fix(audio): handle streaming WAV placeholder size 0xFFFFFFFF from Groq TTS

README.md

@@ -55,7 +55,7 @@ The current WebUI ships with 23 WebUI pages and 42 authenticated WebUI API route
 | **V2 platform features** | Semantic vector memory, associative memory graph, memory prioritization, prediction engine, predictive cache, anomaly detection, agent registry, task delegation, pipeline execution, self-correction loop, temporal context, zero-trust execution, audit trail, integrations, event bus/webhooks, dynamic dashboards, AI widget generator, feedback learning, adaptive prompting, and multi-agent network. |
 | **WebUI** | Dashboard widgets, tools controls, plugins marketplace, Soul editor, Memory, Workspace, Tasks, Workflows, Pipelines, Events, MCP, Integrations, Network, Hooks, Sessions, Analytics, Feedback, Security, Self-Improve, Autonomous Mode, Configuration, setup/login flow, and bilingual user guide. |
 | **Network and agents** | Managed agent runtimes with personal/bot auth, signed inter-agent messages, trust levels, replay protection, allowlist/recipient enforcement, manual inbox fallback, remote setup docs, and local-agent visibility in Network status totals. |
-| **Telegram and providers** | MTProto proxy recovery, Bot API proxy support, startup resilience when Telegram is unavailable, ffmpeg-free Telegram voice notes, Groq STT/TTS fixes, NVIDIA NIM provider support, OpenRouter free models, and 16-provider model catalog. |
+| **Telegram and providers** | MTProto proxy recovery, Bot API proxy support, startup resilience when Telegram is unavailable, ffmpeg-free Telegram voice notes (pure-JS OGG/Opus encoder; no system ffmpeg required), Groq STT/TTS support including streaming WAV header fix (`0xFFFFFFFF` placeholder), NVIDIA NIM provider support, OpenRouter free models, and 16-provider model catalog. |
 | **Security and reliability** | Wallet encryption, plugin isolation, exec allowlist mode, shell-free allowlist execution, TON-proxy checksum verification, SSRF guards, path/symlink hardening, auth-token hashing, CSRF fixes, restart locks, transcript caps, session TTL enforcement, CI restoration, and audit reports. |
 
 ---

src/utils/__tests__/audio.test.ts

@@ -0,0 +1,58 @@
+import { describe, it, expect } from "vitest";
+import { wavToOggOpus } from "../audio.js";
+
+/** Build a minimal valid 16-bit PCM WAV buffer. */
+function buildWavBuffer({
+  dataSize = 4,
+  chunkSize,
+  pcmData,
+}: {
+  dataSize?: number;
+  chunkSize?: number;
+  pcmData?: Buffer;
+} = {}): Buffer {
+  const payload = pcmData ?? Buffer.alloc(dataSize);
+  const actualDataSize = payload.length;
+  const writtenChunkSize = chunkSize ?? actualDataSize;
+  const fmtChunkSize = 16;
+  const riffSize = 4 + (8 + fmtChunkSize) + (8 + actualDataSize);
+  const buf = Buffer.alloc(8 + riffSize);
+  buf.write("RIFF", 0, "ascii");
+  buf.writeUInt32LE(riffSize, 4);
+  buf.write("WAVE", 8, "ascii");
+  buf.write("fmt ", 12, "ascii");
+  buf.writeUInt32LE(fmtChunkSize, 16);
+  buf.writeUInt16LE(1, 20); // PCM
+  buf.writeUInt16LE(1, 22); // mono
+  buf.writeUInt32LE(48000, 24); // sample rate
+  buf.writeUInt32LE(96000, 28); // byte rate
+  buf.writeUInt16LE(2, 32); // block align
+  buf.writeUInt16LE(16, 34); // bits per sample
+  buf.write("data", 36, "ascii");
+  buf.writeUInt32LE(writtenChunkSize, 40);
+  payload.copy(buf, 44);
+  return buf;
+}
+
+describe("wavToOggOpus", () => {
+  it("converts a valid WAV buffer to OGG/Opus", () => {
+    const wav = buildWavBuffer({ pcmData: Buffer.alloc(960 * 2) }); // 960 samples, 16-bit
+    const ogg = wavToOggOpus(wav);
+    expect(ogg.slice(0, 4).toString("ascii")).toBe("OggS");
+  });
+
+  it("handles streaming WAV where data chunk size is 0xFFFFFFFF (Groq TTS format)", () => {
+    // Groq TTS generates WAV with streaming placeholder size 0xFFFFFFFF
+    // because the total audio length is not known at the time the header is written.
+    const pcmData = Buffer.alloc(960 * 2); // 960 mono 16-bit samples = one Opus frame
+    const wav = buildWavBuffer({ pcmData, chunkSize: 0xffffffff });
+    // Must NOT throw "WAV parse error: chunk 'data' size 4294967295 exceeds buffer"
+    const ogg = wavToOggOpus(wav);
+    expect(ogg.slice(0, 4).toString("ascii")).toBe("OggS");
+  });
+
+  it("rejects a buffer that is not a WAV file", () => {
+    const buf = Buffer.from("not a wav file");
+    expect(() => wavToOggOpus(buf)).toThrow("WAV parse error: missing RIFF/WAVE header");
+  });
+});

src/utils/audio.ts

@@ -49,8 +49,12 @@ function parseWav(buf: Buffer): ParsedWav {
 
   while (offset + 8 <= buf.length) {
     const id = buf.toString("ascii", offset, offset + 4);
-    const size = buf.readUInt32LE(offset + 4);
     const start = offset + 8;
+    // 0xFFFFFFFF is a streaming/unknown-size placeholder used by some TTS providers
+    // (e.g. Groq) that write the WAV header before the total audio length is known.
+    // Treat it as "rest of buffer" rather than rejecting the file.
+    const rawSize = buf.readUInt32LE(offset + 4);
+    const size = rawSize === 0xffffffff ? buf.length - start : rawSize;
     if (start + size > buf.length) {
       throw new Error(`WAV parse error: chunk '${id}' size ${size} exceeds buffer`);
     }