Docker Deployment

Agent Network provides a Docker Compose orchestration solution for one-click startup of a complete agent squad.

Quick Start

bash

cd demos/codex-telegram-squad

# 1. Configure environment variables
cp .env.example .env
# Edit .env to fill in tokens and API keys

# 2. Start (12 containers)
docker compose up -d

# 3. Check status
docker compose ps
docker compose logs -f commander

Architecture

Dockerfile Details

Dockerfile.server (CommHub Server)

dockerfile

FROM oven/bun:1
WORKDIR /app

# Copy only the server directory
COPY server/ ./server/
COPY agent-network/src/ ./agent-network/src/

WORKDIR /app/server
RUN bun install

EXPOSE 9200
CMD ["bun", "run", "src/index.ts"]

Key points:

Based on Bun image (CommHub Server runs on Bun)
Only needs server/ and agent-network/src/
Exposes port 9200

Dockerfile.agent (Agent Node)

dockerfile

FROM node:20-slim
WORKDIR /app

# Install agent-node and CLI
COPY agent-network/ ./agent-network/
COPY agent-node/ ./agent-node/
COPY channel/ ./channel/

RUN cd agent-network && npm install && npm link
RUN cd agent-node && npm install

# Install Codex CLI (optional)
RUN npm install -g @openai/codex

COPY demos/codex-telegram-squad/entrypoint.sh /entrypoint.sh
RUN chmod +x /entrypoint.sh

ENTRYPOINT ["/entrypoint.sh"]

Key points:

Based on Node.js image
Installs agent-network (CLI) and agent-node (runtime)
Starts via entrypoint.sh, which selects the runtime based on environment variables

entrypoint.sh

bash

#!/bin/bash
set -e

# Wait for Server to be ready
until curl -sf http://$COMMHUB_URL/health; do
  sleep 1
done

# Read ntok_ (from shared volume exported by seed container)
if [ -f /shared/ntok ]; then
  export COMMHUB_TOKEN=$(cat /shared/ntok)
fi

# Start Agent Node
exec npx @sleep2agi/agent-node \
  --alias "$ALIAS" \
  --runtime "$RUNTIME" \
  --model "$MODEL" \
  --hub "$COMMHUB_URL" \
  --token "$COMMHUB_TOKEN" \
  ${TOOLS:+--tools "$TOOLS"} \
  ${SYSTEM_PROMPT:+--system-prompt "$SYSTEM_PROMPT"}

docker-compose.yml Details

Shared Configuration

yaml

x-common: &common
  build:
    context: ../..
    dockerfile: demos/codex-telegram-squad/Dockerfile.agent
  volumes:
    - ${HOME}/.codex:/root/.codex:ro          # Codex auth
    - ${HOME}/.claude.json:/root/.claude.json:ro  # Claude auth
    - squad_shared:/shared                     # Shared ntok_
  tmpfs:
    - /root/.claude    # Writable temp directory
    - /tmp
  depends_on:
    seed:
      condition: service_completed_successfully
  restart: unless-stopped

Key design decisions:

Mount	Mode	Description
`~/.codex`	`ro`	Codex auth (read-only)
`~/.claude.json`	`ro`	Claude auth (read-only)
`squad_shared`	`rw`	Shared volume for ntok_
`/root/.claude`	`tmpfs`	Agent SDK needs writable dir, uses tmpfs

Seed Container

The seed container registers the admin and exports ntok_ after the server starts:

yaml

seed:
  image: curlimages/curl:latest
  depends_on:
    server:
      condition: service_healthy
  volumes:
    - squad_shared:/shared
  entrypoint:
    - sh
    - -c
    - |
      # Register admin
      RESP=$(curl -sX POST http://server:9200/api/auth/register \
        -H "Content-Type: application/json" \
        -H "Authorization: Bearer ${COMMHUB_AUTH_TOKEN}" \
        -d '{"username":"admin","password":"admin123"}')

      # Extract ntok_ and write to shared volume
      NTOK=$(echo "$RESP" | sed -n 's/.*"network_token":"\(ntok_[^"]*\)".*/\1/p')
      echo "$NTOK" > /shared/ntok
  restart: "no"

The seed container is one-shot (restart: "no") and only runs on first startup. Subsequent restarts skip if /shared/ntok already exists.

Server Health Check

yaml

server:
  healthcheck:
    test: ["CMD", "curl", "-sf", "http://127.0.0.1:9200/health"]
    interval: 3s
    timeout: 5s
    retries: 10

All agent containers wait for the server via depends_on + condition: service_healthy.

