videoclaw 3.0.0-alpha.1

package/dist/video/audio-platform/native-flowmusic.d.ts

@@ -0,0 +1,21 @@
+/**
+ * FlowMusic music backend — Google Lyria 3 Pro vocal songs via useapi.net.
+ *
+ * Unlike the instrumental-only Lyria backends (native-lyria / native-lyria3),
+ * FlowMusic generates FULL VOCAL SONGS (and instrumentals): it honors
+ * user-supplied `lyrics` and an `instrumental` toggle. It is an async, job-polled
+ * provider, so this backend wraps the providers/flowmusic-useapi.ts transport:
+ *   submit (mode:async) → poll GET /jobs/{jobid} until terminal → download the
+ *   first clip's mp3 via the auth'd /music/download endpoint → write `.mp3`.
+ *
+ * Availability reuses the shared USEAPI_API_TOKEN (the same token as the
+ * dreamina/runway routes) — gated in the registry via `requiresUseApi`, not a
+ * backend-specific requiredEnv. Optional: VCLAW_FLOWMUSIC_ACCOUNT pins the
+ * flowmusic.app account; VCLAW_FLOWMUSIC_GHOSTWRITER ('standard'|'pro') picks the
+ * lyrics-writer when the model writes lyrics. The poll cadence is configurable
+ * via VCLAW_FLOWMUSIC_POLL_INTERVAL_MS / VCLAW_FLOWMUSIC_POLL_MAX_ATTEMPTS
+ * (tests set the interval to 0 to stay fast + offline).
+ */
+import type { MusicBackend } from './types.js';
+export declare const flowMusicBackend: MusicBackend;
+//# sourceMappingURL=native-flowmusic.d.ts.map

package/dist/video/audio-platform/native-flowmusic.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"native-flowmusic.d.ts","sourceRoot":"","sources":["../../../src/video/audio-platform/native-flowmusic.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;GAiBG;AAYH,OAAO,KAAK,EAAE,YAAY,EAAiC,MAAM,YAAY,CAAC;AA4B9E,eAAO,MAAM,gBAAgB,EAAE,YA6G9B,CAAC"}

package/dist/video/audio-platform/native-flowmusic.js

@@ -0,0 +1,137 @@
+/**
+ * FlowMusic music backend — Google Lyria 3 Pro vocal songs via useapi.net.
+ *
+ * Unlike the instrumental-only Lyria backends (native-lyria / native-lyria3),
+ * FlowMusic generates FULL VOCAL SONGS (and instrumentals): it honors
+ * user-supplied `lyrics` and an `instrumental` toggle. It is an async, job-polled
+ * provider, so this backend wraps the providers/flowmusic-useapi.ts transport:
+ *   submit (mode:async) → poll GET /jobs/{jobid} until terminal → download the
+ *   first clip's mp3 via the auth'd /music/download endpoint → write `.mp3`.
+ *
+ * Availability reuses the shared USEAPI_API_TOKEN (the same token as the
+ * dreamina/runway routes) — gated in the registry via `requiresUseApi`, not a
+ * backend-specific requiredEnv. Optional: VCLAW_FLOWMUSIC_ACCOUNT pins the
+ * flowmusic.app account; VCLAW_FLOWMUSIC_GHOSTWRITER ('standard'|'pro') picks the
+ * lyrics-writer when the model writes lyrics. The poll cadence is configurable
+ * via VCLAW_FLOWMUSIC_POLL_INTERVAL_MS / VCLAW_FLOWMUSIC_POLL_MAX_ATTEMPTS
+ * (tests set the interval to 0 to stay fast + offline).
+ */
+import { mkdir, writeFile, rename } from 'node:fs/promises';
+import { dirname, extname } from 'node:path';
+import { VclawError } from '../errors.js';
+import { submitFlowMusicJob, pollFlowMusicJob, downloadFlowMusicClipMp3, } from '../providers/flowmusic-useapi.js';
+/** Lyria 3 Pro full songs land ~40–150s; used only as a fallback duration estimate. */
+const DEFAULT_DURATION_SEC = 150;
+const DEFAULT_POLL_INTERVAL_MS = 5000;
+const DEFAULT_POLL_MAX_ATTEMPTS = 90; // ~7.5 min ceiling at 5s
+/** Force a `.mp3` extension on the output path (we download mp3 bytes). */
+function forceMp3Path(p) {
+    const ext = extname(p);
+    if (ext.toLowerCase() === '.mp3')
+        return p;
+    return ext ? `${p.slice(0, p.length - ext.length)}.mp3` : `${p}.mp3`;
+}
+function sleep(ms) {
+    return new Promise((resolve) => setTimeout(resolve, ms));
+}
+function isGhostwriter(v) {
+    return v === 'standard' || v === 'pro';
+}
+function positiveIntFromEnv(raw, fallback) {
+    if (raw === undefined)
+        return fallback;
+    const n = Number(raw);
+    return Number.isFinite(n) && n >= 0 ? n : fallback;
+}
+export const flowMusicBackend = {
+    id: 'flowmusic',
+    kind: 'music',
+    displayName: 'FlowMusic (Lyria 3 Pro, useapi.net)',
+    // Reuses the shared USEAPI_API_TOKEN; gated in the registry via requiresUseApi
+    // rather than a backend-specific requiredEnv (no new token is introduced).
+    requiredEnv: [],
+    requiresUseApi: true,
+    summary: 'Full vocal songs (and instrumentals) via Google Lyria 3 Pro on useapi.net FlowMusic. ' +
+        'Honors --lyrics and --instrumental. Async submit → poll → mp3 download. Reuses USEAPI_API_TOKEN; ' +
+        'optional VCLAW_FLOWMUSIC_ACCOUNT / VCLAW_FLOWMUSIC_GHOSTWRITER.',
+    async generate(input) {
+        const env = input.env ?? process.env;
+        const outputPath = forceMp3Path(input.outputPath);
+        const fallbackDurationMs = (input.durationSec ?? DEFAULT_DURATION_SEC) * 1000;
+        if (!input.prompt.trim()) {
+            throw new VclawError('music_gen_failed', 'Cannot generate music with an empty prompt.', {
+                backend: 'flowmusic',
+            });
+        }
+        // Dry-run: estimate duration, write a placeholder mp3, touch no network.
+        if (input.dryRun) {
+            await mkdir(dirname(outputPath), { recursive: true });
+            await writeFile(outputPath, Buffer.alloc(0));
+            return { path: outputPath, durationMs: fallbackDurationMs, backendId: 'flowmusic' };
+        }
+        const apiToken = (env['USEAPI_API_TOKEN'] ?? '').trim();
+        if (!apiToken) {
+            throw new VclawError('music_gen_failed', 'FlowMusic requires USEAPI_API_TOKEN.', {
+                backend: 'flowmusic',
+            });
+        }
+        const account = env['VCLAW_FLOWMUSIC_ACCOUNT']?.trim() || undefined;
+        const ghostwriterRaw = env['VCLAW_FLOWMUSIC_GHOSTWRITER']?.trim();
+        const ghostwriter = isGhostwriter(ghostwriterRaw) ? ghostwriterRaw : undefined;
+        const fetchImpl = input.fetcher;
+        const pollIntervalMs = positiveIntFromEnv(env['VCLAW_FLOWMUSIC_POLL_INTERVAL_MS'], DEFAULT_POLL_INTERVAL_MS);
+        const maxAttempts = positiveIntFromEnv(env['VCLAW_FLOWMUSIC_POLL_MAX_ATTEMPTS'], DEFAULT_POLL_MAX_ATTEMPTS);
+        const submit = await submitFlowMusicJob({
+            apiToken,
+            prompt: input.prompt,
+            ...(account ? { account } : {}),
+            ...(input.instrumental !== undefined ? { instrumental: input.instrumental } : {}),
+            ...(input.lyrics ? { lyrics: input.lyrics } : {}),
+            ...(ghostwriter ? { ghostwriter } : {}),
+            ...(fetchImpl ? { fetchImpl } : {}),
+        });
+        // Poll to a terminal state (bounded). The submit may already be terminal in
+        // some responses; the loop body re-polls on 'pending'.
+        let poll = await pollFlowMusicJob({
+            apiToken,
+            jobid: submit.jobid,
+            ...(fetchImpl ? { fetchImpl } : {}),
+        });
+        for (let attempt = 0; poll.status === 'pending' && attempt < maxAttempts; attempt += 1) {
+            if (pollIntervalMs > 0)
+                await sleep(pollIntervalMs);
+            poll = await pollFlowMusicJob({
+                apiToken,
+                jobid: submit.jobid,
+                ...(fetchImpl ? { fetchImpl } : {}),
+            });
+        }
+        if (poll.status === 'failed') {
+            throw new VclawError('music_gen_failed', `FlowMusic generation failed: ${poll.error ?? 'unknown error'}`, { backend: 'flowmusic', jobid: submit.jobid, error: poll.error });
+        }
+        if (poll.status !== 'completed' || poll.clips.length === 0) {
+            throw new VclawError('music_gen_failed', 'FlowMusic job did not complete in time (no clips returned).', { backend: 'flowmusic', jobid: submit.jobid, status: poll.status });
+        }
+        const clip = poll.clips[0]; // A/B pair → MVP takes the first clip.
+        const bytes = await downloadFlowMusicClipMp3({
+            apiToken,
+            clip: clip.clip,
+            ...(fetchImpl ? { fetchImpl } : {}),
+        });
+        if (bytes.length === 0) {
+            throw new VclawError('music_gen_failed', 'FlowMusic returned an empty audio payload.', {
+                backend: 'flowmusic',
+                jobid: submit.jobid,
+            });
+        }
+        await mkdir(dirname(outputPath), { recursive: true });
+        // Atomic write: tmp → rename (mirrors native-dreamina's downloadToFile).
+        const tmp = `${outputPath}.tmp`;
+        await writeFile(tmp, bytes);
+        await rename(tmp, outputPath);
+        // Prefer the clip's real duration; fall back to the request estimate.
+        const durationMs = clip.durationS != null ? Math.round(clip.durationS * 1000) : fallbackDurationMs;
+        return { path: outputPath, durationMs, backendId: 'flowmusic' };
+    },
+};
+//# sourceMappingURL=native-flowmusic.js.map

package/dist/video/audio-platform/native-flowmusic.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"native-flowmusic.js","sourceRoot":"","sources":["../../../src/video/audio-platform/native-flowmusic.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;GAiBG;AAEH,OAAO,EAAE,KAAK,EAAE,SAAS,EAAE,MAAM,EAAE,MAAM,kBAAkB,CAAC;AAC5D,OAAO,EAAE,OAAO,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AAC7C,OAAO,EAAE,UAAU,EAAE,MAAM,cAAc,CAAC;AAC1C,OAAO,EACL,kBAAkB,EAClB,gBAAgB,EAChB,wBAAwB,GAGzB,MAAM,kCAAkC,CAAC;AAG1C,uFAAuF;AACvF,MAAM,oBAAoB,GAAG,GAAG,CAAC;AACjC,MAAM,wBAAwB,GAAG,IAAI,CAAC;AACtC,MAAM,yBAAyB,GAAG,EAAE,CAAC,CAAC,yBAAyB;AAE/D,2EAA2E;AAC3E,SAAS,YAAY,CAAC,CAAS;IAC7B,MAAM,GAAG,GAAG,OAAO,CAAC,CAAC,CAAC,CAAC;IACvB,IAAI,GAAG,CAAC,WAAW,EAAE,KAAK,MAAM;QAAE,OAAO,CAAC,CAAC;IAC3C,OAAO,GAAG,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC,MAAM,GAAG,GAAG,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC;AACvE,CAAC;AAED,SAAS,KAAK,CAAC,EAAU;IACvB,OAAO,IAAI,OAAO,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,UAAU,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC,CAAC;AAC3D,CAAC;AAED,SAAS,aAAa,CAAC,CAAqB;IAC1C,OAAO,CAAC,KAAK,UAAU,IAAI,CAAC,KAAK,KAAK,CAAC;AACzC,CAAC;AAED,SAAS,kBAAkB,CAAC,GAAuB,EAAE,QAAgB;IACnE,IAAI,GAAG,KAAK,SAAS;QAAE,OAAO,QAAQ,CAAC;IACvC,MAAM,CAAC,GAAG,MAAM,CAAC,GAAG,CAAC,CAAC;IACtB,OAAO,MAAM,CAAC,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC;AACrD,CAAC;AAED,MAAM,CAAC,MAAM,gBAAgB,GAAiB;IAC5C,EAAE,EAAE,WAAW;IACf,IAAI,EAAE,OAAO;IACb,WAAW,EAAE,qCAAqC;IAClD,+EAA+E;IAC/E,2EAA2E;IAC3E,WAAW,EAAE,EAAE;IACf,cAAc,EAAE,IAAI;IACpB,OAAO,EACL,uFAAuF;QACvF,mGAAmG;QACnG,iEAAiE;IAEnE,KAAK,CAAC,QAAQ,CAAC,KAAoB;QACjC,MAAM,GAAG,GAAG,KAAK,CAAC,GAAG,IAAI,OAAO,CAAC,GAAG,CAAC;QACrC,MAAM,UAAU,GAAG,YAAY,CAAC,KAAK,CAAC,UAAU,CAAC,CAAC;QAClD,MAAM,kBAAkB,GAAG,CAAC,KAAK,CAAC,WAAW,IAAI,oBAAoB,CAAC,GAAG,IAAI,CAAC;QAE9E,IAAI,CAAC,KAAK,CAAC,MAAM,CAAC,IAAI,EAAE,EAAE,CAAC;YACzB,MAAM,IAAI,UAAU,CAAC,kBAAkB,EAAE,6CAA6C,EAAE;gBACtF,OAAO,EAAE,WAAW;aACrB,CAAC,CAAC;QACL,CAAC;QAED,yEAAyE;QACzE,IAAI,KAAK,CAAC,MAAM,EAAE,CAAC;YACjB,MAAM,KAAK,CAAC,OAAO,CAAC,UAAU,CAAC,EAAE,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC,CAAC;YACtD,MAAM,SAAS,CAAC,UAAU,EAAE,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC;YAC7C,OAAO,EAAE,IAAI,EAAE,UAAU,EAAE,UAAU,EAAE,kBAAkB,EAAE,SAAS,EAAE,WAAW,EAAE,CAAC;QACtF,CAAC;QAED,MAAM,QAAQ,GAAG,CAAC,GAAG,CAAC,kBAAkB,CAAC,IAAI,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC;QACxD,IAAI,CAAC,QAAQ,EAAE,CAAC;YACd,MAAM,IAAI,UAAU,CAAC,kBAAkB,EAAE,sCAAsC,EAAE;gBAC/E,OAAO,EAAE,WAAW;aACrB,CAAC,CAAC;QACL,CAAC;QACD,MAAM,OAAO,GAAG,GAAG,CAAC,yBAAyB,CAAC,EAAE,IAAI,EAAE,IAAI,SAAS,CAAC;QACpE,MAAM,cAAc,GAAG,GAAG,CAAC,6BAA6B,CAAC,EAAE,IAAI,EAAE,CAAC;QAClE,MAAM,WAAW,GAAG,aAAa,CAAC,cAAc,CAAC,CAAC,CAAC,CAAC,cAAc,CAAC,CAAC,CAAC,SAAS,CAAC;QAC/E,MAAM,SAAS,GAAG,KAAK,CAAC,OAAoD,CAAC;QAE7E,MAAM,cAAc,GAAG,kBAAkB,CAAC,GAAG,CAAC,kCAAkC,CAAC,EAAE,wBAAwB,CAAC,CAAC;QAC7G,MAAM,WAAW,GAAG,kBAAkB,CAAC,GAAG,CAAC,mCAAmC,CAAC,EAAE,yBAAyB,CAAC,CAAC;QAE5G,MAAM,MAAM,GAAG,MAAM,kBAAkB,CAAC;YACtC,QAAQ;YACR,MAAM,EAAE,KAAK,CAAC,MAAM;YACpB,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;YAC/B,GAAG,CAAC,KAAK,CAAC,YAAY,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,YAAY,EAAE,KAAK,CAAC,YAAY,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;YACjF,GAAG,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,MAAM,EAAE,KAAK,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;YACjD,GAAG,CAAC,WAAW,CAAC,CAAC,CAAC,EAAE,WAAW,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;YACvC,GAAG,CAAC,SAAS,CAAC,CAAC,CAAC,EAAE,SAAS,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;SACpC,CAAC,CAAC;QAEH,4EAA4E;QAC5E,uDAAuD;QACvD,IAAI,IAAI,GAAG,MAAM,gBAAgB,CAAC;YAChC,QAAQ;YACR,KAAK,EAAE,MAAM,CAAC,KAAK;YACnB,GAAG,CAAC,SAAS,CAAC,CAAC,CAAC,EAAE,SAAS,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;SACpC,CAAC,CAAC;QACH,KAAK,IAAI,OAAO,GAAG,CAAC,EAAE,IAAI,CAAC,MAAM,KAAK,SAAS,IAAI,OAAO,GAAG,WAAW,EAAE,OAAO,IAAI,CAAC,EAAE,CAAC;YACvF,IAAI,cAAc,GAAG,CAAC;gBAAE,MAAM,KAAK,CAAC,cAAc,CAAC,CAAC;YACpD,IAAI,GAAG,MAAM,gBAAgB,CAAC;gBAC5B,QAAQ;gBACR,KAAK,EAAE,MAAM,CAAC,KAAK;gBACnB,GAAG,CAAC,SAAS,CAAC,CAAC,CAAC,EAAE,SAAS,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;aACpC,CAAC,CAAC;QACL,CAAC;QAED,IAAI,IAAI,CAAC,MAAM,KAAK,QAAQ,EAAE,CAAC;YAC7B,MAAM,IAAI,UAAU,CAClB,kBAAkB,EAClB,gCAAgC,IAAI,CAAC,KAAK,IAAI,eAAe,EAAE,EAC/D,EAAE,OAAO,EAAE,WAAW,EAAE,KAAK,EAAE,MAAM,CAAC,KAAK,EAAE,KAAK,EAAE,IAAI,CAAC,KAAK,EAAE,CACjE,CAAC;QACJ,CAAC;QACD,IAAI,IAAI,CAAC,MAAM,KAAK,WAAW,IAAI,IAAI,CAAC,KAAK,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YAC3D,MAAM,IAAI,UAAU,CAClB,kBAAkB,EAClB,6DAA6D,EAC7D,EAAE,OAAO,EAAE,WAAW,EAAE,KAAK,EAAE,MAAM,CAAC,KAAK,EAAE,MAAM,EAAE,IAAI,CAAC,MAAM,EAAE,CACnE,CAAC;QACJ,CAAC;QAED,MAAM,IAAI,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,uCAAuC;QACnE,MAAM,KAAK,GAAG,MAAM,wBAAwB,CAAC;YAC3C,QAAQ;YACR,IAAI,EAAE,IAAI,CAAC,IAAI;YACf,GAAG,CAAC,SAAS,CAAC,CAAC,CAAC,EAAE,SAAS,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;SACpC,CAAC,CAAC;QACH,IAAI,KAAK,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YACvB,MAAM,IAAI,UAAU,CAAC,kBAAkB,EAAE,4CAA4C,EAAE;gBACrF,OAAO,EAAE,WAAW;gBACpB,KAAK,EAAE,MAAM,CAAC,KAAK;aACpB,CAAC,CAAC;QACL,CAAC;QAED,MAAM,KAAK,CAAC,OAAO,CAAC,UAAU,CAAC,EAAE,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC,CAAC;QACtD,yEAAyE;QACzE,MAAM,GAAG,GAAG,GAAG,UAAU,MAAM,CAAC;QAChC,MAAM,SAAS,CAAC,GAAG,EAAE,KAAK,CAAC,CAAC;QAC5B,MAAM,MAAM,CAAC,GAAG,EAAE,UAAU,CAAC,CAAC;QAE9B,sEAAsE;QACtE,MAAM,UAAU,GAAG,IAAI,CAAC,SAAS,IAAI,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC,CAAC,CAAC,CAAC,kBAAkB,CAAC;QACnG,OAAO,EAAE,IAAI,EAAE,UAAU,EAAE,UAAU,EAAE,SAAS,EAAE,WAAW,EAAE,CAAC;IAClE,CAAC;CACF,CAAC"}

