From b3b49bed802eae6df542569615a4a37e6ce09489 Mon Sep 17 00:00:00 2001
From: Marcus Castro <mcaxtr@gmail.com>
Date: Fri, 13 Feb 2026 14:09:04 -0300
Subject: [PATCH] fix(slack): override video/* MIME to audio/* for voice
 messages (#14941)

* fix(slack): override video/* MIME to audio/* for voice messages

* fix(slack): preserve overridden MIME in return value

* test(slack): fix media monitor MIME mock wiring

---------

Co-authored-by: Peter Steinberger <steipete@gmail.com>
---
 src/slack/monitor/media.test.ts | 74 +++++++++++++++++++++++++++++++++
 src/slack/monitor/media.ts      | 22 +++++++++-
 src/slack/types.ts              |  1 +
 3 files changed, 95 insertions(+), 2 deletions(-)

diff --git a/src/slack/monitor/media.test.ts b/src/slack/monitor/media.test.ts
index d9b35ab74b..dd1b3b41ac 100644
--- a/src/slack/monitor/media.test.ts
+++ b/src/slack/monitor/media.test.ts
@@ -238,6 +238,80 @@ describe("resolveSlackMedia", () => {
     expect(mockFetch).not.toHaveBeenCalled();
   });
 
+  it("overrides video/* MIME to audio/* for slack_audio voice messages", async () => {
+    // saveMediaBuffer re-detects MIME from buffer bytes, so it may return
+    // video/mp4 for MP4 containers.  Verify resolveSlackMedia preserves
+    // the overridden audio/* type in its return value despite this.
+    const saveMediaBufferMock = vi.spyOn(mediaStore, "saveMediaBuffer").mockResolvedValue({
+      path: "/tmp/voice.mp4",
+      contentType: "video/mp4",
+    });
+
+    const mockResponse = new Response(Buffer.from("audio data"), {
+      status: 200,
+      headers: { "content-type": "video/mp4" },
+    });
+    mockFetch.mockResolvedValueOnce(mockResponse);
+
+    const result = await resolveSlackMedia({
+      files: [
+        {
+          url_private: "https://files.slack.com/voice.mp4",
+          name: "audio_message.mp4",
+          mimetype: "video/mp4",
+          subtype: "slack_audio",
+        },
+      ],
+      token: "xoxb-test-token",
+      maxBytes: 16 * 1024 * 1024,
+    });
+
+    expect(result).not.toBeNull();
+    // saveMediaBuffer should receive the overridden audio/mp4
+    expect(saveMediaBufferMock).toHaveBeenCalledWith(
+      expect.any(Buffer),
+      "audio/mp4",
+      "inbound",
+      16 * 1024 * 1024,
+    );
+    // Returned contentType must be the overridden value, not the
+    // re-detected video/mp4 from saveMediaBuffer
+    expect(result!.contentType).toBe("audio/mp4");
+  });
+
+  it("preserves original MIME for non-voice Slack files", async () => {
+    const saveMediaBufferMock = vi.spyOn(mediaStore, "saveMediaBuffer").mockResolvedValue({
+      path: "/tmp/video.mp4",
+      contentType: "video/mp4",
+    });
+
+    const mockResponse = new Response(Buffer.from("video data"), {
+      status: 200,
+      headers: { "content-type": "video/mp4" },
+    });
+    mockFetch.mockResolvedValueOnce(mockResponse);
+
+    const result = await resolveSlackMedia({
+      files: [
+        {
+          url_private: "https://files.slack.com/clip.mp4",
+          name: "recording.mp4",
+          mimetype: "video/mp4",
+        },
+      ],
+      token: "xoxb-test-token",
+      maxBytes: 16 * 1024 * 1024,
+    });
+
+    expect(result).not.toBeNull();
+    expect(saveMediaBufferMock).toHaveBeenCalledWith(
+      expect.any(Buffer),
+      "video/mp4",
+      "inbound",
+      16 * 1024 * 1024,
+    );
+  });
+
   it("falls through to next file when first file returns error", async () => {
     vi.spyOn(mediaStore, "saveMediaBuffer").mockResolvedValue({
       path: "/tmp/test.jpg",
diff --git a/src/slack/monitor/media.ts b/src/slack/monitor/media.ts
index c96ca50234..e634a30dcb 100644
--- a/src/slack/monitor/media.ts
+++ b/src/slack/monitor/media.ts
@@ -115,6 +115,23 @@ export async function fetchWithSlackAuth(url: string, token: string): Promise<Re
   return fetch(resolvedUrl.toString(), { redirect: "follow" });
 }
 
+/**
+ * Slack voice messages (audio clips, huddle recordings) carry a `subtype` of
+ * `"slack_audio"` but are served with a `video/*` MIME type (e.g. `video/mp4`,
+ * `video/webm`).  Override the primary type to `audio/` so the
+ * media-understanding pipeline routes them to transcription.
+ */
+function resolveSlackMediaMimetype(
+  file: SlackFile,
+  fetchedContentType?: string,
+): string | undefined {
+  const mime = fetchedContentType ?? file.mimetype;
+  if (file.subtype === "slack_audio" && mime?.startsWith("video/")) {
+    return mime.replace("video/", "audio/");
+  }
+  return mime;
+}
+
 export async function resolveSlackMedia(params: {
   files?: SlackFile[];
   token: string;
@@ -144,16 +161,17 @@ export async function resolveSlackMedia(params: {
       if (fetched.buffer.byteLength > params.maxBytes) {
         continue;
       }
+      const effectiveMime = resolveSlackMediaMimetype(file, fetched.contentType);
       const saved = await saveMediaBuffer(
         fetched.buffer,
-        fetched.contentType ?? file.mimetype,
+        effectiveMime,
         "inbound",
         params.maxBytes,
       );
       const label = fetched.fileName ?? file.name;
       return {
         path: saved.path,
-        contentType: saved.contentType,
+        contentType: effectiveMime ?? saved.contentType,
         placeholder: label ? `[Slack file: ${label}]` : "[Slack file]",
       };
     } catch {
diff --git a/src/slack/types.ts b/src/slack/types.ts
index b87bdd739f..39a8d04ae1 100644
--- a/src/slack/types.ts
+++ b/src/slack/types.ts
@@ -2,6 +2,7 @@ export type SlackFile = {
   id?: string;
   name?: string;
   mimetype?: string;
+  subtype?: string;
   size?: number;
   url_private?: string;
   url_private_download?: string;