feat(voxtral): STT-3B via Transformers (torch cu124, kein Treiber-Upgrade)
Backports brachte keinen neueren Treiber (bleibt 550/CUDA 12.4). Statt Upgrade: Voxtral-Mini-3B-2507 via Transformers mit torch 2.6.0+cu124 (F5-Trick) — laeuft auf 550, ~9 GB bf16 auf GPU 1. Chunked wie whisper mit dem adaptiven M0.1-Endpointer, RVS-Protokoll identisch (drop-in). Ersetzt den vLLM-Realtime-4B-Ansatz (der brauchte 16 GB). Compose: ein Container, GPU-1-gepinnt, Profil 'voxtral'. Transformers-API in einer Methode gekapselt (verify-on-run). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
Claude Opus 4.8
+12
-35
@@ -138,55 +138,32 @@ services:
|
||||
- LLM_TIMEOUT_SEC=${LLM_TIMEOUT_SEC:-600}
|
||||
restart: unless-stopped
|
||||
|
||||
# ─── Voxtral STT (GPU, Realtime) — PROFIL "voxtral" ───────────
|
||||
# Ersetzt whisper als STT sobald die 24-GB-Karte da ist. Startet NUR mit
|
||||
# docker compose --profile voxtral up -d
|
||||
# (sonst kollidiert es mit whisper — beide wuerden stt_* beantworten).
|
||||
#
|
||||
# ⚠️ VRAM: Voxtral-Mini-4B-Realtime-2602 braucht >=16 GB (BF16, laut vLLM-
|
||||
# Rezept keine Quant). Laeuft NICHT auf der 3060 (12 GB) — erst 24-GB-Karte.
|
||||
# ⚠️ vLLM: Version >=0.20.0 noetig. Entrypoint/Serve-Form beim ersten Lauf
|
||||
# gegen das offizielle Rezept pruefen (siehe voxtral/README.md).
|
||||
voxtral-vllm:
|
||||
image: vllm/vllm-openai:latest
|
||||
container_name: aria-voxtral-vllm
|
||||
# ─── Voxtral STT-3B (Transformers, GPU) — PROFIL "voxtral" ─────
|
||||
# Ersetzt whisper als STT. Laeuft auf Treiber 550/CUDA 12.4 (torch cu124, KEIN
|
||||
# Treiber-Upgrade noetig). Modell Voxtral-Mini-3B-2507 (~9 GB bf16) → GPU 1.
|
||||
# Startet NUR mit: docker compose --profile voxtral up -d --build
|
||||
# Vorher whisper stoppen, sonst beantworten beide stt_* (Kollision):
|
||||
# docker compose stop whisper-bridge
|
||||
voxtral-bridge:
|
||||
build: ./voxtral
|
||||
container_name: aria-voxtral-bridge
|
||||
profiles: ["voxtral"]
|
||||
deploy:
|
||||
resources:
|
||||
reservations:
|
||||
devices:
|
||||
- driver: nvidia
|
||||
count: 1
|
||||
device_ids: ["1"] # 12-GB-Karte (STT-3B ~9 GB); GPU 0 (8 GB) bleibt fuer F5/LLM
|
||||
capabilities: [gpu]
|
||||
volumes:
|
||||
- ./hf-cache:/root/.cache/huggingface # gleicher Modell-Cache wie whisper/f5
|
||||
environment:
|
||||
- VLLM_DISABLE_COMPILE_CACHE=1
|
||||
- HUGGING_FACE_HUB_TOKEN=${HF_TOKEN:-}
|
||||
# Serve-Command aus dem vLLM-Rezept (Voxtral-Mini-4B-Realtime-2602).
|
||||
command:
|
||||
- --model
|
||||
- mistralai/Voxtral-Mini-4B-Realtime-2602
|
||||
- --tokenizer-mode
|
||||
- mistral
|
||||
- --compilation_config
|
||||
- '{"cudagraph_mode":"PIECEWISE"}'
|
||||
restart: unless-stopped
|
||||
|
||||
# ─── Voxtral-Bridge — RVS <-> vLLM-Realtime-WS (CPU-Glue) ─────
|
||||
voxtral-bridge:
|
||||
build: ./voxtral
|
||||
container_name: aria-voxtral-bridge
|
||||
profiles: ["voxtral"]
|
||||
depends_on:
|
||||
- voxtral-vllm
|
||||
environment:
|
||||
- RVS_HOST=${RVS_HOST}
|
||||
- RVS_PORT=${RVS_PORT:-443}
|
||||
- RVS_TLS=${RVS_TLS:-true}
|
||||
- RVS_TLS_FALLBACK=${RVS_TLS_FALLBACK:-true}
|
||||
- RVS_TOKEN=${RVS_TOKEN}
|
||||
- VOXTRAL_VLLM_URL=ws://voxtral-vllm:8000/v1/realtime
|
||||
- VOXTRAL_MODEL=mistralai/Voxtral-Mini-4B-Realtime-2602
|
||||
- VOXTRAL_MODEL=mistralai/Voxtral-Mini-3B-2507
|
||||
- VOXTRAL_LANGUAGE=${WHISPER_LANGUAGE:-de}
|
||||
- HUGGING_FACE_HUB_TOKEN=${HF_TOKEN:-} # falls das Modell HF-gated ist
|
||||
restart: unless-stopped
|
||||
|
||||
+19
-7
@@ -1,14 +1,26 @@
|
||||
# Voxtral-BRIDGE (nicht das Modell!) — leichte CPU-Glue zwischen RVS und dem
|
||||
# vLLM-Realtime-Server. Das eigentliche Voxtral-Modell laeuft im Container
|
||||
# `voxtral-vllm` (GPU, vllm/vllm-openai). Deshalb hier kein CUDA-Base noetig.
|
||||
FROM python:3.11-slim
|
||||
# Voxtral-STT-3B Bridge (Transformers). Laeuft auf Treiber 550/CUDA 12.4 via
|
||||
# torch cu124 — KEIN Treiber-Upgrade noetig (gleicher Trick wie f5tts).
|
||||
# Modell: Voxtral-Mini-3B-2507 (~9 GB in bf16) → passt auf die 12-GB-Karte (GPU 1).
|
||||
FROM nvidia/cuda:12.2.2-cudnn8-runtime-ubuntu22.04
|
||||
|
||||
ENV DEBIAN_FRONTEND=noninteractive
|
||||
ENV PYTHONUNBUFFERED=1
|
||||
WORKDIR /app
|
||||
|
||||
COPY requirements.txt .