package/dist/video/audio-platform/native-gemini-tts.d.ts

@@ -0,0 +1,26 @@
+/**
+ * SCAFFOLD: Gemini TTS request/response per docs; verify against a real key.
+ *
+ * Gemini TTS backend — wraps the generativelanguage.googleapis.com
+ * `gemini-2.5-flash-preview-tts:generateContent` endpoint. This is an
+ * API-KEY product (NOT Vertex), so it resolves a key from the existing
+ * Gemini key pool (GEMINI_API_KEYS / GOOGLE_API_KEYS / GOOGLE_API_KEY) via
+ * `fetchGeminiWithPool`. When a caller passes `input.env` carrying a key the
+ * process.env-backed pool cannot see, that key is forwarded as the pool's
+ * `keyOverride` so library callers authenticate without mutating process.env.
+ *
+ * The model returns RAW PCM (signed 16-bit little-endian, 24kHz, mono) as
+ * base64 in candidates[0].content.parts[].inlineData.data. We wrap that PCM
+ * in a 44-byte WAV header and write the result to `outputPath` (a .wav).
+ *
+ * Duration is computed deterministically from the PCM byte count — no ffprobe:
+ *   durationMs = round(pcmBytes / (24000 * 2) * 1000)
+ *
+ * Availability: a Gemini key is resolvable (pool size > 0 or one of the env
+ * vars is set). The backend is inert (unavailable) without a key and never
+ * throws on the default path; on a missing/empty audio response it throws
+ * VclawError('tts_failed', ...).
+ */
+import type { TtsBackend } from './types.js';
+export declare const geminiTts: TtsBackend;
+//# sourceMappingURL=native-gemini-tts.d.ts.map

