@tandryukha/aidemo 0.15.1 → 0.16.0

package/src/mcp/server.ts

@@ -4,7 +4,13 @@ import { z } from "zod";
 import { zodToJsonSchema } from "zod-to-json-schema";
 import { readFile } from "node:fs/promises";
 import { resolve } from "node:path";
-import { ENGINE_ROOT, engineVersion, captionsAutoOffline, chromeProfileDir } from "../config.js";
+import {
+  ENGINE_ROOT,
+  engineVersion,
+  captionsAutoOffline,
+  chromeProfileDir,
+  applyProjectEnv,
+} from "../config.js";
 import { Project, parseStoryboard } from "../project.js";
 import { StoryboardSchema, SeedCookieSchema, type SeedCookie, type Storyboard } from "../types.js";
 import { generateVoice } from "../voice.js";
@@ -498,6 +504,8 @@ export function buildMcpServer(): { server: McpServer; jobs: JobManager } {
       },
     },
     async (args) => {
+      // fromUrl launches Chrome: honour the target repo's .env (#77).
+      if (args.dir && !jobs.running()) await applyProjectEnv(args.dir);
       const demoDir = await scaffoldDemo(
         args.dir ? resolve(args.dir) : process.cwd(),
         args.name,
@@ -523,7 +531,8 @@ export function buildMcpServer(): { server: McpServer; jobs: JobManager } {
     {
       title: "Environment preflight",
       description:
-        "Check prereqs: node, ffmpeg, Chrome, TTS/STT endpoint (flags " +
+        "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.",
       inputSchema: {
         dir: z.string().optional().describe("repo to check for an installed skill"),
@@ -533,6 +542,15 @@ export function buildMcpServer(): { server: McpServer; jobs: JobManager } {
         node: z.string(),
         ffmpeg: z.string().nullable(),
         chrome: z.string().nullable(),
+        browser: z.object({
+          source: z.enum(["default", "channel", "executable"]),
+          channel: z.string().optional(),
+          executablePath: z.string().optional(),
+          description: z.string(),
+          found: z.boolean(),
+          version: z.string().optional(),
+          error: z.string().optional(),
+        }),
         gh: z.string().nullable(),
         endpoint: z.object({
           url: z.string(),
@@ -558,10 +576,13 @@ export function buildMcpServer(): { server: McpServer; jobs: JobManager } {
       },
       annotations: { readOnlyHint: true },
     },
-    async (args) =>
-      jsonResult(
+    async (args) => {
+      // Report the env the pipeline would actually run with for this repo (#77).
+      if (args.dir && !jobs.running()) await applyProjectEnv(args.dir);
+      return jsonResult(
         (await doctorReport(args.dir)) as unknown as Record<string, unknown>
-      )
+      );
+    }
   );
 
   server.registerTool(
@@ -788,6 +809,9 @@ export function buildMcpServer(): { server: McpServer; jobs: JobManager } {
       (async (args: z.infer<z.ZodObject<Shape>>) => {
         const project = new Project((args as { dir: string }).dir);
         await project.ensureDirs();
+        // The demo's own .env, as the CLI run from its repo would see it
+        // (#77). Never swapped under a running job — startJob rejects anyway.
+        if (!jobs.running()) await applyProjectEnv(project.dir);
         try {
           const job = jobs.startJob(kind, project, makeRunner(project, args));
           return jsonResult({
@@ -1131,6 +1155,7 @@ export function buildMcpServer(): { server: McpServer; jobs: JobManager } {
         .boolean()
         .optional()
         .describe("approximate captions from the script — no network/STT"),
+      alignScript: z.boolean().optional().describe("use script spelling and punctuation with STT word timing"),
       params: PARAMS_INPUT,
       lang: LANG_INPUT,
     },
@@ -1142,7 +1167,7 @@ export function buildMcpServer(): { server: McpServer; jobs: JobManager } {
         if (args.offline) {
           await generateCaptionsOffline(lp, sb, sceneProgress(job));
         } else {
-          await generateCaptions(lp, sb, sceneProgress(job));
+          await generateCaptions(lp, sb, { ...sceneProgress(job), alignScript: args.alignScript });
         }
         return {
           srt: lp.captionsSrtPath,
@@ -1375,6 +1400,7 @@ export function buildMcpServer(): { server: McpServer; jobs: JobManager } {
         });
         const file = project.p("logs", `inspect-${stamp}.json`);
         await writeJson(file, res);
+        if (res.target.status === "blocked") log(`inspect blocked: ${res.target.reason}`);
         log(`inspect: ${res.elements.length} element(s), ${res.headings.length} heading(s) → ${file}`);
         return { ...res, file };
       })

package/src/qa.ts

@@ -14,10 +14,12 @@
 
 import { spawn } from "node:child_process";
 import { relative } from "node:path";
+import { stat } from "node:fs/promises";
 import type { Project } from "./project.js";
 import {
   probeDurationMs,
   probeFlatLeadMs,
+  probeVideoDims,
 } from "./ffmpeg.js";
 import { ComposeReportSchema, type Storyboard } from "./types.js";
 import { exists, log, ok, readJson } from "./util.js";
@@ -105,6 +107,24 @@ function imageYavg(file: string): Promise<number | null> {
   });
 }
 
+/**
+ * `blankOpenMs` from the compose report that belongs to `video` (written at or
+ * after it), or null when there is no such report.
+ */
+async function reportedBlankOpenMs(project: Project, video: string): Promise<number | null> {
+  if (!(await exists(project.reportPath))) return null;
+  try {
+    const [rs, vs] = await Promise.all([stat(project.reportPath), stat(video)]);
+    if (rs.mtimeMs < vs.mtimeMs) return null;
+    const report = ComposeReportSchema.safeParse(await readJson(project.reportPath));
+    return report.success && typeof report.data.blankOpenMs === "number"
+      ? report.data.blankOpenMs
+      : null;
+  } catch {
+    return null;
+  }
+}
+
 /** Host of a URL-ish string, or null. */
 function hostOf(url: string): string | null {
   try {
@@ -137,8 +157,17 @@ export async function runQa(project: Project, storyboard: Storyboard): Promise<Q
   }
 
