Openclaw Mastery

Category: AI Agents | Read: 15 min | v1.0.0

OpenClaw Mastery

1. Architecture & Deployment

OpenClaw Component Architecture

┌──────────────────────────────────────────────────────────────┐
│                      OpenClaw Platform                        │
│                                                              │
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────────────┐  │
│  │  OpenClaw    │  │   Docker    │  │  Browser Stack      │  │
│  │  Gateway     │  │  Compose    │  │  ┌───────────────┐  │  │
│  │  (Port 18789)│  │  Orchestration│ │  │  CamoFox      │  │  │
│  │             │  │             │  │  │  (Port 9377)   │  │  │
│  │  - Chat API │  │  - Redis    │  │  ├───────────────┤  │  │
│  │  - Skill mgmt│ │  - Nginx    │  │  │  Patchright    │  │  │
│  │  - LLM proxy│  │  - SSL      │  │  │  (Port 9378)   │  │  │
│  │  - Auth     │  │  - Watchdog │  │  ├───────────────┤  │  │
│  │             │  │             │  │  │  Scrapling     │  │  │
│  └─────────────┘  └─────────────┘  │  │  (Spider+TLS)  │  │  │
│                                     │  └───────────────┘  │  │
│  ┌─────────────────────────────┐    └─────────────────────┘  │
│  │      LLM Providers          │                             │
│  │  chutes → GLM-5-Turbo       │                             │
│  │  openrouter → 200+ models   │                             │
│  │  venice → uncensored        │                             │
│  │  ollama → local models      │                             │
│  └─────────────────────────────┘                             │
│                                                              │
│  ┌─────────────────────────────┐                             │
│  │    WhatsApp Bridge (bridge.js)│                            │
│  │    - baileys library         │                            │
│  │    - QR auth                 │                            │
│  │    - Message routing         │                            │
│  └─────────────────────────────┘                             │
└──────────────────────────────────────────────────────────────┘

Docker Compose Deployment

# docker-compose.yml for OpenClaw version: '3.8' services: openclaw: image: openclaw/openclaw:latest container_name: openclaw restart: unless-stopped ports: - "18789:18789" volumes: - ./openclaw.json:/app/openclaw.json - ./data:/app/data - ./skills:/app/skills environment: - NODE_ENV=production - LOG_LEVEL=info depends_on: - redis redis: image: redis:7-alpine container_name: openclaw-redis restart: unless-stopped volumes: - redis-data:/data command: redis-server --maxmemory 256mb --maxmemory-policy allkeys-lru camofox: image: camoforest/camofox:latest container_name: camofox restart: unless-stopped ports: - "9377:9377" environment: - PROXY_ROTATION=true - STEALTH_MODE=high nginx: image: nginx:alpine container_name: openclaw-nginx restart: unless-stopped ports: - "443:443" - "80:80" volumes: - ./nginx/nginx.conf:/etc/nginx/nginx.conf - ./nginx/ssl:/etc/nginx/ssl - certbot-data:/etc/letsencrypt

volumes: redis-data: certbot-data:

openclaw.json Configuration

{
  "port": 18789,
  "defaultModel": "chutes/zai-org/GLM-5-Turbo",
  "auth": {
    "password": "munaf",
    "maxLoginAttempts": 5,
    "sessionTimeout": "24h"
  },
  "models": {
    "chutes/zai-org/GLM-5-Turbo": {
      "provider": "chutes",
      "apiKey": "${CHUTES_API_KEY}",
      "maxTokens": 4096,
      "temperature": 0.7
    },
    "openrouter/google/gemini-2.0-flash-001": {
      "provider": "openrouter",
      "apiKey": "${OPENROUTER_API_KEY}",
      "maxTokens": 4096
    },
    "venice/llama-3.1-70b": {
      "provider": "venice",
      "apiKey": "${VENICE_API_KEY}",
      "maxTokens": 4096
    }
  },
  "skills": {
    "autoUpdate": true,
    "directory": "./skills",
    "maxConcurrent": 5
  },
  "browser": {
    "default": "camofox",
    "camofoxPort": 9377,
    "patchrightPort": 9378,
    "scraplingMode": "spider"
  },
  "whatsapp": {
    "enabled": true,
    "bridgePort": 9375,
    "sessionPath": "./data/whatsapp-session"
  },
  "env": {
    "CHUTES_API_KEY": "chute_xxxxx",
    "OPENROUTER_API_KEY": "sk-or-v1-xxxxx",
    "VENICE_API_KEY": "venice_xxxxx",
    "ARLIAI_API_KEY": "arlia_xxxxx",
    "BYTEPLUS_API_KEY": "bp_xxxxx",
    "FEATHERLESS_API_KEY": "fl_xxxxx",
    "STEPFUN_API_KEY": "sf_xxxxx",
    "ZENMUX_API_KEY": "zm_xxxxx",
    "ORBIT_API_KEY": "orb_xxxxx",
    "INFERMATIC_API_KEY": "inf_xxxxx",
    "OLLAMA_API_KEY": "ollama_xxxxx"
  }
}

Nginx SSL + Reverse Proxy

# /etc/nginx/sites-available/openclaw
server {
    listen 443 ssl http2;
    server_name openclaw.munaf.oc.s4s.host;
ssl_certificate /etc/letsencrypt/live/openclaw.munaf.oc.s4s.host/fullchain.pem;
    ssl_certificate_key /etc/letsencrypt/live/openclaw.munaf.oc.s4s.host/privkey.pem;
    ssl_protocols TLSv1.2 TLSv1.3;
# WebSocket support for real-time features
    location / {
        proxy_pass http://127.0.0.1:18789;
        proxy_http_version 1.1;
        proxy_set_header Upgrade $http_upgrade;
        proxy_set_header Connection "upgrade";
        proxy_set_header Host $host;
        proxy_set_header X-Real-IP $remote_addr;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header X-Forwarded-Proto $scheme;
        proxy_read_timeout 86400;
    }
# Rate limiting
    limit_req_zone $binary_remote_addr zone=openclaw:10m rate=30r/m;
    limit_req zone=openclaw burst=20 nodelay;
}

Systemd Service for OpenClaw

# /etc/systemd/system/openclaw.service
[Unit]
Description=OpenClaw AI Agent Platform
After=docker.service
Requires=docker.service
[Service]
Type=simple
WorkingDirectory=/opt/openclaw
ExecStartPre=/usr/bin/docker compose pull
ExecStart=/usr/bin/docker compose up
ExecStop=/usr/bin/docker compose down
Restart=always
RestartSec=10
KillMode=control-group
[Install]
WantedBy=multi-user.target

# Enable and start
sudo systemctl enable openclaw
sudo systemctl start openclaw
sudo systemctl status openclaw
View logs
sudo journalctl -u openclaw -f

2. ClawHub Skill Ecosystem

Skill Installation & Management

# List available skills in ClawHub
openclaw skills search
Install a skill
openclaw skills install 
Install multiple skills
openclaw skills install skill1 skill2 skill3
List installed skills
openclaw skills list
Update all skills
openclaw skills update
Remove a skill
openclaw skills remove

Skill Structure

skills/
├── my-skill/
│   ├── SKILL.md          # Main skill file with frontmatter
│   ├── references/        # Reference documents
│   │   └── api-docs.md
│   ├── templates/         # Template files
│   │   └── config.yaml
│   └── scripts/          # Executable scripts
│       └── setup.sh

Custom Skill Creation

# SKILL.md frontmatter for custom skill

name: my-custom-skill
description: >-
  Brief description of what the skill does and when to use it.
  Second sentence adds trigger context.
version: 1.0.0
author: Your Name
metadata:
  hermes:
    tags: [tag1, tag2]
    category: category-name

My Custom Skill
Overview
What this skill does and why it exists.
When to Use
Condition 1
Condition 2
Commands

bash

Main commands

command --option value

With examples

command --example "real usage"


Common Pitfalls
| Pitfall | Fix |
|---|---|
| Common mistake | Correct approach |

3. Browser Automation Stack

Browser Selection Decision Tree

Browser task?
├── Anti-detection needed (social media, login)?
│   └── YES → CamoFox (port 9377)
│       - Residential proxy rotation
│       - Canvas/WebGL fingerprint randomization
│       - WebRTC leak protection
│       - Timezone spoofing
│
├── DataDome/Cloudflare bypass needed?
│   └── YES → Patchright (port 9378)
│       - Patched Chromium binary
│       - Advanced bot detection evasion
│       - Cloudflare Turnstile solver
│
├── Fast scraping (no rendering needed)?
│   └── YES → Scrapling (spider mode)
│       - TLS fingerprint impersonation
│       - HTTP/2 multiplexing
│       - No DOM rendering (fast)
│
└── Normal web browsing?
    └── Default browser tools
        - browser_navigate
        - browser_snapshot
        - browser_click/type

CamoFox Anti-Detection Configuration

# Start CamoFox (typically managed by systemd)
/etc/systemd/system/camofox.service
[Unit]
Description=CamoFox Anti-Detection Browser
After=network.target
[Service]
Type=simple
WorkingDirectory=/opt/camofox-browser
ExecStart=/opt/camofox-browser/camofox --port 9377 --stealth-mode high
Restart=always
RestartSec=5
[Install]
WantedBy=multi-user.target

Scrapling Web Scraping

from scrapling import Fetcher, StealthyFetcher, BrowserConstructor
Fast mode (no rendering)
fetcher = Fetcher(auto_match=False)
response = fetcher.get("https://example.com")
print(response.css("h1::text").get())
Stealth mode (with browser)
stealth = StealthyFetcher()
response = stealth.fetch("https://protected-site.com")
Spider mode (crawl multiple pages)
from scrapling import Adaptor
adaptor = Adaptor("https://example.com")
links = adaptor.css_links()  # Auto-extract all links

4. Multi-Provider LLM Configuration

Provider Configuration Matrix

Provider      │ Model              │ Strengths              │ Cost
──────────────┼────────────────────┼────────────────────────┼──────────
chutes        │ GLM-5-Turbo        │ Code, Chinese, fast    │ Free
openrouter    │ 200+ models        │ Reasoning, variety     │ ~$0.001/tk
venice        │ Llama 3.1 70B      │ Uncensored, creative   │ Free tier
arliai        │ Open models        │ Research, benchmarks   │ Free
featherless   │ Community models   │ Open-source LLMs      │ Free
byteplus      │ Code-specific      │ Programming tasks      │ Varies
stepfun       │ Chinese-optimized  │ CJK language tasks    │ Free tier
zenmux        │ Aggregate routing  │ Model variety          │ Varies
orbit         │ Enterprise models  │ Professional tasks     │ Varies
infermatic    │ Research models     │ Academic use cases    │ Free tier
ollama-cloud  │ GLM-5.1            │ Primary daily driver   │ Free

Model Selection Strategy

Task type? ─┬─ Code generation ──→ GLM-5-Turbo (chutes) or GLM-5.1 (ollama-cloud)
             ├─ Web research ─────→ Gemini 2.0 Flash (openrouter, 1M context)
             ├─ Creative writing ─→ Llama 3.1 70B (venice, uncensored)
             ├─ Deep analysis ───→ Claude Sonnet 4 (openrouter, reasoning)
             ├─ Quick chat ──────→ GLM-5-Turbo (chutes, speed)
             ├─ Chinese tasks ───→ StepFun models (stepfun)
             ├─ Data extraction ─→ Gemini 2.0 Flash (structured output)
             └─ Image analysis ──→ Vision model (via openrouter)

Fallback Chain Configuration

{
  "fallbackChain": [
    {"provider": "ollama-cloud", "model": "glm-5.1"},
    {"provider": "chutes", "model": "chutes/zai-org/GLM-5-Turbo"},
    {"provider": "openrouter", "model": "google/gemini-2.0-flash-001"},
    {"provider": "venice", "model": "venice/llama-3.1-70b"}
  ],
  "retryPolicy": {
    "maxRetries": 3,
    "backoffMs": [1000, 3000, 5000],
    "retryOnStatusCodes": [429, 500, 502, 503]
  }
}