|
||||
RUN pip install --no-cache-dir -r requirements.txt
|
||||
RUN apt-get update && apt-get install -y --no-install-recommends \
|
||||
python3 python3-pip ffmpeg git \
|
||||
&& rm -rf /var/lib/apt/lists/*
|
||||
|
||||
COPY bridge.py ./
|
||||
# torch FEST auf cu124 (Treiber 550 = CUDA 12.4). 2.6.0 ist der neueste cu124-Build;
|
||||
# torch 2.7+ gibt es nur fuer cu126+ und braeuchte einen neueren Treiber.
|
||||
RUN pip3 install --no-cache-dir torch==2.6.0 torchaudio==2.6.0 \
|
||||
--index-url https://download.pytorch.org/whl/cu124
|
||||
|
||||
COPY requirements.txt .
|
||||
# Constraint haelt transformers/mistral-common davon ab, torch wieder hochzuziehen.
|
||||
RUN printf 'torch==2.6.0\ntorchaudio==2.6.0\n' > /tmp/torch-constraint.txt && \
|
||||
pip3 install --no-cache-dir -c /tmp/torch-constraint.txt -r requirements.txt
|
||||
|
||||
COPY bridge.py .
|
||||
|
||||
CMD ["python3", "bridge.py"]
|
||||
|
||||
+31
-54
@@ -1,70 +1,47 @@
|
||||
# Voxtral-STT-Satellit (M0.3)
|
||||
# Voxtral-STT-3B-Satellit (Transformers)
|
||||
|
||||
Streaming-STT via **Voxtral-Mini-4B-Realtime-2602** (Mistral, Apache 2.0) auf
|
||||
vLLM. Ersetzt whisper als STT — genauer (~5,9 % WER vs 7,4 % FLEURS) und mit
|
||||
echtem Realtime-Streaming. Deutsch ist in den 13 Sprachen abgedeckt.
|
||||
Streaming-STT via **Voxtral-Mini-3B-2507** (Mistral, Apache 2.0) über
|
||||
🤗 Transformers. Ersetzt whisper als STT — genauer, und **ohne Treiber-Upgrade**:
|
||||
läuft auf dem Trixie-Standardtreiber (550/CUDA 12.4) via **torch 2.6.0+cu124**
|
||||
(derselbe Trick wie bei f5tts).
|
||||
|
||||
Zwei Container:
|
||||
- **`voxtral-vllm`** — das Modell auf vLLM (GPU). Exponiert die Realtime-WS-API.
|
||||
- **`voxtral-bridge`** — CPU-Glue: RVS ⇄ vLLM-Realtime-WS. Macht das Endpointing
|
||||
selbst (adaptiver Rausch-Boden-VAD, identisch zur whisper-Bridge / M0.1).
|
||||
- Modell ~9 GB (bf16) → **GPU 1** (die 12-GB-Karte; per Compose gepinnt).
|
||||
- **F5-TTS + LLM** bleiben auf GPU 0 (8 GB).
|
||||
- Arbeitsweise = chunked wie whisper: live PCM → alle ~1 s transkribieren
|
||||
(Partials) → **adaptiver Endpointer** (Rausch-Boden-VAD + semantische
|
||||
Stagnation, aus M0.1) feuert `stt_endpoint`. RVS-Protokoll identisch → drop-in.
|
||||
|
||||
## ⚠️ Hardware-Realität — läuft NICHT auf der 3060
|
||||
|
||||
Das Realtime-Modell braucht laut [vLLM-Rezept](https://recipes.vllm.ai/mistralai/Voxtral-Mini-4B-Realtime-2602)
|
||||
**≥ 16 GB VRAM (BF16, keine Quant)**. Die RTX 3060 hat 12 GB → passt nicht.
|
||||
|
||||
- **Interim-gpubox (nur 3060):** whisper (mit M0.1-Fix) + F5-TTS bleiben aktiv.
|
||||
Voxtral NICHT starten.
|
||||
- **Ab der 24-GB-Karte:** Voxtral-Profil hochziehen, whisper wird Fallback.
|
||||
|
||||
Deshalb liegen beide Services hinter dem Compose-**Profil `voxtral`** und starten
|
||||
NUR explizit — sonst würden whisper *und* voxtral dieselben `stt_*`-Messages
|
||||
beantworten (Kollision).
|
||||
|
||||
## Starten (erst wenn die 24-GB-Karte drin ist)
|
||||
## Starten (Profil `voxtral`)
|
||||
|
||||
```bash
|
||||
cd xtts
|
||||
docker compose stop whisper-bridge # sonst beantworten beide stt_*
|
||||
docker compose --profile voxtral up -d --build
|
||||
docker logs -f aria-voxtral-vllm # laedt Modell (mehrere GB, dauert)
|
||||
docker logs -f aria-voxtral-bridge # "RVS verbunden" + service_status ready
|
||||
```
|
||||
Whisper vorher stoppen, damit nur eine STT-Engine antwortet:
|
||||
```bash
|
||||
docker compose stop whisper-bridge
|
||||
docker logs -f aria-voxtral-bridge # Modell laedt (mehrere GB), dann "RVS verbunden"
|
||||
```
|
||||
Zurück zu whisper: `docker compose --profile voxtral down && docker compose up -d whisper-bridge`.
|
||||
|
||||
## ⚠️ Auf echter Hardware verifizieren (blind gebaut, kein Test hier)
|
||||
## ⚠️ Auf echter Hardware verifizieren (blind gebaut)
|
||||
|
||||
1. **vLLM-Version ≥ 0.20.0** und die **Serve-Form**. Das Rezept nutzt
|
||||
`vllm serve <model> …`. Falls das `vllm/vllm-openai`-Image einen anderen
|
||||
Entrypoint hat, das `command:` in `docker-compose.yml` anpassen
|
||||
(Rezept-Command steht dort als Kommentar).
|
||||
2. **Realtime-WS-Frames.** Die exakten Event-Namen sind in `bridge.py` ganz oben
|
||||
als Konstanten gebündelt (`VLLM_SEND_APPEND`, `VLLM_DELTA_SUFFIXES`, …),
|
||||
modelliert nach dem OpenAI-Realtime-Schema. Gegen das offizielle
|
||||
**vLLM-Realtime-Client-Beispiel** prüfen und dort anpassen — nur an dieser
|
||||
einen Stelle. Das Response-Handling ist bereits defensiv (mehrere Feldnamen).
