Add profile presets and min-p controls, embeddings and VoxCPM workers

server.py

@@ -14,6 +14,7 @@ from tts_runtime import TTSRuntime
 from stt import STT,read_upload
 from audio_policy import AudioPolicy
 from tts_test import TTSTests
+from voxcpm_test import VoxCPMTests
 from profiles import Profiles
 from capacity import assess, overview
 from runtime import Runtime
@@ -104,8 +105,14 @@ class Server(ThreadingHTTPServer):
             raise RuntimeError('Serverzugang muss vor dem Start eingerichtet werden.')
         self.worker=LlamaWorker(self.catalog.root.parent/'llama-worker',self.catalog,self.runtime)
         self.scheduler=Scheduler(self.worker)
+        self.voxcpm_tests=VoxCPMTests(self.catalog.root.parent/'voice-tests',self.profiles,self.voxcpm_runtime)
+        self.voxcpm_tests.acquire=lambda wait=False:self.scheduler.image_reservation(wait,kind='voice')
         self.separator_runtime=SeparatorRuntime(self.catalog.root.parent/'separator-runtime')
         self.separator_tests=SeparatorTests(self.catalog.root.parent/'separator-tests',self.separator_runtime,self.scheduler)
+        from embeddings import Embeddings
+        import tempfile
+        self.embeddings=Embeddings(tempfile.mkdtemp(prefix='deck-embeddings-'),self.catalog,self.worker)
+        self.profiles.embedding_blockers=self.embeddings.blockers
         self.profiles.chat_blockers=self.worker.blockers
         self.image_tests.acquire=self.scheduler.image_reservation
         self.prompt_enhancer.acquire=self.scheduler.image_reservation
@@ -114,12 +121,14 @@ class Server(ThreadingHTTPServer):
         self.stt=STT(self.profiles,self.catalog,self.runtime)
         self.profiles.stt_blockers=self.stt.blockers
         self.endpoint=Endpoint(self.catalog.root.parent,self.profiles,self.worker,self.scheduler,self.image_tests,self.credentials,self.server_port)
+        self.endpoint.embeddings=self.embeddings
         self.endpoint.tts=self.tts_tests
+        self.endpoint.voxcpm=self.voxcpm_tests
         self.endpoint.stt=self.stt
         self.chat_tests=ChatTests(self.profiles,self.worker,self.scheduler)
         self.auto_tests=AutoTests(self.catalog.root.parent/'auto-tests.json',self.profiles,self.worker,self.scheduler)
         def stop_gpu_work():
-            self.separator_tests.stop();self.music.stop();self.auto_tests.stop();self.chat_tests.stop();self.image_tests.stop();self.prompt_enhancer.stop_preview();self.tts_tests.stop();self.worker.stop()
+            self.voxcpm_tests.stop();self.separator_tests.stop();self.music.stop();self.auto_tests.stop();self.chat_tests.stop();self.image_tests.stop();self.prompt_enhancer.stop_preview();self.tts_tests.stop();self.worker.stop()
         from video_original import VideoRuntimes
         from ltx_original_runtime import LTXOriginalRuntime
         from audio_cpp_runtime import AudioCppRuntime
@@ -306,7 +315,7 @@ class Handler(BaseHTTPRequestHandler):
         video_runtime = (self.server.video.original_runtime
                          if self.server.video.runtime_kind() == 'original'
                          else self.server.video.runtime)
-        return {'stt':chat,'chat':chat,
+        return {'embeddings':chat,'stt':chat,'chat':chat,
                 'image':self.server.image_runtime.status().get('installed',False),
                 'audio':self.server.tts_runtime.status().get('installed',False),
                 'video':video_runtime.status().get('installed',False),
@@ -327,7 +336,7 @@ class Handler(BaseHTTPRequestHandler):
             return self.respond({'error':'Anmeldung erforderlich.'},401)
         if not self.authenticated() and self.path == '/':
             return self.respond((ROOT/'login.html').read_bytes(), mime='text/html; charset=utf-8')
