feat(import): optional transcription job

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

packages/feature-import/test/transcribe.test.ts

@@ -0,0 +1,57 @@
+import { afterAll, expect, test } from "bun:test";
+import { readFile } from "node:fs/promises";
+import { join } from "node:path";
+import { importPacktVideo } from "../src/jobs/packt-video";
+import { segmentsToMarkdown, transcribeCourse } from "../src/jobs/transcribe";
+import { makeEnv } from "./harness";
+import { makePacktZip, tmp } from "./fixtures";
+
+let calls = 0;
+let auth: string | null = null;
+const server = Bun.serve({
+  port: 0,
+  async fetch(req) {
+    calls++;
+    auth = req.headers.get("authorization");
+    const form = await req.formData();
+    const f = form.get("file") as File;
+    if (!f || f.size < 100 || form.get("model") !== "whisper-1") return new Response("bad", { status: 400 });
+    return Response.json({ text: "hello world", segments: [{ start: 0, end: 2, text: " hello" }, { start: 35, end: 40, text: " world" }] });
+  },
+});
+afterAll(() => server.stop(true));
+
+test("segments are grouped into timestamped paragraphs", () => {
+  expect(segmentsToMarkdown("T", [{ start: 0, end: 1, text: "a" }, { start: 31, end: 32, text: "b" }, { start: 33, end: 34, text: "c" }], "")).toBe("# T\n\n[0:00] a\n\n[0:31] b c\n");
+});
+
+test("skipped when STC_TRANSCRIBE_URL is unset", async () => {
+  const env = await makeEnv();
+  const r = await transcribeCourse(env.deps, { courseId: "x" }, env.jc);
+  expect(r).toMatchObject({ skipped: true, transcribed: 0 });
+});
+
+test("transcribes video lessons through a Whisper-compatible endpoint", async () => {
+  const env = await makeEnv({ STC_TRANSCRIBE_URL: `http://127.0.0.1:${server.port}/v1/audio/transcriptions`, STC_TRANSCRIBE_KEY: "k" });
+  const imp = await importPacktVideo(env.deps, { path: await makePacktZip(await tmp()), options: {} }, env.jc);
+  const r = await transcribeCourse(env.deps, { courseId: imp.courseId }, env.jc);
+  expect(r.warnings).toEqual([]);
+  expect(r.transcribed).toBe(3);
+  expect(calls).toBe(3);
+  expect(auth).toBe("Bearer k");
+  const c = env.deps.ctx.courses.get(imp.courseId)!;
+  expect(c.status).toBe("ok");
+  expect(c.manifest!.modules[0]!.lessons[0]!.transcript).toBe("transcripts/m1-l1.md");
+  expect(await env.deps.ctx.courses.readLessonMarkdown(imp.courseId, "m1-l1")).toContain("[0:35] world");
+  expect(JSON.parse(await readFile(join(env.dataDir, "cache/transcripts", imp.courseId, "m1-l1.json"), "utf8")).text).toBe("hello world");
+  // a second run has nothing left to do
+  expect((await transcribeCourse(env.deps, { courseId: imp.courseId }, env.jc)).transcribed).toBe(0);
+});
+
+test("an endpoint failure becomes a warning, not a failed job", async () => {
+  const env = await makeEnv({ STC_TRANSCRIBE_URL: `http://127.0.0.1:1/x` });
+  const imp = await importPacktVideo(env.deps, { path: await makePacktZip(await tmp()), options: {} }, env.jc);
+  const r = await transcribeCourse(env.deps, { courseId: imp.courseId }, env.jc);
+  expect(r.transcribed).toBe(0);
+  expect(r.warnings).toHaveLength(3);
+});