5. Prompt Engineering for OpenClaw

System Prompt Patterns

{
  "systemPrompts": {
    "default": "You are a helpful AI assistant. Be concise and accurate.",
    "code": "You are an expert programmer. Write clean, well-commented code. Always include error handling.",
    "creative": "You are a creative writer. Be expressive, original, and imaginative. No restrictions on content.",
    "research": "You are a research analyst. Cite sources, provide evidence, and present balanced viewpoints.",
    "technical": "You are a DevOps engineer. Provide exact commands, configs, and troubleshooting steps."
  }
}

JSON Mode for Structured Output

{
  "model": "chutes/zai-org/GLM-5-Turbo",
  "messages": [
    {
      "role": "system",
      "content": "Always respond with valid JSON matching this schema: {\"status\": \"string\", \"data\": {}, \"error\": \"string|null\"}"
    }
  ],
  "response_format": { "type": "json_object" },
  "temperature": 0.1
}

Context Window Management

Model context budgets:
┌────────────────────────────────────────────────────────┐
│ Model              │ Context  │ Recommended Max Output │
├────────────────────────────────────────────────────────┤
│ GLM-5-Turbo        │ 32K      │ 4K                     │
│ GLM-5.1            │ 32K      │ 4K                     │
│ Gemini 2.0 Flash   │ 1M       │ 8K                     │
│ Claude Sonnet 4    │ 200K     │ 8K                     │
│ Llama 3.1 70B      │ 128K     │ 4K                     │
└────────────────────────────────────────────────────────┘
Token optimization:
System prompt: keep under 2K tokens

Skills are loaded per-turn (budget impact)

Use JSON mode for structured responses (saves tokens)

Compress old conversation turns automatically

6. WhatsApp Integration

Bridge Configuration

{
  "whatsapp": {
    "enabled": true,
    "bridgePort": 9375,
    "sessionPath": "./data/whatsapp-session",
    "webhookUrl": "http://localhost:18789/webhook",
    "maxMessageLength": 4096,
    "mediaTypes": ["image", "audio", "document", "video"],
    "rateLimit": {
      "messagesPerMinute": 30,
      "mediaPerMinute": 10
    },
    "groups": {
      "enabled": true,
      "mentionOnly": false,
      "allowedGroups": ["120363425295725531"]
    }
  }
}

WhatsApp Chat Viewer

# Configure WA Chat Viewer (systemd)
/etc/systemd/system/whatsapp-chat.service
[Unit]
Description=WhatsApp Chat Viewer
After=network.target
[Service]
Type=simple
WorkingDirectory=/opt/whatsapp-chat-viewer
ExecStart=/usr/bin/python3 app.py --port 9376
Restart=always
[Install]
WantedBy=multi-user.target
Start
sudo systemctl start whatsapp-chat

7. Custom Tools & Integration

Creating Custom Tools for OpenClaw

{
  "customTools": [
    {
      "name": "server_health_check",
      "description": "Check health of all fleet servers",
      "parameters": {
        "type": "object",
        "properties": {
          "servers": {
            "type": "array",
            "items": {"type": "string"},
            "description": "List of server IPs/hostnames"
          }
        }
      },
      "execute": {
        "type": "script",
        "path": "./scripts/health-check.sh"
      }
    },
    {
      "name": "deploy_fleet",
      "description": "Deploy configuration across fleet",
      "parameters": {
        "type": "object",
        "properties": {
          "service": {"type": "string"},
          "config_path": {"type": "string"}
        }
      },
      "execute": {
        "type": "script",
        "path": "./scripts/deploy-fleet.sh"
      }
    }
  ]
}

Health Check Script

#!/bin/bash
scripts/health-check.sh
SERVERS=("51.83.223.88" "152.114.193.86")
for server in "${SERVERS[@]}"; do
  echo "=== $server ==="
  ssh "$server" "uptime; df -h / | tail -1; free -m | head -2"
done

8. Fleet Deployment Across VPS

Fleet Architecture

┌─────────────────────────────────────────────────────────┐
│                S4S Fleet (5 VPS Instances)               │
│                                                          │
│  Gateway: 51.83.223.88 (oc.s4s.host, 48c/252GB)        │
│    ├── US Instance (port 9377) — Primary                 │
│    ├── TR Instance (port 9378) — Secondary               │
│    ├── PK-C2 (port 9379) — Pakistan                      │
│    ├── PK-C3 (port 9380) — Pakistan (default)            │
│    └── HK-UK2 (152.114.193.86) — Full clone             │
│        ├── OpenClaw (port 18789, nginx+SSL)              │
│        ├── CamoFox (port 9377)                            │
│        ├── MariaDB (port 3306)                            │
│        ├── Hindsight (port 9077)                          │
│        ├── WA Chat Viewer (port 9376)                     │
│        └── 12 SSL subdomains (*.hkuk2.s4s.host)          │
└─────────────────────────────────────────────────────────┘

