Switch transcription to Whisper large v3

deploy/ai-server/docker-compose.audio.yml

@@ -1,12 +1,12 @@
 services:
-  voxtral-small:
+  whisper-large-v3:
     build:
       context: .
       dockerfile: vllm-audio.Dockerfile
     image: vllm-audio:local
-    container_name: voxtral-small
+    container_name: whisper-large-v3
     profiles:
-      - voxtral-small
+      - whisper-large-v3
     restart: unless-stopped
     ipc: host
     runtime: nvidia
@@ -29,32 +29,19 @@ services:
       - "10.2.3.5:8004:8000"
     command:
       - "--model"
-      - "mistralai/Voxtral-Small-24B-2507"
+      - "openai/whisper-large-v3"
       - "--served-model-name"
-      - "mistralai/Voxtral-Small-24B-2507"
-      - "--tokenizer-mode"
-      - "mistral"
-      - "--config-format"
-      - "mistral"
-      - "--load-format"
-      - "mistral"
-      - "--tool-call-parser"
-      - "mistral"
-      - "--enable-auto-tool-choice"
+      - "openai/whisper-large-v3"
+      - "--task"
+      - "transcription"
       - "--host"
       - "0.0.0.0"
       - "--port"
       - "8000"
-      - "--max-model-len"
-      - "16384"
       - "--gpu-memory-utilization"
       - "0.55"
       - "--api-key"
       - "${VLLM_API_KEY}"
-      - "--max-num-seqs"
-      - "1"
-      - "--max-num-batched-tokens"
-      - "4096"
     healthcheck:
       test: ["CMD", "python3", "-c", "import urllib.request; urllib.request.urlopen('http://localhost:8000/health')"]
       interval: 30s