fix(media): proper JSON error responses + fix false-positive empty transcript credential error

open-sse/handlers/audioSpeech.ts

@@ -24,13 +24,28 @@ import { errorResponse } from "../utils/error.ts";
  * Return a CORS error response from an upstream fetch failure
  */
 function upstreamErrorResponse(res, errText) {
-  return new Response(errText, {
-    status: res.status,
-    headers: {
-      "Content-Type": "application/json",
-      "Access-Control-Allow-Origin": getCorsOrigin(),
-    },
-  });
+  // Always return JSON so the client can detect 401/credential errors reliably
+  let errorMessage: string;
+  try {
+    const parsed = JSON.parse(errText);
+    errorMessage =
+      parsed?.err_msg ||
+      parsed?.error?.message ||
+      parsed?.error ||
+      parsed?.message ||
+      parsed?.detail ||
+      errText;
+  } catch {
+    errorMessage = errText || `Upstream error (${res.status})`;
+  }
+
+  return Response.json(
+    { error: { message: errorMessage, code: res.status } },
+    {
+      status: res.status,
+      headers: { "Access-Control-Allow-Origin": getCorsOrigin() },
+    }
+  );
 }
 
 /**

open-sse/handlers/audioTranscription.ts

@@ -26,13 +26,28 @@ type TranscriptionCredentials = {
  * Return a CORS error response from an upstream fetch failure
  */
 function upstreamErrorResponse(res, errText) {
-  return new Response(errText, {
-    status: res.status,
-    headers: {
-      "Content-Type": "application/json",
-      "Access-Control-Allow-Origin": getCorsOrigin(),
-    },
-  });
+  // Always return JSON so the client can parse the error reliably
+  let errorMessage: string;
+  try {
+    const parsed = JSON.parse(errText);
+    errorMessage =
+      parsed?.err_msg ||
+      parsed?.error?.message ||
+      parsed?.error ||
+      parsed?.message ||
+      parsed?.detail ||
+      errText;
+  } catch {
+    errorMessage = errText || `Upstream error (${res.status})`;
+  }
+
+  return Response.json(
+    { error: { message: errorMessage, code: res.status } },
+    {
+      status: res.status,
+      headers: { "Access-Control-Allow-Origin": getCorsOrigin() },
+    }
+  );
 }
 
 /**
@@ -71,9 +86,14 @@ async function handleDeepgramTranscription(providerConfig, file, modelId, token)
 
   const data = await res.json();
   // Transform Deepgram response to OpenAI Whisper format
-  const text = data.results?.channels?.[0]?.alternatives?.[0]?.transcript || "";
+  const text = data.results?.channels?.[0]?.alternatives?.[0]?.transcript ?? null;
 
-  return Response.json({ text }, { headers: { "Access-Control-Allow-Origin": getCorsOrigin() } });
+  // null means the audio had no recognizable speech (music, silence, etc.)
+  // Return it explicitly so the client can distinguish from a credentials error
+  return Response.json(
+    { text: text ?? "", noSpeechDetected: text === null || text === "" },
+    { headers: { "Access-Control-Allow-Origin": getCorsOrigin() } }
+  );
 }
 
 /**

src/app/(dashboard)/dashboard/media/MediaPageClient.tsx

@@ -328,6 +328,7 @@ function getVoiceList(providerId: string) {
 function parseApiError(raw: any, statusCode: number): { message: string; isCredentials: boolean } {
   const msg =
     raw?.error?.message ||
+    raw?.err_msg ||
     raw?.error ||
     raw?.message ||
     raw?.detail ||
@@ -340,6 +341,7 @@ function parseApiError(raw: any, statusCode: number): { message: string; isCrede
       msg.toLowerCase().includes("invalid api key") ||
       msg.toLowerCase().includes("unauthorized") ||
       msg.toLowerCase().includes("authentication") ||
+      msg.toLowerCase().includes("api key") ||
       statusCode === 401 ||
       statusCode === 403);
 
@@ -519,12 +521,22 @@ export default function MediaPageClient() {
           throw new Error(message);
         }
         const data = await res.json();
-        // Warn if text is empty (likely missing credentials that returned silently)
+        // Check for noSpeechDetected flag (music, silence, etc.) — NOT a credential error
+        if (data?.noSpeechDetected) {
+          setError(
+            `No speech detected in the audio file. If you uploaded music or a silent file, try an audio file with spoken words. Provider: "${selectedProvider}".`
+          );
+          setIsCredentialsError(false);
+          setLoading(false);
+          return;
+        }
+        // Warn if text is empty without the noSpeechDetected flag (unexpected)
         if (data && typeof data.text === "string" && data.text.trim() === "") {
           setError(
-            `Transcription returned empty text. Make sure you have a valid API key for "${selectedProvider}" configured in /dashboard/providers.`
+            `Transcription returned empty text. The audio may contain no recognizable speech, or the "${selectedProvider}" API key may be invalid. Check Dashboard → Logs → Proxy for details.`
           );
-          setIsCredentialsError(true);
+          // Only mark as credential error if we can confirm it from context
+          setIsCredentialsError(false);
           setLoading(false);
           return;
         }