|
||||
3. **Endpoint-URL/Port.** Default `ws://voxtral-vllm:8000/v1/realtime` — prüfen ob
|
||||
vLLM auf 8000 lauscht und `/v1/realtime` registriert (Log-Zeile
|
||||
`Route: /v1/realtime`).
|
||||
4. **Endpointing.** Voxtral liefert keine eigene VAD → unser adaptiver Endpointer
|
||||
entscheidet (akustisch + semantisch am Delta-Wachstum). `endpointMs` kommt wie
|
||||
bei whisper aus der App; `voiceFactor` per Session tunebar.
|
||||
1. **Transformers-API.** Die exakte Voxtral-Transkriptions-API ist in `bridge.py`
|
||||
in **einer** Methode gekapselt (`VoxtralRunner._transcribe_blocking`), modelliert
|
||||
nach der HF-Modelcard (`apply_transcription_request` → `generate` → `batch_decode`).
|
||||
Beim ersten Lauf gegen die Modelcard prüfen und dort anpassen.
|
||||
2. **HF-Gating.** Ist `Voxtral-Mini-3B-2507` gated, `HF_TOKEN` in `xtts/.env` setzen
|
||||
(wird als `HUGGING_FACE_HUB_TOKEN` durchgereicht).
|
||||
3. **VRAM/Tempo.** 3B in bf16 ~9 GB auf der 12-GB-Karte — Rest fürs KV-Cache. Ist die
|
||||
Partial-Transkription (alle 1 s) zu schwer, `STREAM_TRANSCRIBE_INTERVAL_MS` hochsetzen.
|
||||
4. **torch-Konflikt.** Falls `transformers`/`mistral-common` beim Build torch>2.6
|
||||
erzwingen, meldet der Constraint einen Konflikt → dann brauchen wir doch das
|
||||
Treiber-Upgrade (`bootstrap.sh --upgrade-driver` via NVIDIA-CUDA-Repo) + cu126-torch.
|
||||
|
||||
## Protokoll (RVS, identisch zu whisper — drop-in)
|
||||
|
||||
Rein: `stt_stream_start`, `stt_audio_chunk` (16 kHz mono s16le, base64), `stt_stream_end`.
|
||||
Raus: `stt_partial`, `stt_endpoint` (das Event, auf das aria-bridge horcht), `stt_stream_done`.
|
||||
Raus: `stt_partial`, `stt_endpoint`, `stt_stream_done`.
|
||||
|
||||
## TTS-Hinweis
|
||||
|
||||
Voxtral-**TTS** (Voice-Cloning) ist hier NICHT enthalten — braucht ebenfalls
|
||||
16 GB VRAM und ist ein eigener Bau. Bis zur 24-GB-Karte bleibt **F5-TTS** aktiv.
|
||||
Danach: eigener `voxtral-tts`-Satellit (separates Ticket).
|
||||
## TTS
|
||||
Bleibt **F5-TTS** (klingt gut, passt auf GPU 0). Voxtral-TTS bräuchte ~24 GB — separates Thema.
|
||||
|
||||
## Quellen
|
||||
- Rezept: https://recipes.vllm.ai/mistralai/Voxtral-Mini-4B-Realtime-2602
|
||||
- vLLM Speech-to-Text: https://docs.vllm.ai/en/latest/serving/online_serving/speech_to_text/
|
||||
- Modell: https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-2602
|
||||
- Modell: https://huggingface.co/mistralai/Voxtral-Mini-3B-2507
|
||||
- Transformers-Nutzung: HF-Modelcard (Voxtral) + `mistral-common`
|
||||
|
||||
+147
-231
@@ -1,39 +1,28 @@
|
||||
#!/usr/bin/env python3
|
||||
"""
|
||||
ARIA Voxtral Bridge — Streaming-STT via Voxtral-Mini-4B-Realtime-2602 (vLLM).
|
||||
ARIA Voxtral-STT-3B Bridge (Transformers) — Ersatz fuer whisper.
|
||||
|
||||
Zwilling der whisper-Bridge, aber das Transkribieren macht NICHT faster-whisper
|
||||
im selben Prozess, sondern der separate vLLM-Realtime-Server (Container
|
||||
`voxtral-vllm`) ueber dessen WebSocket-API `/v1/realtime`. Diese Bridge ist
|
||||
reine Glue:
|
||||
Laeuft auf Treiber 550/CUDA 12.4 via torch cu124 (kein Treiber-Upgrade noetig).
|
||||
Modell: Voxtral-Mini-3B-2507 (bf16, ~9 GB) → GPU 1 (12 GB, per Compose gepinnt).
|
||||
|
||||
App ──(RVS: stt_stream_start / stt_audio_chunk / stt_stream_end)──▶ diese Bridge
|
||||
diese Bridge ──(WS /v1/realtime: PCM16-b64 append)──▶ voxtral-vllm
|
||||
voxtral-vllm ──(transcription.delta / transcription.done)──▶ diese Bridge
|
||||
diese Bridge ──(RVS: stt_partial / stt_endpoint / stt_stream_done)──▶ App/aria-bridge
|
||||
Arbeitsweise = Zwilling der whisper-Bridge: App schickt live PCM-Chunks; wir
|
||||
transkribieren alle ~STREAM_TRANSCRIBE_INTERVAL_MS auf dem Ringbuffer (Partials)
|
||||
und feuern stt_endpoint, sobald der ADAPTIVE Endpointer (Rausch-Boden-VAD +
|
||||
semantische Stagnation, aus M0.1) "fertig" sagt. RVS-Wire-Protokoll identisch zu
|
||||
whisper → drop-in (die App merkt nur bessere Genauigkeit).
|
||||
|
||||
Das RVS-Wire-Protokoll ist IDENTISCH zur whisper-Bridge (drop-in). Das
|
||||
Endpointing (wann hat der User aufgehoert zu sprechen) macht diese Bridge
|
||||
selbst — mit demselben ADAPTIVEN Rausch-Boden-Endpointer wie whisper (Voxtral
|
||||
Realtime liefert laut vLLM-Doku keine eigene VAD/„speaker done"-Semantik, nur
|
||||
transcription.delta/.done). Die akustische Energie messen wir auf unserer
|
||||
eigenen PCM-Kopie, die semantische Stagnation am Delta-Textwachstum.
|
||||
|
||||
⚠️ HARDWARE: Voxtral-Mini-4B-Realtime-2602 braucht >=16 GB VRAM (BF16). Auf der
|
||||
RTX 3060 (12 GB) laeuft es NICHT — erst auf der 24-GB-Karte. Bis dahin
|
||||
bleibt die whisper-Bridge aktiv (Profil-gesteuert im docker-compose).
|
||||
|
||||
⚠️ VERIFY-ON-FIRST-RUN: Die exakten vLLM-Realtime-FRAME-Namen (Audio-Append,
|
||||
Delta/Done-Event-Typen) sind unten als Konstanten gebuendelt und nach dem
|
||||
OpenAI-Realtime-Schema modelliert. Gegen das offizielle vLLM-Realtime-
|
||||
Client-Beispiel pruefen und ggf. anpassen — sie stehen bewusst an EINER
|
||||
Stelle. Response-Handling ist defensiv (mehrere moegliche Feldnamen).
|
||||
⚠️ VERIFY-ON-FIRST-RUN: Die exakte Transformers-Transkriptions-API von Voxtral
|
||||
(apply_transcription_request / generate / decode) ist unten in EINER Methode
|
||||
(VoxtralRunner._transcribe_blocking) gekapselt und nach dem HF-Modelcard-Muster
|
||||
modelliert. Beim ersten echten Lauf gegen die Voxtral-Modelcard pruefen und dort
|
||||
anpassen. Alles andere (RVS, Endpointer) ist bewaehrt.
|
||||
|
||||
Env:
|
||||
RVS_HOST, RVS_PORT, RVS_TLS, RVS_TLS_FALLBACK, RVS_TOKEN
|
||||
VOXTRAL_VLLM_URL Default: ws://voxtral-vllm:8000/v1/realtime
|
||||
VOXTRAL_MODEL Default: mistralai/Voxtral-Mini-4B-Realtime-2602
|
||||
VOXTRAL_LANGUAGE Default: de
|
||||
VOXTRAL_MODEL Default: mistralai/Voxtral-Mini-3B-2507
|
||||
VOXTRAL_LANGUAGE Default: de
|
||||
VOXTRAL_DEVICE Default: cuda
|
||||
STREAM_TRANSCRIBE_INTERVAL_MS Default 1000 (3B ist schwerer als whisper-small)
|
||||
"""
|
||||
import asyncio
|
||||
import base64
|
||||
@@ -60,29 +49,18 @@ RVS_TLS = os.getenv("RVS_TLS", "true").lower() == "true"
|
||||
RVS_TLS_FALLBACK = os.getenv("RVS_TLS_FALLBACK", "true").lower() == "true"
|
||||
RVS_TOKEN = os.getenv("RVS_TOKEN", "").strip()
|
||||
|
||||
VOXTRAL_VLLM_URL = os.getenv("VOXTRAL_VLLM_URL", "ws://voxtral-vllm:8000/v1/realtime")
|
||||
VOXTRAL_MODEL = os.getenv("VOXTRAL_MODEL", "mistralai/Voxtral-Mini-4B-Realtime-2602")
|
||||
VOXTRAL_MODEL = os.getenv("VOXTRAL_MODEL", "mistralai/Voxtral-Mini-3B-2507")
|
||||
VOXTRAL_LANGUAGE = os.getenv("VOXTRAL_LANGUAGE", "de")
|
||||
VOXTRAL_DEVICE = os.getenv("VOXTRAL_DEVICE", "cuda")
|
||||
|
||||
# ── vLLM-Realtime-Frames — HIER anpassen falls das Client-Beispiel abweicht ──
|
||||
# Senderichtung (wir → vLLM): PCM16-16kHz-mono base64 anhaengen + committen.
|
||||
VLLM_SEND_APPEND = "input_audio_buffer.append" # {"type":..., "audio": "<b64>"}
|
||||
VLLM_SEND_COMMIT = "input_audio_buffer.commit" # Buffer abschliessen
|
||||
VLLM_AUDIO_FIELD = "audio"
|
||||
# Empfangsrichtung (vLLM → wir): inkrementeller Text + final. Defensiv geprueft.
|
||||
VLLM_DELTA_SUFFIXES = ("transcription.delta",) # msg["type"] endet hierauf
|
||||
VLLM_DONE_SUFFIXES = ("transcription.done", "transcription.completed")
|
||||
VLLM_DELTA_FIELDS = ("delta", "text", "transcription") # eins davon traegt den Text
|
||||
|
||||
# ── Streaming-/Endpointing-Parameter (analog whisper-Bridge) ──
|
||||
STREAM_TRANSCRIBE_INTERVAL_MS = int(os.getenv("STREAM_TRANSCRIBE_INTERVAL_MS", "1000"))
|
||||
STREAM_DEFAULT_ENDPOINT_MS = 2400
|
||||
STREAM_DEFAULT_HARD_CAP_MS = 60000
|
||||
STREAM_DEFAULT_HARD_CAP_MS = 300000
|
||||
STREAM_MIN_AUDIO_MS = 600
|
||||
STREAM_SESSION_TTL_S = 120
|
||||
STREAM_ENERGY_WINDOW_MS = 300
|
||||
STREAM_SEMANTIC_BACKUP_FACTOR = 2.0
|
||||
# Adaptiver Voice-Schwellwert (siehe whisper-Bridge M0.1): Grenze relativ zum
|
||||
# gemessenen Rausch-Boden statt fix — schneidet leises Sprechen nicht ab.
|
||||
# Adaptiver Voice-Schwellwert (M0.1): relativ zum gemessenen Rausch-Boden.
|
||||
STREAM_VOICE_FACTOR = 2.5
|
||||
STREAM_VOICE_RMS_MIN = 0.005
|
||||
STREAM_VOICE_RMS_MAX = 0.020
|
||||
@@ -91,21 +69,60 @@ STREAM_VOICE_RMS_MAX = 0.020
|
||||
def pcm_s16le_to_float32(data: bytes) -> np.ndarray:
|
||||
if not data:
|
||||
return np.zeros(0, dtype=np.float32)
|
||||
arr = np.frombuffer(data, dtype=np.int16).astype(np.float32) / 32768.0
|
||||
return arr
|
||||
return np.frombuffer(data, dtype=np.int16).astype(np.float32) / 32768.0
|
||||
|
||||
|
||||
async def _send(ws, mtype: str, payload: dict) -> None:
|
||||
try:
|
||||
await ws.send(json.dumps({
|
||||
"type": mtype,
|
||||
"payload": payload,
|
||||
"timestamp": int(time.time() * 1000),
|
||||
"type": mtype, "payload": payload, "timestamp": int(time.time() * 1000),
|
||||
}))
|
||||
except Exception as e:
|
||||
logger.warning("RVS-Send fehlgeschlagen (%s): %s", mtype, e)
|
||||
|
||||
|
||||
class VoxtralRunner:
|
||||
"""Haelt das Voxtral-Modell (Transformers). transcribe() blockiert → aus dem
|
||||
Event-Loop via run_in_executor aufrufen. Ein Lock serialisiert GPU-Zugriffe."""
