the language-detection pass is optional now

every transcription on this machine failed with "Language detection failed
(400)" before it started. the server was fine — it had answered /inference
correctly minutes earlier — and the platform was fine. the plugin's FIRST
call sends detect_language=true, which is a whisper.cpp parameter, and
whisperx answers 400 to it.

so a pass that exists to HELP became the only thing that could fail, on a
backend that does its own detection anyway. whisperx logs "No language
specified, language will be detected for each audio file" and gets it
right per file.

a backend refusing that request is telling us it detects for itself. skip
the pass, send no `language`, and read back whatever the transcription
reports. the two-pass path is untouched for whisper.cpp, which needs it,
and the spoken-language tie-break still applies wherever detection runs.

no prompt priming when the language is unknown either — TRANSCRIPT_PROMPTS
is keyed by language, and priming with the wrong one is worse than not
priming.

sidecar/transcribe.ts

@@ -330,7 +330,17 @@ export async function transcribeAudio({
 }: TranscribeArgs): Promise<TranscribeResult> {
   const base = whisperUrl.replace(/\/+$/, '');
 
-  // Step 1: detect language, and only fall back to the spoken-language bias when Whisper is unsure.
+  /*
+    Step 1: detect the language, and only fall back to the spoken-language bias when Whisper is unsure.
+
+    OPTIONAL, because `detect_language` is a whisper.cpp parameter and not every backend has it. WhisperX
+    answers 400 to this request and detects per file on its own — so a pass that exists to help became the
+    thing that failed every transcription before it started, on a server that was working perfectly.
+
+    A backend that refuses it is telling us it does its own detection. Skip the pass, send no `language`,
+    and read back whatever the transcription reports. The two-pass path is unchanged for whisper.cpp,
+    which needs it.
+  */
   const detectForm = new FormData();
   detectForm.append('file', file);
   detectForm.append('temperature', '0.0');
@@ -341,21 +351,33 @@ export async function transcribeAudio({
     method: 'POST',
     body: detectForm,
     timeout: false,
-  } as BunRequestInit);
-  if (!detectRes.ok) throw new Error(`Language detection failed (${detectRes.status})`);
-  const detectJson = (await detectRes.json()) as {
-    language?: string;
-    detected_language?: string;
-    detected_language_probability?: number;
-    language_probabilities?: Record<string, number>;
-  };
+  } as BunRequestInit).catch(() => null);
 
-  const probs = detectJson.language_probabilities;
+  const detectJson =
+    detectRes && detectRes.ok
+      ? ((await detectRes.json()) as {
+          language?: string;
+          detected_language?: string;
+          detected_language_probability?: number;
+          language_probabilities?: Record<string, number>;
+        })
+      : null;
+
+  if (!detectJson) {
+    console.log(
+      `[transcribe] no separate language detection (${detectRes ? `HTTP ${detectRes.status}` : 'unreachable'}) — letting the server detect`,
+    );
+  }
+
+  const probs = detectJson?.language_probabilities;
   const ranked = probs ? Object.entries(probs).sort((a, b) => b[1] - a[1]) : [];
   const [topCode, topProb] = ranked[0] ?? [];
 
-  let detectedLanguage: string;
-  if (topCode && (topProb ?? 0) >= CONFIDENT_DETECTION) {
+  let detectedLanguage: string | null;
+  if (!detectJson) {
+    // The server detects for itself. Nothing to bias and nothing to send.
+    detectedLanguage = null;
+  } else if (topCode && (topProb ?? 0) >= CONFIDENT_DETECTION) {
     // Whisper is sure. Trust it even when it is not a language the user speaks — that is exactly the
     // case translation exists for.
     detectedLanguage = topCode;
@@ -377,7 +399,9 @@ export async function transcribeAudio({
   // Step 2: transcribe in the detected language
   const transcribeForm = new FormData();
   transcribeForm.append('file', file);
-  transcribeForm.append('language', detectedLanguage);
+  // Absent when step 1 was skipped: the server detects, and telling it a language we did not detect
+  // would be a guess dressed as an instruction.
+  if (detectedLanguage) transcribeForm.append('language', detectedLanguage);
   transcribeForm.append('temperature', '0.0');
   transcribeForm.append('temperature_inc', '0.2');
   transcribeForm.append('beam_size', String(BEAM_SIZE));
@@ -391,7 +415,7 @@ export async function transcribeAudio({
   // work on every segment and nothing here reads them.
   transcribeForm.append('token_timestamps', 'false');
 
-  const prompt = TRANSCRIPT_PROMPTS[detectedLanguage];
+  const prompt = detectedLanguage ? TRANSCRIPT_PROMPTS[detectedLanguage] : undefined;
   if (prompt) {
     transcribeForm.append('prompt', prompt);
     transcribeForm.append('carry_initial_prompt', 'true');
@@ -410,10 +434,14 @@ export async function transcribeAudio({
   } as BunRequestInit);
   if (!res.ok) throw new Error(`Transcription failed (${res.status})`);
 
-  const json = (await res.json()) as { text?: string; segments?: WhisperSegment[] };
+  const json = (await res.json()) as { text?: string; segments?: WhisperSegment[]; language?: string };
   // Fall back to the flat text for any backend that does not return segments.
   const segments = json.segments ?? [];
   const text = segments.length > 0 ? toParagraphs(segments) : (json.text ?? '').trim();
 
-  return { text, detectedLanguage, segments };
+  // When step 1 was skipped, the transcription response is where the language comes from.
+  const reported = (json.language ?? '').toLowerCase();
+  const language = detectedLanguage ?? (LANG_NAME_TO_ISO[reported] ?? reported) ?? 'unknown';
+
+  return { text, detectedLanguage: language || 'unknown', segments };
 }