package/dist/video/audio-platform/native-gemini-tts.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"native-gemini-tts.d.ts","sourceRoot":"","sources":["../../../src/video/audio-platform/native-gemini-tts.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;GAsBG;AAMH,OAAO,KAAK,EAAE,UAAU,EAA6B,MAAM,YAAY,CAAC;AAqFxE,eAAO,MAAM,SAAS,EAAE,UAgHvB,CAAC"}

package/dist/video/audio-platform/native-gemini-tts.js

@@ -0,0 +1,202 @@
+/**
+ * SCAFFOLD: Gemini TTS request/response per docs; verify against a real key.
+ *
+ * Gemini TTS backend — wraps the generativelanguage.googleapis.com
+ * `gemini-2.5-flash-preview-tts:generateContent` endpoint. This is an
+ * API-KEY product (NOT Vertex), so it resolves a key from the existing
+ * Gemini key pool (GEMINI_API_KEYS / GOOGLE_API_KEYS / GOOGLE_API_KEY) via
+ * `fetchGeminiWithPool`. When a caller passes `input.env` carrying a key the
+ * process.env-backed pool cannot see, that key is forwarded as the pool's
+ * `keyOverride` so library callers authenticate without mutating process.env.
+ *
+ * The model returns RAW PCM (signed 16-bit little-endian, 24kHz, mono) as
+ * base64 in candidates[0].content.parts[].inlineData.data. We wrap that PCM
+ * in a 44-byte WAV header and write the result to `outputPath` (a .wav).
+ *
+ * Duration is computed deterministically from the PCM byte count — no ffprobe:
+ *   durationMs = round(pcmBytes / (24000 * 2) * 1000)
+ *
+ * Availability: a Gemini key is resolvable (pool size > 0 or one of the env
+ * vars is set). The backend is inert (unavailable) without a key and never
+ * throws on the default path; on a missing/empty audio response it throws
+ * VclawError('tts_failed', ...).
+ */
+import { mkdir, writeFile } from 'node:fs/promises';
+import { dirname } from 'node:path';
+import { fetchGeminiWithPool } from '../gemini-key-pool.js';
+import { VclawError } from '../errors.js';
+const TTS_MODEL = 'gemini-2.5-flash-preview-tts';
+const TTS_SAMPLE_RATE_HZ = 24_000;
+const TTS_BYTES_PER_SAMPLE = 2; // s16le
+const TTS_CHANNELS = 1; // mono
+const DEFAULT_VOICE = 'Kore';
+/** Chars-per-second heuristic for dry-run duration estimation (~14 cps speech). */
+const DRYRUN_CHARS_PER_SEC = 14;
+/**
+ * Gemini API key sources, in the same precedence as the global key pool
+ * (gemini-key-pool.ts). Used to resolve an explicit key from a caller-supplied
+ * `input.env` so the pool (which only reads process.env) can be bypassed.
+ */
+const GEMINI_KEY_VARS = ['GEMINI_API_KEYS', 'GOOGLE_API_KEYS', 'GOOGLE_API_KEY'];
+/**
+ * Resolve the first key from a caller-supplied env that differs from
+ * process.env. Returns undefined when no env was passed, or when the passed env
+ * adds no key beyond what process.env already exposes (in which case the global
+ * pool handles key selection). Splits on the same delimiters the pool uses and
+ * returns the first token.
+ */
+function resolveKeyOverrideFromEnv(env) {
+    if (!env || env === process.env)
+        return undefined;
+    for (const varName of GEMINI_KEY_VARS) {
+        const raw = env[varName];
+        if (typeof raw !== 'string' || raw.trim() === '')
+            continue;
+        // process.env already carries this exact value → let the pool handle it.
+        if (process.env[varName] === raw)
+            continue;
+        const first = raw.split(/[,;\n\s]+/).map((t) => t.trim()).find((t) => t.length > 0);
+        if (first)
+            return first;
+    }
+    return undefined;
+}
+function buildTtsUrl(key) {
+    return `https://generativelanguage.googleapis.com/v1beta/models/${TTS_MODEL}:generateContent?key=${encodeURIComponent(key)}`;
+}
+/**
+ * durationMs from raw PCM byte count (s16le 24kHz mono).
+ * pcmBytes / (sampleRate * bytesPerSample * channels) * 1000
+ */
+function pcmDurationMs(pcmBytes) {
+    return Math.round((pcmBytes / (TTS_SAMPLE_RATE_HZ * TTS_BYTES_PER_SAMPLE * TTS_CHANNELS)) * 1000);
+}
+/** Wrap raw PCM (s16le 24kHz mono) bytes in a canonical 44-byte WAV header. */
+function wrapPcmInWav(pcm) {
+    const byteRate = TTS_SAMPLE_RATE_HZ * TTS_CHANNELS * TTS_BYTES_PER_SAMPLE;
+    const blockAlign = TTS_CHANNELS * TTS_BYTES_PER_SAMPLE;
+    const header = Buffer.alloc(44);
+    header.write('RIFF', 0, 'ascii');
+    header.writeUInt32LE(36 + pcm.length, 4); // ChunkSize
+    header.write('WAVE', 8, 'ascii');
+    header.write('fmt ', 12, 'ascii');
+    header.writeUInt32LE(16, 16); // Subchunk1Size (PCM)
+    header.writeUInt16LE(1, 20); // AudioFormat = PCM
+    header.writeUInt16LE(TTS_CHANNELS, 22);
+    header.writeUInt32LE(TTS_SAMPLE_RATE_HZ, 24);
+    header.writeUInt32LE(byteRate, 28);
+    header.writeUInt16LE(blockAlign, 32);
+    header.writeUInt16LE(TTS_BYTES_PER_SAMPLE * 8, 34); // BitsPerSample
+    header.write('data', 36, 'ascii');
+    header.writeUInt32LE(pcm.length, 40); // Subchunk2Size
+    return Buffer.concat([header, pcm]);
+}
+/** Defensive extraction of the first base64 inlineData audio payload. */
+function extractAudioBase64(parsed) {
+    if (typeof parsed !== 'object' || parsed === null)
+        return undefined;
+    const candidates = parsed.candidates;
+    if (!Array.isArray(candidates) || candidates.length === 0)
+        return undefined;
+    const parts = candidates[0]?.content?.parts;
+    if (!Array.isArray(parts))
+        return undefined;
+    for (const part of parts) {
+        const data = part?.inlineData?.data;
+        if (typeof data === 'string' && data.length > 0)
+            return data;
+    }
+    return undefined;
+}
+export const geminiTts = {
+    id: 'gemini-tts',
+    kind: 'tts',
+    outputExtension: 'wav',
+    displayName: 'Gemini TTS (gemini-2.5-flash-preview-tts)',
+    // API-key product, not Vertex. Availability is resolved from the Gemini key
+    // pool (see isTtsBackendAvailable), so requiredEnv stays empty and the
+    // registry gates on requiresGeminiKey instead of this backend's id.
+    requiredEnv: [],
+    requiresVertex: false,
+    requiresGeminiKey: true,
+    summary: 'Single-speaker text-to-speech via the Gemini API (gemini-2.5-flash-preview-tts). Returns 24kHz mono PCM wrapped as WAV. Requires a Gemini API key (GEMINI_API_KEYS / GOOGLE_API_KEYS / GOOGLE_API_KEY).',
+    async generate(input) {
+        const voice = input.voice && input.voice.trim() ? input.voice.trim() : DEFAULT_VOICE;
+        if (!input.text || !input.text.trim()) {
+            throw new VclawError('tts_failed', 'Cannot synthesize narration from empty text.', {
+                backendId: 'gemini-tts',
+            });
+        }
+        // Dry-run: estimate duration from text length; do NOT touch network or do
+        // real synthesis. Write a placeholder WAV (44-byte header, no samples) so
+        // downstream existsSync() checks stay consistent.
+        if (input.dryRun) {
+            const estSec = Math.max(1, input.text.trim().length / DRYRUN_CHARS_PER_SEC);
+            const durationMs = Math.round(estSec * 1000);
+            await mkdir(dirname(input.outputPath), { recursive: true });
+            await writeFile(input.outputPath, wrapPcmInWav(Buffer.alloc(0)));
+            return { path: input.outputPath, durationMs, backendId: 'gemini-tts' };
+        }
+        const requestBody = JSON.stringify({
+            contents: [{ parts: [{ text: input.text }] }],
+            generationConfig: {
+                responseModalities: ['AUDIO'],
+                speechConfig: {
+                    voiceConfig: {
+                        prebuiltVoiceConfig: { voiceName: voice },
+                    },
+                },
+            },
+        });
+        // Resolve a key from input.env when it carries one the global pool (which
+        // only reads process.env) cannot see, so a library caller passing
+        // env: { GEMINI_API_KEYS: '...' } actually authenticates instead of hitting
+        // 'No Gemini API keys configured'. Absent that, the pool handles selection.
+        const keyOverride = resolveKeyOverrideFromEnv(input.env);
+        const poolOptions = {};
+        if (input.fetcher)
+            poolOptions.fetcher = input.fetcher;
+        if (keyOverride)
+            poolOptions.keyOverride = keyOverride;
+        const response = await fetchGeminiWithPool(buildTtsUrl, {
+            method: 'POST',
+            headers: { 'Content-Type': 'application/json' },
+            body: requestBody,
+        }, poolOptions);
+        if (!response.ok) {
+            const body = await response.text().catch(() => '');
+            throw new VclawError('tts_failed', `Gemini TTS request failed with HTTP ${response.status}.`, { backendId: 'gemini-tts', status: response.status, body: body.slice(0, 500) });
+        }
+        let parsed;
+        try {
+            parsed = await response.json();
+        }
+        catch (err) {
+            throw new VclawError('tts_failed', 'Gemini TTS returned a non-JSON response.', {
+                backendId: 'gemini-tts',
+                error: err instanceof Error ? err.message : String(err),
+            });
+        }
+        const b64 = extractAudioBase64(parsed);
+        if (!b64) {
+            throw new VclawError('tts_failed', 'Gemini TTS response contained no audio data.', {
+                backendId: 'gemini-tts',
+            });
+        }
+        const pcm = Buffer.from(b64, 'base64');
+        if (pcm.length === 0) {
+            throw new VclawError('tts_failed', 'Gemini TTS returned an empty audio payload.', {
+                backendId: 'gemini-tts',
+            });
+        }
+        const wav = wrapPcmInWav(pcm);
+        await mkdir(dirname(input.outputPath), { recursive: true });
+        await writeFile(input.outputPath, wav);
+        return {
+            path: input.outputPath,
+            durationMs: pcmDurationMs(pcm.length),
+            backendId: 'gemini-tts',
+        };
+    },
+};
+//# sourceMappingURL=native-gemini-tts.js.map

