feat: clearer split between local (managed) and remote URLs and server status

3 files changed, 32 insertions, 10 deletions

diff --git a/app/converter/config.py b/app/converter/config.py
index 5e35ee3..9f70a73 100644
--- a/app/converter/config.py
+++ b/app/converter/config.py
@@ -25,6 +25,15 @@ BACKEND = "audiocpp"
 QWEN_API_URL     = "http://127.0.0.1:7860"  # CustomVoice model
 CLONE_API_URL    = "http://127.0.0.1:7861"  # Base model
 
+# Remote (externally-run) server URLs. The hub probes these and offers a
+# "[remote]" backend entry when one answers with the expected backend, so an
+# externally-started server can be used alongside a locally-managed one.
+# Leave empty to disable remote probing for that backend. The defaults match
+# the local ports so an external server squatting the local port is found
+# without any configuration.
+QWEN_REMOTE_URL  = "http://127.0.0.1:7860"  # CustomVoice model
+CLONE_REMOTE_URL = "http://127.0.0.1:7861"  # Base model
+
 # Custom voice options
 SPEAKER = "Vivian"  #Vivian, Serena, Uncle_Fu, Dylan, Eric, Ryan, Aiden, Ono_Anna, Sohee
 INSTRUCT = "Speak naturally and clearly, as if reading a dramatic book to an adult audience."
@@ -42,6 +51,7 @@ CONSTANT_SEED = False
 # BACKEND 2: faster-qwen-tts options                                          #
 ###############################################################################
 FASTER_API_URL   = "http://127.0.0.1:8000"    # faster-qwen3-tts server (Base model only)
+FASTER_REMOTE_URL = "http://127.0.0.1:8000"   # externally-run faster-qwen3-tts server ("" disables probing)
 
 # Default voice if no --voice is passed
 FASTER_VOICE = "default"
@@ -50,6 +60,7 @@ FASTER_VOICE = "default"
 # BACKEND 3: audio.cpp options                                                #
 ###############################################################################
 AUDIOCPP_API_URL = "http://127.0.0.1:8080"  # audio.cpp audiocpp_server
+AUDIOCPP_REMOTE_URL = "http://127.0.0.1:8080"  # externally-run audiocpp_server ("" disables probing)
 
 # Model ids in the audio.cpp server.json config. AUDIOCPP_MODEL_ID may point
 # at any TTS model entry the server hosts (qwen3_tts, higgs_audio_tts,
diff --git a/app/converter/converter.py b/app/converter/converter.py
index 4c1ffad..5f67f4a 100644
--- a/app/converter/converter.py
+++ b/app/converter/converter.py
@@ -142,7 +142,8 @@ class AudiobookConverter:
                  voice: Optional[str] = None, debug: bool = False,
                  model_id: Optional[str] = None,
                  instructions: Optional[str] = None,
-                 request_options: Optional[Dict[str, str]] = None):
+                 request_options: Optional[Dict[str, str]] = None,
+                 api_url: Optional[str] = None):
         if speed <= 0:
             raise ValueError(f"Speed must be a positive number, got {speed}")
         if output_format not in AUDIO_FORMATS:
@@ -171,7 +172,7 @@ class AudiobookConverter:
         if backend == BACKEND_FASTER:
             # The faster backend always voice-clones using a reference voice
             # configured on the server, so no local reference audio is needed.
-            self.tts = FasterTTSClient(voice=voice)
+            self.tts = FasterTTSClient(voice=voice, api_url=api_url)
         elif backend == BACKEND_AUDIOCPP:
             # Speaker mode (no voice) uses a built-in CustomVoice speaker;
             # an explicit voice selects a server-side preset (cloning).
@@ -181,7 +182,8 @@ class AudiobookConverter:
             self.tts = AudioCppTTSClient(voice=voice, language=self.language,
                                          model_id=model_id,
                                          instructions=instructions,
-                                         request_options=self.request_options)
+                                         request_options=self.request_options,
+                                         api_url=api_url)
         else:
             self.tts = QwenTTSClient(
                 voice_mode=voice_mode,
@@ -189,6 +191,7 @@ class AudiobookConverter:
                 voice_clone_ref_text=voice_clone_ref_text,
                 skip_transcription=skip_transcription,
                 language=self.language,
+                api_url=api_url,
             )
 
     def _validate_configuration(self) -> None:
@@ -585,8 +588,11 @@ class AudiobookConverter:
                 print(f"Request options: {self.request_options}")
             print(f"Language: {self.language}")
         else:
-            api_url = (config.CLONE_API_URL if self.voice_mode == VOICE_MODE_CLONE
-                       else config.QWEN_API_URL)
+            tts_client = getattr(self, "tts", None)
+            api_url = (getattr(tts_client, "api_url", None)
+                       or (config.CLONE_API_URL
+                           if self.voice_mode == VOICE_MODE_CLONE
+                           else config.QWEN_API_URL))
             print(f"Qwen API endpoint: {api_url}")
             print(f"Voice mode: {self.voice_mode}")
             print(f"Model size: {MODEL_SIZE} (always)")
diff --git a/app/converter/tts.py b/app/converter/tts.py
index a90dbbe..41e3aa5 100644
--- a/app/converter/tts.py
+++ b/app/converter/tts.py
@@ -348,7 +348,7 @@ class QwenTTSClient(_BaseTTSClient):
 
     def __init__(self, voice_mode: str = "custom_voice", voice_clone_ref_audio: Optional[str] = None,
                  voice_clone_ref_text: Optional[str] = None, skip_transcription: bool = False,
-                 language: Optional[str] = None):
+                 language: Optional[str] = None, api_url: Optional[str] = None):
         if voice_mode not in VOICE_MODES:
             raise ValueError(
                 f"Unknown voice mode: {voice_mode!r} (expected one of {VOICE_MODES})"
@@ -357,6 +357,9 @@ class QwenTTSClient(_BaseTTSClient):
         self.voice_clone_ref_audio = voice_clone_ref_audio
         self.voice_clone_ref_text = (voice_clone_ref_text or "").strip()
         self.skip_transcription = skip_transcription
+        # api_url overrides the configured endpoint for the active voice mode
+        # (used by the hub's "[remote]" backend entries and --api-url).
+        self.api_url = (api_url or "").strip() or None
         # Seed sent with every request: config.SEED as-is, or (with
         # CONSTANT_SEED and SEED < 0) one random value drawn per run and
         # reused for every request so the voice stays consistent across
@@ -379,16 +382,18 @@ class QwenTTSClient(_BaseTTSClient):
     # ------------------------------------------------------------------
 
     def _connect(self) -> None:
-        api_url = config.CLONE_API_URL if self.voice_mode == VOICE_MODE_CLONE else config.QWEN_API_URL
+        api_url = self.api_url or (
+            config.CLONE_API_URL if self.voice_mode == VOICE_MODE_CLONE
+            else config.QWEN_API_URL)
         try:
             if self.voice_mode == VOICE_MODE_CLONE:
                 # Voice clone uses the Base-model demo, which is a separate server
                 # from the CustomVoice demo (that one only exposes /run_instruct).
-                self._init_client(config.CLONE_API_URL, clone=True)
-                print(f"[OK] Connected to Voice Clone API at {config.CLONE_API_URL}")
+                self._init_client(api_url, clone=True)
+                print(f"[OK] Connected to Voice Clone API at {api_url}")
                 self._resolve_reference_text()
             else:
-                self._init_client(config.QWEN_API_URL, clone=False)
+                self._init_client(api_url, clone=False)
                 print("[OK] Connected to Qwen API")
         except Exception as exc:
             raise RuntimeError(