From 38703ed9a1b5f128b6c79041c1043bea5ea9bf3d Mon Sep 17 00:00:00 2001
From: Peter Steinberger <steipete@gmail.com>
Date: Sat, 25 Apr 2026 11:05:33 +0100
Subject: [PATCH] fix(discord): identify voice attachment metadata

---
 .../discord/src/monitor/message-utils.test.ts | 26 ++++++++++++++++
 .../discord/src/monitor/message-utils.ts      |  7 +++++
 .../src/monitor/preflight-audio.test.ts       | 30 +++++++++++++++++++
 .../discord/src/monitor/preflight-audio.ts    | 16 ++++++++--
 4 files changed, 76 insertions(+), 3 deletions(-)

diff --git a/extensions/discord/src/monitor/message-utils.test.ts b/extensions/discord/src/monitor/message-utils.test.ts
index 6d33b0b385d..c81b0f0ce94 100644
--- a/extensions/discord/src/monitor/message-utils.test.ts
+++ b/extensions/discord/src/monitor/message-utils.test.ts
@@ -560,6 +560,32 @@ describe("resolveMediaList", () => {
     ]);
   });
 
+  it("classifies Discord voice attachments by waveform metadata", async () => {
+    const attachment = {
+      id: "att-voice-metadata",
+      url: "https://cdn.discordapp.com/attachments/1/voice",
+      filename: "voice",
+      duration_secs: 1.5,
+      waveform: "AAAA",
+    };
+    fetchRemoteMedia.mockRejectedValueOnce(new Error("blocked by ssrf guard"));
+
+    const result = await resolveMediaList(
+      asMessage({
+        attachments: [attachment],
+      }),
+      512,
+    );
+
+    expect(result).toEqual([
+      {
+        path: attachment.url,
+        contentType: undefined,
+        placeholder: "<media:audio>",
+      },
+    ]);
+  });
+
   it("falls back to URL when saveMediaBuffer fails", async () => {
     const attachment = {
       id: "att-save-fail",
diff --git a/extensions/discord/src/monitor/message-utils.ts b/extensions/discord/src/monitor/message-utils.ts
index 4ac0c8587a8..c047720e2a9 100644
--- a/extensions/discord/src/monitor/message-utils.ts
+++ b/extensions/discord/src/monitor/message-utils.ts
@@ -44,6 +44,10 @@ function isDiscordAudioAttachmentFileName(fileName?: string | null): boolean {
   return Boolean(ext && AUDIO_ATTACHMENT_EXTENSIONS.has(ext));
 }
 
+function hasDiscordVoiceAttachmentFields(attachment: APIAttachment): boolean {
+  return typeof attachment.duration_secs === "number" || typeof attachment.waveform === "string";
+}
+
 function mergeHostnameList(...lists: Array<string[] | undefined>): string[] | undefined {
   const merged = lists
     .flatMap((list) => list ?? [])
@@ -578,6 +582,9 @@ function inferPlaceholder(attachment: APIAttachment): string {
   if (mime.startsWith("audio/")) {
     return "<media:audio>";
   }
+  if (hasDiscordVoiceAttachmentFields(attachment)) {
+    return "<media:audio>";
+  }
   if (isDiscordAudioAttachmentFileName(attachment.filename ?? attachment.url)) {
     return "<media:audio>";
   }
diff --git a/extensions/discord/src/monitor/preflight-audio.test.ts b/extensions/discord/src/monitor/preflight-audio.test.ts
index be42324d893..9d492f2724e 100644
--- a/extensions/discord/src/monitor/preflight-audio.test.ts
+++ b/extensions/discord/src/monitor/preflight-audio.test.ts
@@ -77,6 +77,36 @@ describe("resolveDiscordPreflightAudioMentionContext", () => {
     );
   });
 
+  it("preflights Discord voice attachments by waveform metadata", async () => {
+    transcribeFirstAudioMock.mockResolvedValue("metadata transcript");
+
+    await resolveDiscordPreflightAudioMentionContext({
+      message: {
+        attachments: [
+          {
+            url: " https://cdn.discordapp.com/attachments/voice ",
+            filename: "voice",
+            duration_secs: 1.5,
+            waveform: "AAAA",
+          },
+        ],
+      },
+      isDirectMessage: true,
+      shouldRequireMention: false,
+      mentionRegexes: [],
+      cfg,
+    });
+
+    expect(transcribeFirstAudioMock).toHaveBeenCalledWith(
+      expect.objectContaining({
+        ctx: expect.objectContaining({
+          MediaUrls: ["https://cdn.discordapp.com/attachments/voice"],
+          MediaTypes: ["audio/ogg"],
+        }),
+      }),
+    );
+  });
+
   it("does not preflight typed direct-message audio", async () => {
     const result = await resolveDiscordPreflightAudioMentionContext({
       message: {
diff --git a/extensions/discord/src/monitor/preflight-audio.ts b/extensions/discord/src/monitor/preflight-audio.ts
index d1d6acb3624..98706f47b83 100644
--- a/extensions/discord/src/monitor/preflight-audio.ts
+++ b/extensions/discord/src/monitor/preflight-audio.ts
@@ -1,6 +1,7 @@
 import type { OpenClawConfig } from "openclaw/plugin-sdk/config-runtime";
 import { getFileExtension } from "openclaw/plugin-sdk/media-mime";
 import { logVerbose } from "openclaw/plugin-sdk/runtime-env";
+import { normalizeOptionalString } from "openclaw/plugin-sdk/text-runtime";
 
 type DiscordPreflightAudioRuntime = typeof import("./preflight-audio.runtime.js");
 
@@ -13,8 +14,10 @@ function loadDiscordPreflightAudioRuntime(): Promise<DiscordPreflightAudioRuntim
 
 type DiscordAudioAttachment = {
   content_type?: string;
+  duration_secs?: number;
   filename?: string;
   url?: string;
+  waveform?: string;
 };
 
 const AUDIO_ATTACHMENT_MIME_BY_EXT = new Map([
@@ -30,10 +33,16 @@ const AUDIO_ATTACHMENT_MIME_BY_EXT = new Map([
 ]);
 
 function inferAudioAttachmentMime(attachment: DiscordAudioAttachment): string | undefined {
-  const contentType = attachment.content_type?.trim();
+  const contentType = normalizeOptionalString(attachment.content_type);
   if (contentType?.startsWith("audio/")) {
     return contentType;
   }
+  if (
+    typeof attachment.duration_secs === "number" ||
+    typeof normalizeOptionalString(attachment.waveform) === "string"
+  ) {
+    return "audio/ogg";
+  }
   const ext = getFileExtension(attachment.filename ?? attachment.url);
   return ext ? AUDIO_ATTACHMENT_MIME_BY_EXT.get(ext) : undefined;
 }
@@ -45,7 +54,7 @@ function collectAudioAttachments(
     return [];
   }
   return attachments.filter(
-    (att) => typeof att.url === "string" && att.url.length > 0 && inferAudioAttachmentMime(att),
+    (att) => normalizeOptionalString(att.url) && inferAudioAttachmentMime(att),
   );
 }
 
@@ -91,7 +100,8 @@ export async function resolveDiscordPreflightAudioMentionContext(params: {
       }
       const audioUrls = audioAttachments
         .map((att) => att.url)
-        .filter((url): url is string => typeof url === "string" && url.length > 0);
+        .map((url) => normalizeOptionalString(url))
+        .filter((url): url is string => Boolean(url));
       if (audioUrls.length > 0) {
         transcript = await transcribeFirstAudio({
           ctx: {