package/dist/video/audio-platform/native-gemini-tts.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"native-gemini-tts.js","sourceRoot":"","sources":["../../../src/video/audio-platform/native-gemini-tts.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;GAsBG;AAEH,OAAO,EAAE,KAAK,EAAE,SAAS,EAAE,MAAM,kBAAkB,CAAC;AACpD,OAAO,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AACpC,OAAO,EAAE,mBAAmB,EAAE,MAAM,uBAAuB,CAAC;AAC5D,OAAO,EAAE,UAAU,EAAE,MAAM,cAAc,CAAC;AAG1C,MAAM,SAAS,GAAG,8BAA8B,CAAC;AACjD,MAAM,kBAAkB,GAAG,MAAM,CAAC;AAClC,MAAM,oBAAoB,GAAG,CAAC,CAAC,CAAC,QAAQ;AACxC,MAAM,YAAY,GAAG,CAAC,CAAC,CAAC,OAAO;AAC/B,MAAM,aAAa,GAAG,MAAM,CAAC;AAE7B,mFAAmF;AACnF,MAAM,oBAAoB,GAAG,EAAE,CAAC;AAEhC;;;;GAIG;AACH,MAAM,eAAe,GAAG,CAAC,iBAAiB,EAAE,iBAAiB,EAAE,gBAAgB,CAAU,CAAC;AAE1F;;;;;;GAMG;AACH,SAAS,yBAAyB,CAAC,GAAkC;IACnE,IAAI,CAAC,GAAG,IAAI,GAAG,KAAK,OAAO,CAAC,GAAG;QAAE,OAAO,SAAS,CAAC;IAClD,KAAK,MAAM,OAAO,IAAI,eAAe,EAAE,CAAC;QACtC,MAAM,GAAG,GAAG,GAAG,CAAC,OAAO,CAAC,CAAC;QACzB,IAAI,OAAO,GAAG,KAAK,QAAQ,IAAI,GAAG,CAAC,IAAI,EAAE,KAAK,EAAE;YAAE,SAAS;QAC3D,yEAAyE;QACzE,IAAI,OAAO,CAAC,GAAG,CAAC,OAAO,CAAC,KAAK,GAAG;YAAE,SAAS;QAC3C,MAAM,KAAK,GAAG,GAAG,CAAC,KAAK,CAAC,WAAW,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;QACpF,IAAI,KAAK;YAAE,OAAO,KAAK,CAAC;IAC1B,CAAC;IACD,OAAO,SAAS,CAAC;AACnB,CAAC;AAED,SAAS,WAAW,CAAC,GAAW;IAC9B,OAAO,2DAA2D,SAAS,wBAAwB,kBAAkB,CAAC,GAAG,CAAC,EAAE,CAAC;AAC/H,CAAC;AAED;;;GAGG;AACH,SAAS,aAAa,CAAC,QAAgB;IACrC,OAAO,IAAI,CAAC,KAAK,CAAC,CAAC,QAAQ,GAAG,CAAC,kBAAkB,GAAG,oBAAoB,GAAG,YAAY,CAAC,CAAC,GAAG,IAAI,CAAC,CAAC;AACpG,CAAC;AAED,+EAA+E;AAC/E,SAAS,YAAY,CAAC,GAAW;IAC/B,MAAM,QAAQ,GAAG,kBAAkB,GAAG,YAAY,GAAG,oBAAoB,CAAC;IAC1E,MAAM,UAAU,GAAG,YAAY,GAAG,oBAAoB,CAAC;IACvD,MAAM,MAAM,GAAG,MAAM,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;IAChC,MAAM,CAAC,KAAK,CAAC,MAAM,EAAE,CAAC,EAAE,OAAO,CAAC,CAAC;IACjC,MAAM,CAAC,aAAa,CAAC,EAAE,GAAG,GAAG,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC,CAAC,YAAY;IACtD,MAAM,CAAC,KAAK,CAAC,MAAM,EAAE,CAAC,EAAE,OAAO,CAAC,CAAC;IACjC,MAAM,CAAC,KAAK,CAAC,MAAM,EAAE,EAAE,EAAE,OAAO,CAAC,CAAC;IAClC,MAAM,CAAC,aAAa,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,CAAC,sBAAsB;IACpD,MAAM,CAAC,aAAa,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,oBAAoB;IACjD,MAAM,CAAC,aAAa,CAAC,YAAY,EAAE,EAAE,CAAC,CAAC;IACvC,MAAM,CAAC,aAAa,CAAC,kBAAkB,EAAE,EAAE,CAAC,CAAC;IAC7C,MAAM,CAAC,aAAa,CAAC,QAAQ,EAAE,EAAE,CAAC,CAAC;IACnC,MAAM,CAAC,aAAa,CAAC,UAAU,EAAE,EAAE,CAAC,CAAC;IACrC,MAAM,CAAC,aAAa,CAAC,oBAAoB,GAAG,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,gBAAgB;IACpE,MAAM,CAAC,KAAK,CAAC,MAAM,EAAE,EAAE,EAAE,OAAO,CAAC,CAAC;IAClC,MAAM,CAAC,aAAa,CAAC,GAAG,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC,CAAC,gBAAgB;IACtD,OAAO,MAAM,CAAC,MAAM,CAAC,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC,CAAC;AACtC,CAAC;AAED,yEAAyE;AACzE,SAAS,kBAAkB,CAAC,MAAe;IACzC,IAAI,OAAO,MAAM,KAAK,QAAQ,IAAI,MAAM,KAAK,IAAI;QAAE,OAAO,SAAS,CAAC;IACpE,MAAM,UAAU,GAAI,MAAmC,CAAC,UAAU,CAAC;IACnE,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,UAAU,CAAC,IAAI,UAAU,CAAC,MAAM,KAAK,CAAC;QAAE,OAAO,SAAS,CAAC;IAC5E,MAAM,KAAK,GAAI,UAAU,CAAC,CAAC,CAAuC,EAAE,OAAO,EAAE,KAAK,CAAC;IACnF,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC;QAAE,OAAO,SAAS,CAAC;IAC5C,KAAK,MAAM,IAAI,IAAI,KAAK,EAAE,CAAC;QACzB,MAAM,IAAI,GAAI,IAA4C,EAAE,UAAU,EAAE,IAAI,CAAC;QAC7E,IAAI,OAAO,IAAI,KAAK,QAAQ,IAAI,IAAI,CAAC,MAAM,GAAG,CAAC;YAAE,OAAO,IAAI,CAAC;IAC/D,CAAC;IACD,OAAO,SAAS,CAAC;AACnB,CAAC;AAED,MAAM,CAAC,MAAM,SAAS,