|
||||
|
||||
def __init__(self) -> None:
|
||||
self.model = None
|
||||
self.processor = None
|
||||
self._lock = asyncio.Lock()
|
||||
|
||||
def load(self) -> None:
|
||||
import torch
|
||||
from transformers import AutoProcessor, VoxtralForConditionalGeneration
|
||||
t0 = time.time()
|
||||
logger.info("Lade Voxtral '%s' (device=%s, bf16)…", VOXTRAL_MODEL, VOXTRAL_DEVICE)
|
||||
self.processor = AutoProcessor.from_pretrained(VOXTRAL_MODEL)
|
||||
self.model = VoxtralForConditionalGeneration.from_pretrained(
|
||||
VOXTRAL_MODEL, torch_dtype=torch.bfloat16, device_map=VOXTRAL_DEVICE,
|
||||
)
|
||||
logger.info("Voxtral geladen in %.1fs", time.time() - t0)
|
||||
|
||||
def _transcribe_blocking(self, audio_f32: np.ndarray, language: str) -> str:
|
||||
# ⚠️ VERIFY: exakte Voxtral-Transformers-API gegen die HF-Modelcard.
|
||||
import torch
|
||||
proc, model = self.processor, self.model
|
||||
if proc is None or model is None or audio_f32.size == 0:
|
||||
return ""
|
||||
inputs = proc.apply_transcription_request(
|
||||
language=language, audio=audio_f32, model_id=VOXTRAL_MODEL, sampling_rate=16000,
|
||||
)
|
||||
inputs = inputs.to(VOXTRAL_DEVICE, dtype=torch.bfloat16)
|
||||
with torch.no_grad():
|
||||
outputs = model.generate(**inputs, max_new_tokens=512)
|
||||
trimmed = outputs[:, inputs.input_ids.shape[1]:]
|
||||
text = proc.batch_decode(trimmed, skip_special_tokens=True)
|
||||
return (text[0] if text else "").strip()
|
||||
|
||||
async def transcribe(self, audio_f32: np.ndarray, language: str) -> str:
|
||||
loop = asyncio.get_running_loop()
|
||||
async with self._lock:
|
||||
return await loop.run_in_executor(None, self._transcribe_blocking, audio_f32, language)
|
||||
|
||||
|
||||
@dataclass
|
||||
class StreamSession:
|
||||
request_id: str
|
||||
@@ -126,28 +143,26 @@ class StreamSession:
|
||||
last_chunk_at: float = field(default_factory=time.time)
|
||||
last_partial: str = ""
|
||||
last_growth_at: float = 0.0
|
||||
last_transcribe_at: float = 0.0
|
||||
last_voice_at: float = 0.0
|
||||
noise_floor: float = 0.0
|
||||
closed: bool = False
|
||||
endpoint_sent: bool = False
|
||||
# vLLM-Realtime-Session
|
||||
vllm_ws: object = None
|
||||
vllm_reader: object = None
|
||||
|
||||
|
||||
class SessionManager:
|
||||
def __init__(self) -> None:
|
||||
def __init__(self, runner: VoxtralRunner) -> None:
|
||||
self.runner = runner
|
||||
self._sessions: dict[str, StreamSession] = {}
|
||||
self._ws = None # RVS
|
||||
self._ws = None
|
||||
|
||||
def attach_ws(self, ws) -> None:
|
||||
self._ws = ws
|
||||
|
||||
async def start_session(self, payload: dict) -> Optional[StreamSession]:
|
||||
request_id = (payload.get("requestId") or "").strip()
|
||||
if not request_id:
|
||||
logger.warning("stt_stream_start ohne requestId — ignoriert")
|
||||
return None
|
||||
def start_session(self, payload: dict) -> None:
|
||||
rid = (payload.get("requestId") or "").strip()
|
||||
if not rid:
|
||||
return
|
||||
try:
|
||||
endpoint_ms = int(payload.get("endpointMs") or STREAM_DEFAULT_ENDPOINT_MS)
|
||||
except (TypeError, ValueError):
|
||||
@@ -160,8 +175,8 @@ class SessionManager:
|
||||
voice_factor = float(payload.get("voiceFactor") or STREAM_VOICE_FACTOR)
|
||||
except (TypeError, ValueError):
|
||||
voice_factor = STREAM_VOICE_FACTOR
|
||||
sess = StreamSession(
|
||||
request_id=request_id,
|
||||
self._sessions[rid] = StreamSession(
|
||||
request_id=rid,
|
||||
audio_request_id=payload.get("audioRequestId", "") or "",
|
||||
language=payload.get("language") or VOXTRAL_LANGUAGE,
|
||||
endpoint_ms=endpoint_ms,
|
||||
@@ -173,159 +188,44 @@ class SessionManager:
|
||||
location=payload.get("location") or None,
|
||||
sample_rate=int(payload.get("sampleRate") or 16000),
|
||||
)
|
||||
# vLLM-Realtime-Session oeffnen + Reader starten.
|
||||
try:
|
||||
sess.vllm_ws = await websockets.connect(VOXTRAL_VLLM_URL, max_size=8 * 1024 * 1024)
|
||||
await self._vllm_configure(sess)
|
||||
sess.vllm_reader = asyncio.create_task(self._vllm_read_loop(sess))
|
||||
except Exception as e:
|
||||
logger.exception("Stream %s: vLLM-Realtime-Connect fehlgeschlagen: %s",
|
||||
request_id[:8], e)
|
||||
# ohne Backend keine Transkription → sofort leeres Endpoint melden
|
||||
self._sessions[request_id] = sess
|
||||
await self._finalize(sess, reason="vllm_unavailable")
|
||||
return None
|
||||
self._sessions[request_id] = sess
|
||||
logger.info("Voxtral-Session offen: id=%s lang=%s endpointMs=%d",
|
||||
request_id[:8], sess.language, sess.endpoint_ms)
|
||||
return sess
|
||||
|
||||
async def _vllm_configure(self, sess: StreamSession) -> None:
|
||||
"""Optionale Session-Konfig an vLLM (Modell/Sprache/temperature=0).