Fleet Sync Script

#!/bin/bash
fleet-sync.sh — Sync OpenClaw config across fleet
CONFIG_SOURCE="/opt/openclaw/openclaw.json"
TARGETS=("us:51.83.223.88" "tr:51.83.223.88" "pkoc2:51.83.223.88" "hkuk2:152.114.193.86")
echo "🔄 Syncing OpenClaw configuration..."
for target in "${TARGETS[@]}"; do
  IFS=':' read -r name host <<< "$target"
  echo "  → Syncing to $name ($host)..."
  rsync -avz "$CONFIG_SOURCE" "$host:/opt/openclaw/openclaw.json"
  ssh "$host" "cd /opt/openclaw && docker compose restart openclaw"
  echo "  ✅ $name synced"
done
echo "🎉 Fleet sync complete"

9. Troubleshooting

OpenClaw Won't Start

OpenClaw won't start? ─┬─ Docker not running? ──→ systemctl start docker
                        ├─ Port 18789 in use? ──→ lsof -i :18789 → kill or change port
                        ├─ Config error? ──→ jq . openclaw.json
                        ├─ Redis not running? ──→ docker compose restart redis
                        ├─ Permission denied? ──→ sudo chown -R $USER:$USER /opt/openclaw
                        └─ Image pull failure? ──→ docker compose pull

WhatsApp Connection Issues

WA issues? ─┬─ QR code expired? ──→ Regenerate: delete session, restart bridge
              ├─ Bridge process dead? ──→ docker compose restart openclaw
              ├─ Rate limited? ──→ Reduce message frequency, wait 5min
              ├─ Media not sending? ──→ Check file size (max 16MB WA limit)
              └─ Group messages not working? ──→ Check groups.enabled in config

Browser Automation Failing

Browser fails? ─┬─ CamoFox port not responding? ──→ Check systemd status, restart
                  ├─ Cloudflare block? ──→ Switch to Patchright (port 9378)
                  ├─ DataDome detection? ──→ Use Patchright + residential proxy
                  ├─ Rate limited? ──→ Add delays between requests (2-5s)
                  ├─ Element not found? ──→ Refresh snapshot, use full=True
                  └─ SSL error? ──→ Update CA certs: apt install ca-certificates

LLM Provider Errors

Provider error? ─┬─ 401 ──→ Check API key in openclaw.json env block
                   ├─ 429 ──→ Rate limited, wait or switch provider
                   ├─ 500 ──→ Provider down, try fallback chain
                   ├─ Timeout ──→ Increase maxTokens or reduce input
                   └─ Invalid model ──→ Check model name in provider docs

Performance Optimization

┌──────────────────────────────────────────────────────────────┐
│ Optimization               │ Impact          │ Complexity    │
├──────────────────────────────────────────────────────────────┤
│ Redis for session caching   │ -40% latency    │ Easy          │
│ Nginx SSL termination      │ -30% TLS time   │ Easy          │
│ Docker resource limits      │ Stable perf     │ Easy          │
│ Model response caching     │ -60% repeated    │ Medium        │
│ Skill lazy-loading         │ -50% cold start │ Medium        │
│ WebSocket connection pool   │ +2x throughput   │ Medium        │
│ Redis queue for messages   │ +3x throughput   │ Hard          │
└──────────────────────────────────────────────────────────────┘

10. Best Practices & Common Pitfalls

Best Practices

Always use systemd — Never run with nohup or background &. Orphaned bridge processes break WhatsApp.

