michael/AI-Profile-Router
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Remove Beta 1 and Piper fallback

Browse Source
This commit is contained in:
Mikei386
2026-09-10 13:17:57 +02:00
parent 52482627be
commit 08ff3d7c4e
38 changed files with 94 additions and 484 deletions
Download Patch File Download Diff File Expand all files Collapse all files
router/ai_profile_router.py
+11 -12
View File
@@ -2,21 +2,20 @@
"""AI Profile Router – OpenAI-kompatibler Proxy vor llama.cpp.
Leitet OpenAI-kompatible Requests transparent an den lokalen llama.cpp-Server
weiter (Streaming, Tool Calls, JSON) und schaltet zwischen sechs festen
weiter (Streaming, Tool Calls, JSON) und schaltet zwischen fünf festen
Profilen um:
Profil Kontext
------ --------
fast 76800
medium 160000
beta1 192000
large 192000
ultra 262144
uncensored 80000
Virtuelle Modelle: qwen-fast, qwen-medium, qwen-beta-1, qwen-large,
Virtuelle Modelle: qwen-fast, qwen-medium, qwen-large,
qwen-ultra, qwen-uncensored
Kommandos: POST /fast, /medium, /beta1, /large, /ultra,
Kommandos: POST /fast, /medium, /large, /ultra,
/uncensored
GET /status (Zustand)
@@ -26,7 +25,7 @@ Bildgenerierung und Editing (FLUX.2 Klein 9B FP8 beta):
GET /images (Liste)
GET /images/<datei> (PNG-Download)
Sprachausgabe (Qwen3-TTS auf RTX 3060, Piper als CPU-Fallback):
Sprachausgabe (Qwen3-TTS auf RTX 3060):
POST /v1/audio/speech (OpenAI-kompatibel)
GET /v1/audio/voices (verfügbare Stimmen)
@@ -177,15 +176,15 @@ IMAGE_QUALITY = {"standard": 4, "high": 4}
IMAGE_DEFAULT_QUALITY = "standard"
IMAGE_MAX_N = 4
# --- Sprachausgabe (austauschbarer interner TTS-Worker, CPU-only) ---
# --- Sprachausgabe (Qwen3-TTS über das interne Normalisierungs-Gateway) ---
TTS_WORKER_URL = os.environ.get("TTS_WORKER_URL", "http://127.0.0.1:8085")
TTS_TIMEOUT = float(os.environ.get("TTS_TIMEOUT", "300")) # s, pro Synthese
TTS_CONNECT_TIMEOUT = float(os.environ.get("TTS_CONNECT_TIMEOUT", "5"))
TTS_MODEL = os.environ.get("TTS_MODEL", "xtts-v2")
TTS_MODEL = os.environ.get("TTS_MODEL", "qwen3-tts")
TTS_VOICES = tuple(v.strip() for v in os.environ.get(
"TTS_VOICES", "claribel").split(",") if v.strip())
"TTS_VOICES", "alloy").split(",") if v.strip())
TTS_DEFAULT_VOICE = os.environ.get(
"TTS_DEFAULT_VOICE", TTS_VOICES[0] if TTS_VOICES else "claribel")
"TTS_DEFAULT_VOICE", TTS_VOICES[0] if TTS_VOICES else "alloy")
TTS_FORMATS = ("mp3", "wav", "pcm")
TTS_DEFAULT_FORMAT = "mp3"
@@ -2363,7 +2362,7 @@ class Handler(BaseHTTPRequestHandler):
self.end_headers()
self.wfile.write(data)
# ---------- Sprachausgabe (Qwen3-TTS mit Piper-Fallback) ----------
# ---------- Sprachausgabe (Qwen3-TTS) ----------
def _speech(self) -> None:
try:
@@ -2422,7 +2421,7 @@ class Handler(BaseHTTPRequestHandler):
"invalid_request_error", "invalid_speed")
return
# Modell-Name optional; falls angegeben, muss es xtts-v2 sein.
# Modell-Name optional; falls angegeben, muss es Qwen3-TTS sein.
model = data.get("model")
if model is not None and model != TTS_MODEL:
self._send_error(400, f"unbekanntes Modell: {model!r} "
@@ -2532,7 +2531,7 @@ class Handler(BaseHTTPRequestHandler):
models.append({
"id": TTS_MODEL,
"object": "model",
"owned_by": "coqui-xtts",
"owned_by": "qwen",
"type": "speech",
})
return {"object": "list", "data": models}