|
||||
VERIFY: exaktes session.update-Schema gegen vLLM-Realtime-Beispiel.
|
||||
Best-effort — Fehler hier sind nicht fatal."""
|
||||
try:
|
||||
await sess.vllm_ws.send(json.dumps({
|
||||
"type": "session.update",
|
||||
"session": {
|
||||
"model": VOXTRAL_MODEL,
|
||||
"language": sess.language,
|
||||
"temperature": 0.0,
|
||||
"input_audio_format": "pcm16",
|
||||
},
|
||||
}))
|
||||
except Exception:
|
||||
pass
|
||||
|
||||
async def _vllm_read_loop(self, sess: StreamSession) -> None:
|
||||
"""Liest transcription.delta/.done vom vLLM-Realtime-Server."""
|
||||
ws = sess.vllm_ws
|
||||
try:
|
||||
async for raw in ws:
|
||||
try:
|
||||
msg = json.loads(raw)
|
||||
except Exception:
|
||||
continue
|
||||
mtype = str(msg.get("type", ""))
|
||||
if any(mtype.endswith(s) for s in VLLM_DELTA_SUFFIXES):
|
||||
text = self._extract_text(msg)
|
||||
if text:
|
||||
await self._on_delta(sess, text)
|
||||
elif any(mtype.endswith(s) for s in VLLM_DONE_SUFFIXES):
|
||||
text = self._extract_text(msg)
|
||||
if text:
|
||||
await self._on_delta(sess, text, final=True)
|
||||
except Exception:
|
||||
logger.debug("Stream %s: vLLM-Reader beendet", sess.request_id[:8])
|
||||
|
||||
@staticmethod
|
||||
def _extract_text(msg: dict) -> str:
|
||||
for f in VLLM_DELTA_FIELDS:
|
||||
v = msg.get(f)
|
||||
if isinstance(v, str) and v:
|
||||
return v
|
||||
return ""
|
||||
|
||||
async def _on_delta(self, sess: StreamSession, text: str, final: bool = False) -> None:
|
||||
"""Neuer/finaler Transkript-Text vom vLLM. delta = inkrementell; wir
|
||||
haengen an, wenn er den bisherigen Partial verlaengert, sonst ersetzen
|
||||
wir (Voxtral kann korrigieren)."""
|
||||
if final or text.startswith(sess.last_partial):
|
||||
new_full = text if final else text
|
||||
else:
|
||||
new_full = (sess.last_partial + text).strip()
|
||||
new_full = new_full.strip()
|
||||
if new_full and new_full != sess.last_partial:
|
||||
sess.last_partial = new_full
|
||||
sess.last_growth_at = time.time()
|
||||
if self._ws is not None:
|
||||
await _send(self._ws, "stt_partial", {
|
||||
"requestId": sess.request_id,
|
||||
"audioRequestId": sess.audio_request_id,
|
||||
"text": new_full,
|
||||
})
|
||||
rid[:8], self._sessions[rid].language, endpoint_ms)
|
||||
|
||||
def feed_chunk(self, payload: dict) -> bool:
|
||||
request_id = payload.get("requestId", "")
|
||||
sess = self._sessions.get(request_id)
|
||||
sess = self._sessions.get(payload.get("requestId", ""))
|
||||
if sess is None or sess.closed:
|
||||
return False
|
||||
pcm_b64 = payload.get("pcm", "")
|
||||
if not pcm_b64:
|
||||
return True
|
||||
try:
|
||||
pcm = base64.b64decode(pcm_b64)
|
||||
except Exception:
|
||||
return True
|
||||
sess.pcm_buffer.extend(pcm)
|
||||
if pcm_b64:
|
||||
try:
|
||||
sess.pcm_buffer.extend(base64.b64decode(pcm_b64))
|
||||
except Exception:
|
||||
pass
|
||||
sess.last_chunk_at = time.time()