GAAe;IACnC,EAAE,EAAE,YAAY;IAChB,IAAI,EAAE,KAAK;IACX,eAAe,EAAE,KAAK;IACtB,WAAW,EAAE,2CAA2C;IACxD,4EAA4E;IAC5E,uEAAuE;IACvE,oEAAoE;IACpE,WAAW,EAAE,EAAE;IACf,cAAc,EAAE,KAAK;IACrB,iBAAiB,EAAE,IAAI;IACvB,OAAO,EACL,yMAAyM;IAE3M,KAAK,CAAC,QAAQ,CAAC,KAAkB;QAC/B,MAAM,KAAK,GAAG,KAAK,CAAC,KAAK,IAAI,KAAK,CAAC,KAAK,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC,KAAK,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,aAAa,CAAC;QAErF,IAAI,CAAC,KAAK,CAAC,IAAI,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,IAAI,EAAE,EAAE,CAAC;YACtC,MAAM,IAAI,UAAU,CAAC,YAAY,EAAE,8CAA8C,EAAE;gBACjF,SAAS,EAAE,YAAY;aACxB,CAAC,CAAC;QACL,CAAC;QAED,0EAA0E;QAC1E,0EAA0E;QAC1E,kDAAkD;QAClD,IAAI,KAAK,CAAC,MAAM,EAAE,CAAC;YACjB,MAAM,MAAM,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,KAAK,CAAC,IAAI,CAAC,IAAI,EAAE,CAAC,MAAM,GAAG,oBAAoB,CAAC,CAAC;YAC5E,MAAM,UAAU,GAAG,IAAI,CAAC,KAAK,CAAC,MAAM,GAAG,IAAI,CAAC,CAAC;YAC7C,MAAM,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC,UAAU,CAAC,EAAE,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC,CAAC;YAC5D,MAAM,SAAS,CAAC,KAAK,CAAC,UAAU,EAAE,YAAY,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;YACjE,OAAO,EAAE,IAAI,EAAE,KAAK,CAAC,UAAU,EAAE,UAAU,EAAE,SAAS,EAAE,YAAY,EAAE,CAAC;QACzE,CAAC;QAED,MAAM,WAAW,GAAG,IAAI,CAAC,SAAS,CAAC;YACjC,QAAQ,EAAE,CAAC,EAAE,KAAK,EAAE,CAAC,EAAE,IAAI,EAAE,KAAK,CAAC,IAAI,EAAE,CAAC,EAAE,CAAC;YAC7C,gBAAgB,EAAE;gBAChB,kBAAkB,EAAE,CAAC,OAAO,CAAC;gBAC7B,YAAY,EAAE;oBACZ,WAAW,EAAE;wBACX,mBAAmB,EAAE,EAAE,SAAS,EAAE,KAAK,EAAE;qBAC1C;iBACF;aACF;SACF,CAAC,CAAC;QAEH,0EAA0E;QAC1E,kEAAkE;QAClE,4EAA4E;QAC5E,4EAA4E;QAC5E,MAAM,WAAW,GAAG,yBAAyB,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC;QAEzD,MAAM,WAAW,GAGb,EAAE,CAAC;QACP,IAAI,KAAK,CAAC,OAAO;YAAE,WAAW,CAAC,OAAO,GAAG,KAAK,CAAC,OAAO,CAAC;QACvD,IAAI,WAAW;YAAE,WAAW,CAAC,WAAW,GAAG,WAAW,CAAC;QAEvD,MAAM,QAAQ,GAAG,MAAM,mBAAmB,CACxC,WAAW,EACX;YACE,MAAM,EAAE,MAAM;YACd,OAAO,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE;YAC/C,IAAI,EAAE,WAAW;SAClB,EACD,WAAW,CACZ,CAAC;QAEF,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;YACjB,MAAM,IAAI,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC,EAAE,CAAC,CAAC;YACnD,MAAM,IAAI,UAAU,CAClB,YAAY,EACZ,uCAAuC,QAAQ,CAAC,MAAM,GAAG,EACzD,EAAE,SAAS,EAAE,YAAY,EAAE,MAAM,EAAE,QAAQ,CAAC,MAAM,EAAE,IAAI,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,EAAE,CAC/E,CAAC;QACJ,CAAC;QAED,IAAI,MAAe,CAAC;QACpB,IAAI,CAAC;YACH,MAAM,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;QACjC,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,MAAM,IAAI,UAAU,CAAC,YAAY,EAAE,0CAA0C,EAAE;gBAC7E,SAAS,EAAE,YAAY;gBACvB,KAAK,EAAE,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,GAAG,CAAC;aACxD,CAAC,CAAC;QACL,CAAC;QAED,MAAM,GAAG,GAAG,kBAAkB,CAAC,MAAM,CAAC,CAAC;QACvC,IAAI,CAAC,GAAG,EAAE,CAAC;YACT,MAAM,IAAI,UAAU,CAAC,YAAY,EAAE,8CAA8C,EAAE;gBACjF,SAAS,EAAE,YAAY;aACxB,CAAC,CAAC;QACL,CAAC;QAED,MAAM,GAAG,GAAG,MAAM,CAAC,IAAI,CAAC,GAAG,EAAE,QAAQ,CAAC,CAAC;QACvC,IAAI,GAAG,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YACrB,MAAM,IAAI,UAAU,CAAC,YAAY,EAAE,6CAA6C,EAAE;gBAChF,SAAS,EAAE,YAAY;aACxB,CAAC,CAAC;QACL,CAAC;QAED,MAAM,GAAG,GAAG,YAAY,CAAC,GAAG,CAAC,CAAC;QAC9B,MAAM,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC,UAAU,CAAC,EAAE,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC,CAAC;QAC5D,MAAM,SAAS,CAAC,KAAK,CAAC,UAAU,EAAE,GAAG,CAAC,CAAC;QAEvC,OAAO;YACL,IAAI,EAAE,KAAK,CAAC,UAAU;YACtB,UAAU,EAAE,aAAa,CAAC,GAAG,CAAC,MAAM,CAAC;YACrC,SAAS,EAAE,YAAY;SACxB,CAAC;IACJ,CAAC;CACF,CAAC"}