Environment Variables

.env File

bash

# CommHub auth
COMMHUB_AUTH_TOKEN=squad-token

# Telegram Bot
TELEGRAM_BOT_TOKEN=123456789:ABCdefGhIJKlmNoPQRsTUVwxyz
TELEGRAM_ALLOW_USER=7612221352

# MiniMax API
MINIMAX_API_KEY=your-minimax-api-key

# Dashboard password
DASHBOARD_PASSWORD=squad-dash

Container Environment Variables

Variable	Description	Example
`ALIAS`	Agent name	`coder-1`
`RUNTIME`	Runtime engine	`codex-sdk` / `claude-agent-sdk`
`MODEL`	Model	`gpt-5.5` / `claude-3-5-haiku-20241022`
`COMMHUB_URL`	Server address	`http://server:9200`
`COMMHUB_TOKEN`	Auth token	`ntok_xxx` or read from /shared/ntok
`TOOLS`	Tool list	`Read,Write,Edit,Bash,Glob,Grep`
`SYSTEM_PROMPT`	System prompt	Commander's task dispatch rules
`ANTHROPIC_BASE_URL`	MiniMax API URL	`https://api.minimaxi.com/anthropic`
`ANTHROPIC_AUTH_TOKEN`	MiniMax API key	MiniMax Key

Common Operations

Start

bash

# Start all
docker compose up -d

# Start only Server + Commander
docker compose up -d server seed commander

# Start and follow logs
docker compose up

Check Status

bash

# Container status
docker compose ps

# All logs
docker compose logs

# Specific container logs
docker compose logs -f commander
docker compose logs -f worker-1

# CommHub API check
curl http://localhost:9299/api/status
curl http://localhost:9299/health

Scale Up/Down

bash

# Add workers (must be defined in compose)
docker compose up -d --scale worker=10

# Stop a specific worker
docker compose stop worker-5

Stop and Clean Up

bash

# Stop all
docker compose down

# Stop and remove data volumes
docker compose down -v

# Rebuild images
docker compose build --no-cache
docker compose up -d

Port Mapping

Service	Container Port	Host Port	Description
Server	9200	9299	CommHub API
Dashboard	3000	9999	Web UI

Persistence

Data	Storage	Description
CommHub database	Inside server container	Not persisted by default, lost on restart
ntok_	`squad_shared` volume	Persisted to Docker volume
Agent logs	tmpfs	Not persisted

To persist the database:

yaml

server:
  volumes:
    - ./data:/root/.commhub  # Persist SQLite database

Custom Compose

Adding More Workers

yaml

# Add to docker-compose.yml
worker-11:
  <<: *common
  environment:
    - ALIAS=coder-6
    - RUNTIME=codex-sdk
    - MODEL=gpt-5.5
    - COMMHUB_URL=http://server:9200
    - TOOLS=Read,Write,Edit,Bash,Glob,Grep

Using Different Models

yaml

# DeepSeek Worker
worker-deepseek:
  <<: *common
  environment:
    - ALIAS=deep-1
    - RUNTIME=claude-agent-sdk
    - MODEL=deepseek-chat
    - ANTHROPIC_BASE_URL=https://api.deepseek.com/anthropic
    - ANTHROPIC_AUTH_TOKEN=${DEEPSEEK_API_KEY}
    - COMMHUB_URL=http://server:9200

Docker Deployment ​

Quick Start ​

Architecture ​

Dockerfile Details ​

Dockerfile.server (CommHub Server) ​

Dockerfile.agent (Agent Node) ​

entrypoint.sh ​

docker-compose.yml Details ​

Shared Configuration ​

Seed Container ​

Server Health Check ​

Environment Variables ​

.env File ​

Container Environment Variables ​

Common Operations ​

Start ​

Check Status ​

Scale Up/Down ​

Stop and Clean Up ​

Port Mapping ​

Persistence ​

Custom Compose ​

Adding More Workers ​

Using Different Models ​

Docker Deployment

Quick Start

Architecture

Dockerfile Details

Dockerfile.server (CommHub Server)

Dockerfile.agent (Agent Node)

entrypoint.sh

docker-compose.yml Details

Shared Configuration

Seed Container

Server Health Check

Environment Variables

.env File

Container Environment Variables

Common Operations

Start

Check Status

Scale Up/Down

Stop and Clean Up

Port Mapping

Persistence

Custom Compose

Adding More Workers

Using Different Models