-        routes = {'/voxcpm-ui.js':('voxcpm-ui.js','text/javascript'),'/separator-ui.js':('separator-ui.js','text/javascript'),'/music-ui.js':('music-ui.js','text/javascript'),'/audio-cpp-ui.js':('audio-cpp-ui.js','text/javascript'),'/ltx-original-ui.js':('ltx-original-ui.js','text/javascript'),'/backup-ui.js':('backup-ui.js','text/javascript'),'/dashboard-ui.js':('dashboard-ui.js','text/javascript'),'/dashboard.css':('dashboard.css','text/css'),'/themes.css':('themes.css','text/css'),'/video-ui.js':('video-ui.js','text/javascript'),'/stt-ui.js':('stt-ui.js','text/javascript'),'/tts-ui.js':('tts-ui.js','text/javascript'),'/auto-test-ui.js':('auto-test-ui.js','text/javascript'),'/chat-test-ui.js':('chat-test-ui.js','text/javascript'),'/endpoint-ui.js':('endpoint-ui.js','text/javascript'),'/docker-ui.js': ('docker-ui.js','text/javascript'), '/': ('index.html', 'text/html; charset=utf-8'), '/app.js': ('app.js', 'text/javascript'), '/style.css': ('style.css', 'text/css'), '/network-ui.js': ('network-ui.js', 'text/javascript'), '/access-ui.js': ('access-ui.js', 'text/javascript'), '/studio.js': ('studio.js', 'text/javascript'), '/catalog-ui.js': ('catalog-ui.js','text/javascript'), '/runtime-ui.js': ('runtime-ui.js','text/javascript'), '/profiles-ui.js': ('profiles-ui.js','text/javascript'), '/image-test-ui.js': ('image-test-ui.js','text/javascript')}
+        routes = {'/voice-ui.js':('voice-ui.js','text/javascript'),'/voxcpm-ui.js':('voxcpm-ui.js','text/javascript'),'/separator-ui.js':('separator-ui.js','text/javascript'),'/music-ui.js':('music-ui.js','text/javascript'),'/audio-cpp-ui.js':('audio-cpp-ui.js','text/javascript'),'/ltx-original-ui.js':('ltx-original-ui.js','text/javascript'),'/backup-ui.js':('backup-ui.js','text/javascript'),'/dashboard-ui.js':('dashboard-ui.js','text/javascript'),'/dashboard.css':('dashboard.css','text/css'),'/themes.css':('themes.css','text/css'),'/video-ui.js':('video-ui.js','text/javascript'),'/stt-ui.js':('stt-ui.js','text/javascript'),'/tts-ui.js':('tts-ui.js','text/javascript'),'/auto-test-ui.js':('auto-test-ui.js','text/javascript'),'/chat-test-ui.js':('chat-test-ui.js','text/javascript'),'/endpoint-ui.js':('endpoint-ui.js','text/javascript'),'/docker-ui.js': ('docker-ui.js','text/javascript'), '/': ('index.html', 'text/html; charset=utf-8'), '/app.js': ('app.js', 'text/javascript'), '/style.css': ('style.css', 'text/css'), '/network-ui.js': ('network-ui.js', 'text/javascript'), '/access-ui.js': ('access-ui.js', 'text/javascript'), '/studio.js': ('studio.js', 'text/javascript'), '/catalog-ui.js': ('catalog-ui.js','text/javascript'), '/runtime-ui.js': ('runtime-ui.js','text/javascript'), '/profiles-ui.js': ('profiles-ui.js','text/javascript'), '/image-test-ui.js': ('image-test-ui.js','text/javascript')}
         if self.path in routes:
             name, mime = routes[self.path]
             return self.respond((ROOT/name).read_bytes(), mime=mime)
@@ -341,6 +350,10 @@ class Handler(BaseHTTPRequestHandler):
         if self.path == '/api/v1/endpoint':return self.respond(self.server.endpoint.status())
         if self.path == '/api/v1/docker':return self.respond(self.server.docker.status())
         if self.path == '/api/v1/stt':return self.respond(self.server.stt.status())
+        if self.path == '/api/v1/voice':return self.respond(self.server.voxcpm_tests.status())
+        if urlsplit(self.path).path == '/api/v1/voice/audio':
+            try:return self.respond(self.server.voxcpm_tests.audio(parse_qs(urlsplit(self.path).query).get('id',[''])[0]),mime='audio/wav')
+            except (ValueError,OSError) as exc:return self.respond({'error':str(exc)},400)
         if self.path == '/api/v1/tts':return self.respond(self.server.tts_tests.status())
         if self.path == '/api/v1/audio-policy':return self.respond(self.server.audio_policy.status())
         if urlsplit(self.path).path == '/api/v1/tts/audio':
@@ -623,6 +636,21 @@ class Handler(BaseHTTPRequestHandler):
                 if set(data)-{'profile_id','lyrics','style','seed','max_tokens','steps'} or not {'profile_id','lyrics','style'}<=set(data):raise ValueError('Ungültige Musikfelder.')