   // 2. The opening frames: a demo that starts on the browser's white pre-paint.
+  // Prefer compose's own measurement (report.json `blankOpenMs`): it is taken
+  // on the page footage before captions are burned in, and the finished MP4
+  // can't tell a caption pill over a blank page from sparse real content
+  // (issue #78). Fall back to probing the MP4 when there's no report for it.
   const introMs = storyboard.intro?.durationMs ?? 0;
-  const blankMs = await probeFlatLeadMs(video, 4000, introMs).catch(() => 0);
+  let blankMs = await reportedBlankOpenMs(project, video);
+  if (blankMs == null) {
+    const dims = await probeVideoDims(video).catch(() => null);
+    const pxScale = dims ? dims.width / storyboard.video.width : 1;
+    blankMs = await probeFlatLeadMs(video, 4000, introMs, pxScale).catch(() => 0);
+  }
   if (blankMs > 250) {
     add(
       "blank-open",

package/src/recorder.ts

@@ -1,4 +1,4 @@
-import { chromium } from "playwright";
+import { describeChromeTarget, launchChromePersistent } from "./browser.js";
 import { promises as fs } from "node:fs";
 import {
   TimelineSchema,
@@ -18,6 +18,8 @@ import {
   captureDevice,
   obsUrl,
   obsPassword,
+  BrowserSetupError,
+  chromeLaunchOptions,
 } from "./config.js";
 import {
   type CaptureMode,
@@ -169,6 +171,11 @@ export async function record(
   // error); clean up a stale lock from a crashed Chrome.
   await ensureProfileUnlocked(profileDir);
   log(`profile: ${profileDir}${options.fresh ? " (fresh)" : ""}`);
+  // Non-default browser (issue #80): resolve it now — a bad
+  // AIDEMO_CHROME_CHANNEL fails here, before any preflight hook runs.
+  if (process.env.AIDEMO_CHROME_EXECUTABLE?.trim() || process.env.AIDEMO_CHROME_CHANNEL?.trim()) {
+    log(`browser: ${describeChromeTarget(chromeLaunchOptions())}`);
+  }
 
   // Preflight hook (issue #44): runs before Chrome launches, with the resolved
   // profile in its env. It may rewrite the storyboard (a rotating fixture), so
@@ -311,8 +318,7 @@ export async function record(
 
   let context;
   try {
-    context = await chromium.launchPersistentContext(profileDir, {
-      channel: "chrome",
+    context = await launchChromePersistent(profileDir, {
       headless: options.headed === false,
       // Hide the automation fingerprint (navigator.webdriver=false) so
       // Cloudflare on chatgpt.com doesn't throw a "Verify you are human"
@@ -345,6 +351,8 @@ export async function record(
           }),
     });
   } catch (err) {
+    // Wrong/missing browser (issue #80) is already actionable on its own.
+    if (err instanceof BrowserSetupError) throw err;
     throw new Error(
       `Failed to launch Chrome with profile ${profileDir}.\n` +
         `If Chrome is already open on this profile, quit it first (Playwright ` +

package/src/safe-regex.ts

@@ -1,3 +1,5 @@
+import { RE2JS } from "re2js";
+
 /**
  * Storyboard-supplied regex patterns (`assert.textMatches`, `assert.url`, and
  * the `textMatches` option the player threads through) are compiled at replay
@@ -29,13 +31,16 @@ export function isUnsafeRegexSource(pattern: string): string | null {
  * Compile a storyboard-supplied pattern, or throw with `where` naming the
  * field so the author can see which matcher to fix.
  */
-export function compileUserRegex(pattern: string, flags: string, where: string): RegExp {
+export function compileUserRegex(pattern: string, flags: string, where: string): RE2JS {
   if (typeof pattern !== "string" || pattern === "")
     throw new Error(`${where}: regex must be a non-empty string`);
   const unsafe = isUnsafeRegexSource(pattern);
   if (unsafe) throw new Error(`${where}: unsafe regex — ${unsafe}`);
   try {
-    return new RegExp(pattern, flags);
+    // RE2JS guarantees linear-time matching for storyboard supplied patterns.
+    // The replay path only uses `test`, so its API is interchangeable here.
+    if (flags && flags !== "i") throw new Error(`unsupported regex flags ${flags}`);
+    return RE2JS.compile(pattern, flags === "i" ? RE2JS.CASE_INSENSITIVE : 0);
   } catch (e) {
     // A raw SyntaxError names neither the field nor the storyboard, which is
     // the only thing the author can act on.

package/src/types.ts

@@ -842,8 +842,10 @@ export const OutputSchema = z
     /**
      * Drop the browser's blank pre-paint at the very start of the video, so
      * the demo opens on the product instead of on ~0.5 s of white under the
-     * first caption (issue #47). At most `blankOpenCapMs` of the leading flat
-     * span is kept. Default false (unchanged opening).
+     * first caption (issue #47). At most `blankOpenCapMs` of the leading blank
+     * span is kept. Blank = empty (one colour, no text/outlines), not merely
+     * static: a motionless terminal or page is never trimmed (issue #78).
+     * Default false (unchanged opening).
      */
     trimLeadingBlank: z.boolean().optional(),
     /** With `trimLeadingBlank`: leading blank kept, ms. Default 100. */