : null,
     skill: {
       target,
       installed: manifest
@@ -741,6 +749,7 @@ export async function doctorReport(dir: string | undefined): Promise<DoctorRepor
     report.ffmpeg &&
       report.browser.found &&
       report.playwright &&
+      (report.recorderFfmpeg?.found ?? true) &&
       // Local TTS needs no key: voice is in-process, captions auto-fall-back
       // to the offline derivation in `render`.
       (report.apiKey !== "missing" || report.tts.provider === "local") &&
@@ -806,6 +815,15 @@ export async function doctor(dir: string | undefined): Promise<void> {
   );
   if (r.playwright) ok(`playwright: resolvable`);
   else fail(`playwright: not installed`);
+  if (r.recorderFfmpeg?.found) {
+    ok(`playwright ffmpeg helper: revision ${r.recorderFfmpeg.revision} (records the take)`);
+  } else if (r.recorderFfmpeg) {
+    fail(
+      `playwright ffmpeg helper: revision ${r.recorderFfmpeg.revision} NOT FOUND at ` +
+        `${r.recorderFfmpeg.path} — every take fails without it (~2 MB, once per ` +
+        `Playwright version): ${r.recorderFfmpeg.install}`
+    );
+  }
 
   if (r.skill.installed) {
     ok(

package/src/mcp/server.ts

@@ -556,7 +556,8 @@ export function buildMcpServer(): { server: McpServer; jobs: JobManager } {
       description:
         "Check prereqs: node, ffmpeg, Chrome (or the AIDEMO_CHROME_CHANNEL / " +
         "AIDEMO_CHROME_EXECUTABLE browser in use), TTS/STT endpoint (flags " +
-        "LLM-only servers like Ollama), API key, playwright, installed skill.",
+        "LLM-only servers like Ollama), API key, playwright + its ffmpeg " +
+        "helper (records the take; missing = every take fails), installed skill.",
       inputSchema: {
         dir: z.string().optional().describe("repo to check for an installed skill"),
       },
@@ -586,6 +587,15 @@ export function buildMcpServer(): { server: McpServer; jobs: JobManager } {
           elevenLabsKey: z.enum(["set", "missing"]).optional(),
         }),
         playwright: z.boolean(),
+        recorderFfmpeg: z
+          .object({
+            found: z.boolean(),
+            path: z.string(),
+            revision: z.string(),
+            install: z.string().describe("exact command to install it"),
+          })
+          .nullable()
+          .describe("Playwright's ffmpeg helper the built-in capture records with"),
         skill: z.object({
           target: z.string(),
           installed: z