                 return self.respond(self.server.music.start(**data))
             except ValueError as exc:return self.respond({'error':str(exc)},400)
+        if self.path in ('/api/v1/voice/start','/api/v1/voice/cancel'):
+            try:
+                if self.path.endswith('/cancel'):
+                    if self.read_json():raise ValueError('Keine Abbruchparameter erwartet.')
+                    return self.respond(self.server.voxcpm_tests.stop())
+                if self.headers.get('Content-Type','').startswith('multipart/form-data'):
+                    fields,audio=read_upload(self,validate=False,allowed_fields=('profile_id','text','device','cfg','steps','seed'),max_field=4096)
+                    fields['reference']=audio
+                    for key in ('steps','seed'):
+                        if key in fields:fields[key]=int(fields[key])
+                    if 'cfg' in fields:fields['cfg']=float(fields['cfg'])
+                else:fields=self.read_json()
+                if set(fields)-{'profile_id','text','device','cfg','steps','seed','reference'} or not {'profile_id','text'}<=set(fields):raise ValueError('Ungültige VoxCPM2-Anfrage.')
+                return self.respond(self.server.voxcpm_tests.start(**fields))
+            except (ValueError,OSError) as exc:return self.respond({'error':str(exc)},400)
         if self.path in ('/api/v1/tts/install','/api/v1/tts/install-cancel','/api/v1/tts/assign','/api/v1/tts/start','/api/v1/tts/cancel'):
             try:
                 data=self.read_json();t=self.server.tts_tests
@@ -683,7 +711,9 @@ class Handler(BaseHTTPRequestHandler):
         if self.path == '/api/v1/profiles/delete':
             try:
                 data=self.read_json()
-                with self.server.image_tests.lock, self.server.tts_tests.lock, self.server.stt.lock:
+                with self.server.image_tests.lock, self.server.tts_tests.lock, self.server.stt.lock, self.server.voxcpm_tests.lock:
+                    voice_job=self.server.voxcpm_tests.job
+                    if voice_job and voice_job.get('state')=='running' and voice_job.get('profile_id')==data.get('id'):raise ValueError('Stimmprofil wird gerade ausgeführt. Zuerst Auftrag abbrechen.')
                     stt_job=self.server.stt.job
                     if stt_job and stt_job.get("state")=="running" and stt_job.get("profile_id")==data.get("id"):raise ValueError("Dieses STT-Profil wird gerade ausgeführt. Zuerst den Auftrag beenden.")
                     tts_job=self.server.tts_tests.job
@@ -696,6 +726,10 @@ class Handler(BaseHTTPRequestHandler):
                     return self.respond(result)
             except ValueError as exc:return self.respond({'error':str(exc)},400)
             except OSError:return self.respond({'error':'Profil konnte nicht gelöscht werden.'},503)
+        if self.path == '/api/v1/profiles/preset':
+            try:return self.respond(self.server.profiles.preset(self.read_json()))
+            except ValueError as exc:return self.respond({'error':str(exc)},400)
+            except OSError:return self.respond({'error':'Preset konnte nicht gespeichert werden.'},503)
         if self.path == '/api/v1/profiles/save':
             try:return self.respond(self.server.profiles.save(self.read_json()))
             except ValueError as exc:return self.respond({'error':str(exc)},400)
@@ -760,7 +794,7 @@ def main():
         server.video.close()
         server.endpoint.close()
         server.stt.stop()
-        server.tts_tests.stop()
+        server.voxcpm_tests.stop();server.tts_tests.stop()
         server.tts_runtime.stop();server.voxcpm_runtime.stop()
         server.image_runtime.stop()
         server.ltx_original_runtime.stop()