|
||||
# An vLLM weiterreichen (fire-and-forget).
|
||||
if sess.vllm_ws is not None:
|
||||
asyncio.create_task(self._vllm_append(sess, pcm_b64))
|
||||
return True
|
||||
|
||||
async def _vllm_append(self, sess: StreamSession, pcm_b64: str) -> None:
|
||||
try:
|
||||
await sess.vllm_ws.send(json.dumps({
|
||||
"type": VLLM_SEND_APPEND,
|
||||
VLLM_AUDIO_FIELD: pcm_b64,
|
||||
}))
|
||||
except Exception:
|
||||
pass
|
||||
|
||||
def end_session(self, request_id: str) -> None:
|
||||
sess = self._sessions.get(request_id)
|
||||
if sess is not None:
|
||||
sess.closed = True
|
||||
|
||||
def drop(self, request_id: str) -> None:
|
||||
sess = self._sessions.pop(request_id, None)
|
||||
if sess is not None:
|
||||
self._teardown_vllm(sess)
|
||||
self._sessions.pop(request_id, None)
|
||||
|
||||
def _teardown_vllm(self, sess: StreamSession) -> None:
|
||||
try:
|
||||
if sess.vllm_reader is not None:
|
||||
sess.vllm_reader.cancel()
|
||||
except Exception:
|
||||
pass
|
||||
if sess.vllm_ws is not None:
|
||||
asyncio.create_task(self._close_ws(sess.vllm_ws))
|
||||
sess.vllm_ws = None
|
||||
|
||||
@staticmethod
|
||||
async def _close_ws(ws) -> None:
|
||||
try:
|
||||
await ws.close()
|
||||
except Exception:
|
||||
pass
|
||||
|
||||
# ── Endpointer (adaptiv, wie whisper-Bridge M0.1) ──
|
||||
def _buffer_duration_ms(self, sess: StreamSession) -> float:
|
||||
# ── Endpointer (adaptiv, M0.1) ──
|
||||
def _buffer_ms(self, sess: StreamSession) -> float:
|
||||
samples = len(sess.pcm_buffer) // 2
|
||||
return (samples / sess.sample_rate) * 1000.0 if samples else 0.0
|
||||
|
||||
def _tail_rms(self, sess: StreamSession) -> float:
|
||||
win_bytes = int(sess.sample_rate * STREAM_ENERGY_WINDOW_MS / 1000) * 2
|
||||
if win_bytes <= 0:
|
||||
win = int(sess.sample_rate * STREAM_ENERGY_WINDOW_MS / 1000) * 2
|
||||
if win <= 0:
|
||||
return 0.0
|
||||
tail = sess.pcm_buffer[-win_bytes:]
|
||||
tail = sess.pcm_buffer[-win:]
|
||||
if len(tail) < 2:
|
||||
return 0.0
|
||||
arr = pcm_s16le_to_float32(bytes(tail))
|
||||
if arr.size == 0:
|
||||
return 0.0
|
||||
return float(np.sqrt(np.mean(arr * arr)))
|
||||
return float(np.sqrt(np.mean(arr * arr))) if arr.size else 0.0
|
||||
|
||||
def _voice_threshold(self, sess: StreamSession) -> float:
|
||||
nf = sess.noise_floor
|
||||
@@ -343,7 +243,8 @@ class SessionManager:
|
||||
sess.noise_floor = 0.98 * nf + 0.02 * rms
|
||||
|
||||
async def run_endpointer(self) -> None:
|
||||
logger.info("Voxtral-Endpointer gestartet (adaptiver VAD)")
|
||||
logger.info("Voxtral-Endpointer gestartet (adaptiver VAD, interval=%dms)",
|
||||
STREAM_TRANSCRIBE_INTERVAL_MS)
|
||||
while True:
|
||||
await asyncio.sleep(0.2)
|
||||
now = time.time()
|
||||
@@ -351,7 +252,7 @@ class SessionManager:
|
||||
try:
|
||||
await self._tick(sess, now)
|
||||
except Exception:
|
||||
logger.exception("Endpointer-Tick crashed (session=%s)", sid[:8])
|
||||
logger.exception("Tick crashed (session=%s)", sid[:8])
|
||||
for sid, sess in list(self._sessions.items()):
|
||||
if now - sess.last_chunk_at > STREAM_SESSION_TTL_S:
|
||||
logger.info("Stream %s: TTL — drop", sid[:8])
|
||||
@@ -360,14 +261,13 @@ class SessionManager:
|
||||
async def _tick(self, sess: StreamSession, now: float) -> None:
|
||||
if sess.endpoint_sent:
|
||||
return
|
||||
elapsed_ms = (now - sess.started_at) * 1000.0
|
||||
if elapsed_ms > sess.hard_cap_ms and not sess.closed:
|
||||
await self._finalize(sess, reason="hardcap")
|
||||
if (now - sess.started_at) * 1000.0 > sess.hard_cap_ms and not sess.closed:
|
||||
await self._finalize(sess, "hardcap")
|
||||
return
|
||||
if sess.closed:
|
||||
await self._finalize(sess, reason="stream_end")
|
||||
await self._finalize(sess, "stream_end")
|
||||
return
|
||||
if self._buffer_duration_ms(sess) < STREAM_MIN_AUDIO_MS:
|
||||
if self._buffer_ms(sess) < STREAM_MIN_AUDIO_MS:
|
||||
return
|
||||
# adaptive akustische Sprach-Aktivitaet
|
||||
rms = self._tail_rms(sess)
|
||||
@@ -375,45 +275,65 @@ class SessionManager:
|
||||
sess.last_voice_at = now
|
||||
else:
|
||||
self._update_noise_floor(sess, rms)
|
||||
# Endpoint: akustisch (Primaer) oder semantisch (Backstop), sobald Text da
|
||||
# Endpoint-Entscheidung, sobald Text erkannt wurde
|
||||
if sess.last_growth_at > 0.0:
|
||||
acoustic_silence_ms = (now - sess.last_voice_at) * 1000.0 if sess.last_voice_at > 0 else 0.0
|
||||
semantic_silence_ms = (now - sess.last_growth_at) * 1000.0
|
||||
acoustic_done = sess.last_voice_at > 0 and acoustic_silence_ms >= sess.endpoint_ms
|
||||
semantic_done = semantic_silence_ms >= sess.endpoint_ms * STREAM_SEMANTIC_BACKUP_FACTOR
|
||||
if acoustic_done or semantic_done:
|
||||
await self._finalize(sess, reason="endpoint" if acoustic_done else "endpoint_semantic")
|
||||
ac_sil = (now - sess.last_voice_at) * 1000.0 if sess.last_voice_at > 0 else 0.0
|
||||
se_sil = (now - sess.last_growth_at) * 1000.0
|
||||
ac_done = sess.last_voice_at > 0 and ac_sil >= sess.endpoint_ms
|
||||
se_done = se_sil >= sess.endpoint_ms * STREAM_SEMANTIC_BACKUP_FACTOR
|
||||
if ac_done or se_done:
|
||||
await self._finalize(sess, "endpoint" if ac_done else "endpoint_semantic")
|
||||
return
|
||||
# Partial-Transkription (throttled)
|
||||
if (now - sess.last_transcribe_at) * 1000.0 < STREAM_TRANSCRIBE_INTERVAL_MS:
|
||||
return
|
||||
sess.last_transcribe_at = now
|
||||
audio = pcm_s16le_to_float32(bytes(sess.pcm_buffer))
|
||||
try:
|
||||
text = (await self.runner.transcribe(audio, sess.language)).strip()
|
||||
except Exception:
|
||||
logger.exception("Stream %s: Partial-Transcribe crashed", sess.request_id[:8])
|
||||
return
|
||||
if text and text != sess.last_partial:
|
||||
sess.last_partial = text
|
||||
sess.last_growth_at = now
|
||||
if self._ws is not None:
|
||||
await _send(self._ws, "stt_partial", {
|
||||
"requestId": sess.request_id,
|
||||
"audioRequestId": sess.audio_request_id,
|
||||
"text": text,
|
||||
})
|
||||
|
||||
async def _finalize(self, sess: StreamSession, reason: str) -> None:
|
||||
if sess.endpoint_sent:
|
||||
return
|
||||
sess.endpoint_sent = True
|
||||
# vLLM ggf. committen, damit ein letztes transcription.done kommt.
|
||||
if sess.vllm_ws is not None:
|
||||
try:
|
||||
await sess.vllm_ws.send(json.dumps({"type": VLLM_SEND_COMMIT}))
|
||||
await asyncio.sleep(0.15) # kurz auf finalen Delta warten
|
||||
except Exception:
|
||||
pass
|
||||
final_text = sess.last_partial.strip()
|
||||
duration_s = self._buffer_duration_ms(sess) / 1000.0
|
||||
logger.info("Stream %s: FINAL (reason=%s, %.1fs): %r",
|
||||
sess.request_id[:8], reason, duration_s, final_text[:120])
|
||||
audio = pcm_s16le_to_float32(bytes(sess.pcm_buffer))
|
||||
t0 = time.time()
|
||||
try:
|
||||
final_text = (await self.runner.transcribe(audio, sess.language)).strip()
|
||||
except Exception:
|
||||
logger.exception("Stream %s: Final-Transcribe crashed", sess.request_id[:8])
|
||||
final_text = sess.last_partial
|
||||
stt_ms = int((time.time() - t0) * 1000)
|
||||
duration_s = audio.size / 16000.0
|
||||
logger.info("Stream %s: FINAL (reason=%s, %.1fs, %dms): %r",
|
||||
sess.request_id[:8], reason, duration_s, stt_ms, final_text[:120])
|
||||
if self._ws is not None:
|
||||
endpoint_payload = {
|
||||
payload = {
|
||||
"requestId": sess.request_id,
|
||||
"audioRequestId": sess.audio_request_id,
|
||||
"text": final_text,
|
||||
"reason": reason,
|
||||
"durationS": duration_s,
|
||||
"sttMs": 0,
|
||||
"sttMs": stt_ms,
|
||||
"voice": sess.voice,
|
||||
"speed": sess.speed,
|
||||
"interrupted": sess.interrupted,
|
||||
}
|
||||
if sess.location:
|
||||
endpoint_payload["location"] = sess.location
|
||||
await _send(self._ws, "stt_endpoint", endpoint_payload)
|
||||
payload["location"] = sess.location
|
||||
await _send(self._ws, "stt_endpoint", payload)
|
||||
await _send(self._ws, "stt_stream_done", {
|
||||
"requestId": sess.request_id,
|
||||
"audioRequestId": sess.audio_request_id,
|
||||
@@ -447,7 +367,6 @@ async def run_loop(sessions: SessionManager) -> None:
|
||||
sessions.attach_ws(ws)
|
||||
await _broadcast_status(ws, "ready", model=VOXTRAL_MODEL)
|
||||
await _send(ws, "config_request", {"service": "voxtral"})
|
||||
|
||||
async for raw in ws:
|
||||
try:
|
||||
msg = json.loads(raw)
|
||||
@@ -456,35 +375,32 @@ async def run_loop(sessions: SessionManager) -> None:
|
||||
mtype = msg.get("type", "")
|
||||
payload = msg.get("payload", {}) or {}
|
||||
if mtype == "stt_stream_start":
|
||||
asyncio.create_task(sessions.start_session(payload))
|
||||
sessions.start_session(payload)
|
||||
elif mtype == "stt_audio_chunk":
|
||||
sessions.feed_chunk(payload)
|
||||
elif mtype == "stt_stream_end":
|
||||
sessions.end_session(payload.get("requestId", ""))
|
||||
# stt_request (Legacy One-Shot) macht Voxtral hier NICHT —
|
||||
# dafuer bleibt die whisper-Bridge (Fallback).
|
||||
except Exception as e:
|
||||
logger.warning("RVS-Verbindung verloren: %s — retry in %ds", e, retry_s)
|
||||
if use_tls and RVS_TLS_FALLBACK and not tls_fallback_tried:
|
||||
use_tls = False
|
||||
tls_fallback_tried = True
|
||||
logger.info("TLS-Fallback: versuche ws:// (kein TLS)")
|
||||
continue
|
||||
await asyncio.sleep(retry_s)
|
||||
retry_s = min(retry_s * 2, 30)
|
||||
use_tls = RVS_TLS # fuer den naechsten Zyklus zuruecksetzen
|
||||
use_tls = RVS_TLS
|
||||
|
||||
|
||||
async def main() -> None:
|
||||
if not RVS_HOST or not RVS_TOKEN:
|
||||
logger.error("RVS_HOST/RVS_TOKEN fehlen — .env pruefen. Abbruch.")
|
||||
return
|
||||
sessions = SessionManager()
|
||||
logger.info("Voxtral-Bridge startet — vLLM=%s Modell=%s", VOXTRAL_VLLM_URL, VOXTRAL_MODEL)
|
||||
await asyncio.gather(
|
||||
run_loop(sessions),
|
||||
sessions.run_endpointer(),
|
||||
)
|
||||
runner = VoxtralRunner()
|
||||
loop = asyncio.get_running_loop()
|
||||
await loop.run_in_executor(None, runner.load) # Modell laden (blockierend)
|
||||
sessions = SessionManager(runner)
|
||||
logger.info("Voxtral-Bridge startet — Modell=%s", VOXTRAL_MODEL)
|
||||
await asyncio.gather(run_loop(sessions), sessions.run_endpointer())
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
|
||||
@@ -1,5 +1,8 @@
|
||||
# Voxtral-Bridge ist reine CPU-Glue (RVS-WS <-> vLLM-Realtime-WS). Das Modell
|
||||
# selbst laeuft im separaten voxtral-vllm-Container (GPU). Deshalb hier KEIN
|
||||
# torch/vllm — nur der WebSocket-Client + numpy fuer die RMS-Energiemessung.
|
||||
websockets>=12.0
|
||||
# Voxtral-3B via Transformers. torch/torchaudio kommen cu124-gepinnt aus dem
|
||||
# Dockerfile (nicht hier, sonst zieht pip das Default-CUDA-Wheel).
|
||||
transformers>=4.54
|
||||
mistral-common[audio]>=1.8.1
|
||||
accelerate>=0.30
|
||||
soundfile>=0.12
|
||||
numpy>=1.24
|
||||
websockets>=12.0
|
||||
|
||||
Reference in New Issue
Block a user