Restart both gateway and bridge together — systemctl restart hermes-gateway with KillMode=control-group

Keep skills under 40KB — Larger skills waste context budget

Use JSON mode for structured output — Reduces token waste on creative writing

Set rate limits on nginx — 30 req/min per IP prevents abuse

Use CamoFox for anti-detection — Not default browser for social media

Monitor disk space — WhatsApp sessions + media can fill disks

Backup sessions — tar czf wa-backup.tar.gz data/whatsapp-session/

Test provider failover — Disable primary provider to verify fallback chain works

Rotate API keys quarterly — Set calendar reminder for key rotation

Common Pitfalls Table

| Pitfall | Symptom | Fix | |---|---|---| | Using nohup for bridge | WhatsApp stops responding randomly | Use systemctl restart hermes-gateway | | Stale WhatsApp session | QR code loop, can't authenticate | Delete session dir, re-authenticate | | Missing API key in env | Provider returns 401 | Add key to openclaw.json env block | | Docker volume not mounted | Data lost on restart | Add volume mounts in compose file | | No nginx SSL termination | Mixed content errors, slow TLS | Configure nginx HTTPS reverse proxy | | Overloaded context window | Agent loses track mid-conversation | Reduce skills loaded, compress history | | Wrong browser for site | Bot detection, captcha | Switch CamoFox → Patchright → Scrapling | | Disk full from media | Agent stops, sessions corrupt | Monitor with cron + df -h alerts | | Orphaned Docker containers | Port conflicts, zombie processes | docker compose down && docker compose up -d | | Skills in wrong directory | Skill not discovered | Check openclaw skills list for path |

11. Security Hardening

API Key Management

# Environment variable approach (NEVER hardcode in config.json)
openclaw.json env block references env vars:
"env": {
  "CHUTES_API_KEY": "chute_xxxxx",
  "OPENROUTER_API_KEY": "sk-or-v1-xxxxx"
}
For production, use Docker secrets or vault:
echo "chute_xxxxx" | docker secret create chutes_api_key -
Rotate keys quarterly
Calendar reminder: Q1=Jan, Q2=Apr, Q3=Jul, Q4=Oct

Input Sanitization & Validation

def validate_openclaw_config(config: dict) -> list[str]:
    """Validate OpenClaw configuration for security issues."""
    errors = []
    
    # Check for hardcoded API keys in non-env sections
    config_str = json.dumps(config)
    api_key_patterns = [
        r'sk-[a-zA-Z0-9]{20,}',  # OpenAI-style keys
        r'cpk_[a-zA-Z0-9]{20,}',  # Chutes keys
        r'gsk_[a-zA-Z0-9]{20,}',  # Groq keys
    ]
    for pattern in api_key_patterns:
        import re
        if re.search(pattern, config_str):
            # Ignore env block references
            matches = re.findall(pattern, config_str)
            for m in matches:
                # Check if it's outside the env block
                if '"env"' not in config_str[:config_str.index(m)]:
                    errors.append(f"Hardcoded API key found: {m[:8]}...")
    
    # Check password strength
    auth = config.get("auth", {})
    if auth.get("password") and len(auth["password"]) < 8:
        errors.append("Auth password too short (min 8 chars)")
    
    # Check for open CORS
    # Validate SSL config
    # Check rate limits
    
    return errors

Container Security

# docker-compose.yml security hardening
services:
  openclaw:
    image: openclaw/openclaw:latest
    security_opt:
      - no-new-privileges:true
    read_only: true
    tmpfs:
      - /tmp
      - /app/data
    cap_drop:
      - ALL
    cap_add:
      - NET_BIND_SERVICE
    deploy:
      resources:
        limits:
          cpus: '4'
          memory: 8G
        reservations:
          cpus: '2'
          memory: 4G
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:18789/health"]
      interval: 30s
      timeout: 10s
      retries: 3

Network Security

# Firewall rules (ufw)
ufw default deny incoming
ufw default allow outgoing
ufw allow 22/tcp       # SSH
ufw allow 80/tcp       # HTTP (for certbot)
ufw allow 443/tcp      # HTTPS (nginx)
ufw deny 18789/tcp     # Block direct OpenClaw access
ufw deny 9377:9378/tcp # Block direct browser access
ufw deny 3306/tcp      # Block direct DB access
ufw deny 6379/tcp      # Block direct Redis access
Verify
ufw status numbered

12. Testing & