VoiceCortex/docker-compose.yml at master · OneByJorah/VoiceCortex · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
version: "3.9"

# ============================================================================
# HERMES VOICE — self-hosted voice assistant call stack
# JorahOne
#
# Call path:
#   Caller --SIP/RTP--> [PBX of your choice] --SIP/RTP--> FreeSWITCH (media)
#                                                          --WebSocket audio--> bot
#                                                            |-- STT: faster-whisper
#                                                            |-- LLM: Ollama / llama.cpp / hosted API
#                                                            |-- TTS: Piper
#                                                          --WebSocket audio--> back into call
#
# PBX OPTIONS (see README.md for overview):
#   - none/standalone : bundled FreeSWITCH below is the only PBX, dial ext. 8500
#   - 3CX             : keep 3CX, trunk a DID/extension into this FreeSWITCH
#   - FreePBX         : same pattern as 3CX, trunk into this FreeSWITCH
#   - Asterisk        : use Asterisk's ARI + externalMedia instead of
#                        FreeSWITCH's audio_fork (see bot/server_asterisk.py)
#   In every case the PBX only needs to get a call's media to FreeSWITCH (or
#   directly to the bot, for the Asterisk path) — the STT/LLM/TTS bot logic
#   is identical regardless of which PBX answers the call.
#
# LLM BACKEND OPTIONS (set LLM_BACKEND in .env):
#   - ollama   : local Ollama server (default, e.g. your RTX 3060 box)
#   - llamacpp : local llama.cpp `server` binary (OpenAI-compatible endpoint)
#   - api      : hosted API — Anthropic by default, or any OpenAI-compatible
#                endpoint (OpenAI, OpenRouter, etc.) via API_PROVIDER
# ============================================================================

services:

  freeswitch:
    image: signalwire/freeswitch:latest
    container_name: hermes-freeswitch
    network_mode: host          # simplest for SIP/RTP NAT behavior on a LAN
    volumes:
      - ./freeswitch/conf:/etc/freeswitch
    environment:
      - TZ=America/New_York     # adjust
    restart: unless-stopped
    profiles: ["freeswitch"]    # `docker compose --profile freeswitch up`
                                  # omit this profile entirely if bridging
                                  # to your own 3CX/FreePBX/Asterisk instead

  bot:
    build: ./bot
    container_name: hermes-bot
    network_mode: host
    volumes:
      - ./bot:/app
      - ./piper-voices:/piper-voices
    env_file:
      - .env
    environment:
      - WHISPER_MODEL=${WHISPER_MODEL:-small.en}
      - WHISPER_DEVICE=${WHISPER_DEVICE:-cuda}
      - PIPER_VOICE=${PIPER_VOICE:-/piper-voices/en_US-amy-medium.onnx}
      - BOT_WS_PORT=${BOT_WS_PORT:-8765}
      - HERMES_SYSTEM_PROMPT_FILE=/app/persona.txt
    restart: unless-stopped

# NOTE: Ollama is assumed to already be running on your host (RTX 3060 box)
# per your existing setup. If you'd rather containerize it here too, add:
#
#  ollama:
#    image: ollama/ollama:latest
#    container_name: hermes-ollama
#    runtime: nvidia
#    environment:
#      - NVIDIA_VISIBLE_DEVICES=all
#    volumes:
#      - ollama-data:/root/.ollama
#    ports:
#      - "11434:11434"
#    restart: unless-stopped
#
# volumes:
#   ollama-data: