From cb60172b514a34ee6b7ae720aab896b84d28c5d0 Mon Sep 17 00:00:00 2001
From: root <root@AI-M.intra.weidenhaupt.net>
Date: Wed, 13 May 2026 16:00:30 +0000
Subject: [PATCH] fix(mac-worker): handle MLX segments as list/tuple in
 addition to dict

lightning-whisper-mlx returns segments as [start, end, text] lists
(not dicts) in some versions, which broke transcribe(). Normalize all
three observed shapes: str output, list/tuple segments, dict segments.
Also fall back to joining segment texts when top-level text is empty.
---
 mac-worker/app/core/whisper_engine.py | 37 +++++++++++++++++++--------
 1 file changed, 26 insertions(+), 11 deletions(-)

diff --git a/mac-worker/app/core/whisper_engine.py b/mac-worker/app/core/whisper_engine.py
index 14244ca..4b5800d 100644
--- a/mac-worker/app/core/whisper_engine.py
+++ b/mac-worker/app/core/whisper_engine.py
@@ -53,19 +53,34 @@ class WhisperEngine:
 
         if self.engine == "mlx":
             out = self._impl.transcribe(audio_path=str(audio_path), language=lang)
-            # MLX-Output: {"text": "...", "segments": [...]} oder String. Robust mappen.
+            # MLX-Output schwankt zwischen Versionen:
+            # - String
+            # - {"text": "...", "segments": [...]}
+            # - {"text": "...", "segments": [[start, end, text], ...]}
+            # - {"text": "...", "segments": [{"start":..,"end":..,"text":..}, ...]}
             if isinstance(out, str):
                 return {"text": out, "language": lang, "segments": []}
-            segments = out.get("segments") or []
-            normalized = [
-                {
-                    "start": float(s.get("start", 0.0)),
-                    "end": float(s.get("end", 0.0)),
-                    "text": (s.get("text") or "").strip(),
-                }
-                for s in segments
-            ]
-            return {"text": out.get("text", "").strip(), "language": lang, "segments": normalized}
+            raw_segments = out.get("segments") or []
+            normalized = []
+            for s in raw_segments:
+                if isinstance(s, dict):
+                    normalized.append({
+                        "start": float(s.get("start", 0.0)),
+                        "end": float(s.get("end", 0.0)),
+                        "text": (s.get("text") or "").strip(),
+                    })
+                elif isinstance(s, (list, tuple)) and len(s) >= 3:
+                    normalized.append({
+                        "start": float(s[0] or 0.0),
+                        "end": float(s[1] or 0.0),
+                        "text": str(s[2] or "").strip(),
+                    })
+                else:
+                    normalized.append({"start": 0.0, "end": 0.0, "text": str(s).strip()})
+            text = (out.get("text") or "").strip()
+            if not text and normalized:
+                text = " ".join(seg["text"] for seg in normalized).strip()
+            return {"text": text, "language": lang, "segments": normalized}
 
         if self.engine == "faster":
             segments_iter, info = self._impl.transcribe(str(audio_path), language=lang, vad_filter=True)