package/dist/video/audio-platform/native-lyria.d.ts

@@ -0,0 +1,35 @@
+/**
+ * Music generation via Google Vertex AI Lyria (predict endpoint).
+ *
+ * VERIFIED LIVE against `lyria-002` (Lyria 2, GA) on 2026-06-01 — the request
+ * body and response field below match the real API. INERT / unavailable
+ * without a Vertex project (GOOGLE_CLOUD_PROJECT or VERTEX_PROJECT must be set).
+ *
+ *   POST https://{REGION}-aiplatform.googleapis.com/v1/projects/{PROJECT}/
+ *        locations/{REGION}/publishers/google/models/{MODEL}:predict
+ *   Authorization: Bearer <token>
+ *   Body: { instances: [{ prompt: string }], parameters: { sampleCount: 1 } }
+ *   Response: { predictions: [{ bytesBase64Encoded: string (WAV), mimeType }] }
+ *
+ * MODEL (via LYRIA_MODEL env, default lyria-002):
+ *   - lyria-002             Lyria 2, GA, 30s, text-only             ($0.06/30s)
+ *   - lyria-3-clip-preview  Lyria 3, 30s, text+image  (PREVIEW — needs project allowlist)
+ *   - lyria-3-pro-preview   Lyria 3 Pro, ~184s full song, text+image (PREVIEW — allowlist)
+ * NOTE: the Lyria 3 image-input (multimodal) request shape is NOT yet wired or
+ * verified here — it needs preview allowlist access; add + verify once granted.
+ *
+ * Auth (tried in order): GOOGLE_ACCESS_TOKEN env -> GCE metadata server ->
+ * `gcloud auth print-access-token` (local dev, dependency-free). For prod/CI,
+ * set GOOGLE_ACCESS_TOKEN or run on a GCE/Cloud Run service account.
+ * requiresVertex: true.
+ */
+import type { MusicBackend } from './types.js';
+/**
+ * True duration (ms) of a RIFF/WAVE buffer, read from its fmt byteRate + data
+ * chunk size. Lyria returns a full WAV, so this reports the real clip length
+ * (~32.8s for lyria-002) rather than the requested-duration default. Returns
+ * null for a non-WAV / unparseable buffer (caller falls back to the request).
+ */
+export declare function wavDurationMs(buf: Buffer): number | null;
+export declare const lyriaBackend: MusicBackend;
+//# sourceMappingURL=native-lyria.d.ts.map

package/dist/video/audio-platform/native-lyria.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"native-lyria.d.ts","sourceRoot":"","sources":["../../../src/video/audio-platform/native-lyria.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;GAwBG;AAOH,OAAO,KAAK,EAAE,YAAY,EAAiC,MAAM,YAAY,CAAC;AAO9E;;;;;GAKG;AACH,wBAAgB,aAAa,CAAC,GAAG,EAAE,MAAM,GAAG,MAAM,GAAG,IAAI,CAwBxD;AA6DD,eAAO,MAAM,YAAY,EAAE,YAqG1B,CAAC"}

