feat: FRIDAY Voice stack on the GPU machine; VoiceCoach history

FRIDAY Voice (speaches GPU STT/TTS + hermes agent + push-to-talk console) on
.150 Docker Desktop, fronted by coach.hobbs.farm. Also captures the VoiceCoach
compose/agent trees and deploy notes it replaced, which had never been committed.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

compose/gpu-machine/hermes/docker-compose.yml

@@ -0,0 +1,30 @@
+# Hermes Agent on .150 (Docker Desktop) — FRIDAY's reasoning agent.
+# Data/settings live in the `hermes-data` named volume, seeded from the user's
+# WSL ~/.hermes (settings preserved; SQLite state.db kept on the VM ext4, not a
+# bridged bind mount). friday-voice calls this via `docker exec hermes hermes ...`.
+services:
+  gateway:
+    image: nousresearch/hermes-agent:latest
+    container_name: hermes
+    restart: unless-stopped
+    volumes:
+      - hermes-data:/opt/data
+    environment:
+      - HERMES_UID=10000
+      - HERMES_GID=10000
+    command: ["gateway", "run"]
+    labels:
+      - "com.centurylinklabs.watchtower.enable=false"
+    logging:
+      driver: "json-file"
+      options:
+        max-size: "50m"
+        max-file: "3"
+
+  # dashboard: disabled — crash-loops under s6 on Docker Desktop and not needed
+  # by friday-voice (gateway + `docker exec hermes` is all that's required).
+  # Re-enable + debug later if the localhost:9119 web UI is wanted.
+
+volumes:
+  hermes-data:
+    external: true

compose/gpu-machine/voicecoach-speaches/docker-compose.yml

@@ -0,0 +1,36 @@
+services:
+  speaches:
+    image: ghcr.io/speaches-ai/speaches:latest-cuda
+    container_name: voicecoach-speaches-gpu
+    restart: unless-stopped
+    environment:
+      - WHISPER__INFERENCE_DEVICE=cuda
+      - WHISPER__COMPUTE_TYPE=float16
+      - WHISPER__TTL=-1
+    volumes:
+      - speaches-models:/home/ubuntu/.cache
+    ports:
+      - "8000:8000"
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: nvidia
+              count: 1
+              capabilities: [gpu]
+    labels:
+      - "com.centurylinklabs.watchtower.enable=false"
+    healthcheck:
+      test: ["CMD", "curl", "-sf", "http://localhost:8000/health"]
+      interval: 30s
+      timeout: 10s
+      retries: 3
+      start_period: 180s
+    logging:
+      driver: "json-file"
+      options:
+        max-size: "50m"
+        max-file: "3"
+
+volumes:
+  speaches-models: