fix(gateway): voice module critical bugs

- shutdown ordering: DB pool closed AFTER voice disconnect (was before)
- muxer: amix (mix) -> concat (sequential) filter for OGG segments
- upload retries: 0->3 with proper backoff 1-10s
- teleUpload retry backoff: minTimeout 0->1000, maxTimeout 0->10000
- mediaSource: 64KB buffer bound on readFirstTwoLines, 1MB on extractMediaInfo
- ffmpegProcess: pipe stderr to structured logger instead of inherit

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

services/discord-gateway/src/app/shutdown.ts

@@ -36,19 +36,23 @@ export function createGracefulShutdown(options: GracefulShutdownOptions) {
     try {
       options.stopMetricsServer?.();
       stopMuxerWorker();
-      options.logger.info("Closing database...");
-      await options.closeDatabase();
-      options.logger.info("Database closed");
 
+      // 1. Voice disconnect (may write final DB records)
       options.logger.info("Stopping voice connection...");
       await options.voiceController.disconnect();
 
+      // 2. Close Redis/pubsub after voice is done
       options.logger.info("Closing event broadcaster...");
       await options.eventBroadcaster.close();
 
       options.logger.info("Closing command handler...");
       await options.commandHandler.close();
 
+      // 3. DB pool LAST – voice disconnect may finalize recordings
+      options.logger.info("Closing database...");
+      await options.closeDatabase();
+      options.logger.info("Database closed");
+
       options.logger.info("Destroying Discord client...");
       try {
         options.client.destroy();

services/discord-gateway/src/modules/voice-recording/ffmpegProcess.ts

@@ -48,7 +48,12 @@ export function runFfmpeg(args: string[]): Promise<void> {
     logger.debug({ args }, "Starting ffmpeg");
 
     const proc = spawn("ffmpeg", args, {
-      stdio: ["ignore", "inherit", "inherit"],
+      stdio: ["ignore", "pipe", "pipe"],
+    });
+
+    let stderrBuf = "";
+    proc.stderr?.on("data", (chunk: Buffer) => {
+      stderrBuf += chunk.toString("utf8");
     });
 
     proc.on("close", (code) => {
@@ -56,8 +61,9 @@ export function runFfmpeg(args: string[]): Promise<void> {
         logger.debug("ffmpeg completed successfully");
         resolve();
       } else {
-        logger.warn({ exitCode: code }, "ffmpeg exited with non-zero code");
-        reject(new Error(`ffmpeg exited with code ${code}`));
+        const detail = stderrBuf.trim().slice(0, 2000);
+        logger.warn({ exitCode: code, stderr: detail }, "ffmpeg exited with non-zero code");
+        reject(new Error(`ffmpeg exited with code ${code}: ${detail}`));
       }
     });
 

services/discord-gateway/src/modules/voice-recording/mediaSource.ts

@@ -61,7 +61,12 @@ function parseSeconds(value: string): number {
  * Returns the parsed header and a new Readable that contains all remaining
  * data (the audio stream).
  */
-function readFirstTwoLines(stdout: Readable): Promise<{
+const MAX_HEADER_BUFFER = 65536; // 64KB safety limit for metadata headers
+
+function readFirstTwoLines(
+  stdout: Readable,
+  maxBufferSize: number = MAX_HEADER_BUFFER,
+): Promise<{
   title: string;
   duration: number;
   remaining: Readable;
@@ -82,6 +87,11 @@ function readFirstTwoLines(stdout: Readable): Promise<{
     function onData(chunk: Buffer) {
       if (stage === "done") return;
       buffer = Buffer.concat([buffer, chunk]);
+      if (buffer.length > maxBufferSize) {
+        cleanup();
+        reject(new Error(`Metadata header exceeded ${maxBufferSize} bytes`));
+        return;
+      }
       processBuffer();
     }
 
@@ -297,10 +307,15 @@ export async function extractMediaInfo(url: string): Promise<MediaInfo> {
     let stdoutBuf = "";
     let stderrBuf = "";
     const MAX_STDERR = 4096;
+    const MAX_STDOUT = 1_048_576; // 1MB safety limit
 
     if (proc.stdout) {
       proc.stdout.on("data", (chunk: Buffer) => {
-        stdoutBuf += chunk.toString("utf8");
+        if (stdoutBuf.length < MAX_STDOUT) {
+          stdoutBuf += chunk
+            .toString("utf8")
+            .slice(0, MAX_STDOUT - stdoutBuf.length);
+        }
       });
     }
 

services/discord-gateway/src/modules/voice-recording/muxer.ts

@@ -134,10 +134,10 @@ async function processJob(
       "Processing muxer job",
     );
 
-    // Build filter: concat all inputs with a crossfade or simple concat
-    // We use amix for mixing multiple audio streams (not sequential concat)
-    const inputLabels = data.inputs.map((_, i) => `[${i}:a]`);
-    const filterComplex = `${inputLabels.join("")}amix=inputs=${data.inputs.length}:duration=first:dropout_transition=2[out]`;
+    // Sequential concat of all OGG segments (amix = simultaneous mix, wrong for this)
+    const inputLabels = data.inputs.map((_, i) => `[${i}:a:0]`);
+    const n = data.inputs.length;
+    const filterComplex = `${inputLabels.join("")}concat=n=${n}:v=0:a=1[out]`;
 
     const args = buildMuxFfmpegArgs({
       inputs: data.inputs,

services/discord-gateway/src/modules/voice-recording/recorder/uploader.ts

@@ -60,7 +60,7 @@ export async function uploadRecordingSegment(input: {
       filename: fileName,
       contentType: "audio/ogg",
       uploadUrl: config.TELE_UPLOAD_URL,
-      retries: 0,
+      retries: 3,
     });
     const downloadUrl = uploadResult.url;
 

services/discord-gateway/src/modules/voice-recording/teleUpload.ts

@@ -71,8 +71,8 @@ export async function uploadToTele(input: {
     },
     {
       retries,
-      minTimeout: 0,
-      maxTimeout: 0,
+      minTimeout: 1000,
+      maxTimeout: 10000,
     },
   );