package/dist/video/audio-platform/native-lyria.js

@@ -0,0 +1,189 @@
+/**
+ * Music generation via Google Vertex AI Lyria (predict endpoint).
+ *
+ * VERIFIED LIVE against `lyria-002` (Lyria 2, GA) on 2026-06-01 — the request
+ * body and response field below match the real API. INERT / unavailable
+ * without a Vertex project (GOOGLE_CLOUD_PROJECT or VERTEX_PROJECT must be set).
+ *
+ *   POST https://{REGION}-aiplatform.googleapis.com/v1/projects/{PROJECT}/
+ *        locations/{REGION}/publishers/google/models/{MODEL}:predict
+ *   Authorization: Bearer <token>
+ *   Body: { instances: [{ prompt: string }], parameters: { sampleCount: 1 } }
+ *   Response: { predictions: [{ bytesBase64Encoded: string (WAV), mimeType }] }
+ *
+ * MODEL (via LYRIA_MODEL env, default lyria-002):
+ *   - lyria-002             Lyria 2, GA, 30s, text-only             ($0.06/30s)
+ *   - lyria-3-clip-preview  Lyria 3, 30s, text+image  (PREVIEW — needs project allowlist)
+ *   - lyria-3-pro-preview   Lyria 3 Pro, ~184s full song, text+image (PREVIEW — allowlist)
+ * NOTE: the Lyria 3 image-input (multimodal) request shape is NOT yet wired or
+ * verified here — it needs preview allowlist access; add + verify once granted.
+ *
+ * Auth (tried in order): GOOGLE_ACCESS_TOKEN env -> GCE metadata server ->
+ * `gcloud auth print-access-token` (local dev, dependency-free). For prod/CI,
+ * set GOOGLE_ACCESS_TOKEN or run on a GCE/Cloud Run service account.
+ * requiresVertex: true.
+ */
+import { mkdir, writeFile } from 'node:fs/promises';
+import { dirname } from 'node:path';
+import { execFile } from 'node:child_process';
+import { promisify } from 'node:util';
+import { VclawError } from '../errors.js';
+const execFileAsync = promisify(execFile);
+const DEFAULT_LYRIA_REGION = 'us-central1';
+const DEFAULT_LYRIA_MODEL = 'lyria-002';
+/**
+ * True duration (ms) of a RIFF/WAVE buffer, read from its fmt byteRate + data
+ * chunk size. Lyria returns a full WAV, so this reports the real clip length
+ * (~32.8s for lyria-002) rather than the requested-duration default. Returns
+ * null for a non-WAV / unparseable buffer (caller falls back to the request).
+ */
+export function wavDurationMs(buf) {
+    if (buf.length < 44 ||
+        buf.toString('ascii', 0, 4) !== 'RIFF' ||
+        buf.toString('ascii', 8, 12) !== 'WAVE') {
+        return null;
+    }
+    let byteRate = 0;
+    let dataSize = 0;
+    let off = 12;
+    while (off + 8 <= buf.length) {
+        const id = buf.toString('ascii', off, off + 4);
+        const size = buf.readUInt32LE(off + 4);
+        if (id === 'fmt ' && off + 24 <= buf.length) {
+            byteRate = buf.readUInt32LE(off + 16); // fmt body: +8 fmt/chan, +12 rate, +16 byteRate
+        }
+        else if (id === 'data') {
+            dataSize = size;
+            break;
+        }
+        off += 8 + size + (size % 2); // chunks are word-aligned
+    }
+    if (byteRate > 0 && dataSize > 0)
+        return Math.round((dataSize / byteRate) * 1000);
+    return null;
+}
+/**
+ * Resolve the Vertex project id from explicit env or well-known fallback vars.
+ */
+function resolveVertexProject(env) {
+    return env['GOOGLE_CLOUD_PROJECT'] ?? env['VERTEX_PROJECT'] ?? env['GCLOUD_PROJECT'];
+}
+/**
+ * Obtain a Vertex bearer token. Tries GOOGLE_APPLICATION_CREDENTIALS / ADC
+ * via the metadata server token endpoint; falls back to the
+ * GOOGLE_ACCESS_TOKEN env var for local-dev / CI injection.
+ *
+ * SCAFFOLD: In production this should use the Google Auth Library
+ * (e.g. google-auth-library) to obtain a proper access token via ADC.
+ * The current implementation accepts a pre-resolved token via env var
+ * GOOGLE_ACCESS_TOKEN for testability without a real service account.
+ */
+async function resolveVertexToken(env, fetcher) {
+    const explicit = env['GOOGLE_ACCESS_TOKEN'];
+    if (explicit)
+        return explicit;
+    // Attempt GCE metadata server (works on Cloud Run / GCE VMs).
+    try {
+        const metaResp = await fetcher('http://metadata.google.internal/computeMetadata/v1/instance/service-accounts/default/token', { headers: { 'Metadata-Flavor': 'Google' } });
+        if (metaResp.ok) {
+            const json = (await metaResp.json());
+            if (json.access_token)
+                return json.access_token;
+        }
+    }
+    catch {
+        // Not on GCE — fall through.
+    }
+    // Local-dev fallback: mint a token via the gcloud CLI if present. Keeps the
+    // Lyria backend dependency-free (no google-auth-library) while removing the
+    // need to refresh GOOGLE_ACCESS_TOKEN by hand. videoclaw already shells out
+    // to provider CLIs elsewhere (e.g. the veo-useapi Bun transport).
+    try {
+        const { stdout } = await execFileAsync('gcloud', ['auth', 'print-access-token'], {
+            timeout: 15_000,
+        });
+        const tok = stdout.trim();
+        if (tok)
+            return tok;
+    }
+    catch {
+        // gcloud absent / not authenticated — fall through to the clear error.
+    }
+    throw new VclawError('env_var_missing', 'No Vertex access token available. Set GOOGLE_ACCESS_TOKEN, run on a GCE/Cloud Run service account, or `gcloud auth login` (the CLI fallback).', { backend: 'lyria' });
+}
+export const lyriaBackend = {
+    id: 'lyria',
+    kind: 'music',
+    displayName: 'Lyria (Vertex AI)',
+    requiredEnv: [],
+    requiresVertex: true,
+    summary: 'Music generation via Google Vertex AI Lyria. Default lyria-002 (verified); LYRIA_MODEL can target lyria-3-clip-preview / lyria-3-pro-preview (preview, allowlist). Requires GOOGLE_CLOUD_PROJECT / VERTEX_PROJECT + a Vertex token.',
+    async generate(input) {
+        const env = input.env ?? process.env;
+        const fetcher = input.fetcher ?? fetch;
+        const durationSec = input.durationSec ?? 30;
+        if (!input.prompt.trim()) {
+            throw new VclawError('music_gen_failed', 'Cannot generate music with an empty prompt.', {
+                backend: 'lyria',
+            });
+        }
+        const project = resolveVertexProject(env);
+        if (!project) {
+            throw new VclawError('env_var_missing', 'GOOGLE_CLOUD_PROJECT (or VERTEX_PROJECT) is not set. The Lyria backend requires a Vertex project.', { backend: 'lyria' });
+        }
+        const region = env['VERTEX_REGION'] ?? DEFAULT_LYRIA_REGION;
+        const model = env['LYRIA_MODEL'] ?? DEFAULT_LYRIA_MODEL;
+        if (input.dryRun) {
+            // Write an empty placeholder so downstream existsSync() stays consistent
+            // with every other audio backend's dry-run path (lyria3/elevenlabs/etc.).
+            const dir = dirname(input.outputPath);
+            if (dir)
+                await mkdir(dir, { recursive: true });
+            await writeFile(input.outputPath, Buffer.alloc(0));
+            return { path: input.outputPath, durationMs: durationSec * 1000, backendId: 'lyria' };
+        }
+        const token = await resolveVertexToken(env, fetcher);
+        const endpoint = `https://${region}-aiplatform.googleapis.com/v1/projects/${project}` +
+            `/locations/${region}/publishers/google/models/${model}:predict`;
+        // Verified against lyria-002 (Lyria 2). Lyria 3 adds optional image input.
+        const body = {
+            instances: [{ prompt: input.prompt }],
+            parameters: { sampleCount: 1 },
+        };
+        const resp = await fetcher(endpoint, {
+            method: 'POST',
+            headers: {
+                Authorization: `Bearer ${token}`,
+                'Content-Type': 'application/json',
+            },
+            body: JSON.stringify(body),
+        });
+        if (!resp.ok) {
+            let detail = '';
+            try {
+                detail = await resp.text();
+            }
+            catch {
+                /* ignore */
+            }
+            throw new VclawError('music_gen_failed', `Vertex Lyria predict request failed: ${resp.status} ${resp.statusText}`, { status: resp.status, detail: detail.slice(0, 300), backend: 'lyria' });
+        }
+        // Verified response shape (lyria-002): predictions[0].bytesBase64Encoded (WAV).
+        const json = (await resp.json());
+        const prediction = json.predictions?.[0];
+        const b64 = prediction?.bytesBase64Encoded;
+        if (!b64) {
+            throw new VclawError('music_gen_failed', 'Vertex Lyria response did not include bytesBase64Encoded audio data.', { backend: 'lyria' });
+        }
+        const bytes = Buffer.from(b64, 'base64');
+        const dir = dirname(input.outputPath);
+        if (dir)
+            await mkdir(dir, { recursive: true });
+        await writeFile(input.outputPath, bytes);
+        // Report the WAV's true length when parseable (Lyria clips are ~32.8s),
+        // falling back to the requested duration only if the header is unreadable.
+        const durationMs = wavDurationMs(bytes) ?? durationSec * 1000;
+        return { path: input.outputPath, durationMs, backendId: 'lyria' };
+    },
+};
+//# sourceMappingURL=native-lyria.js.map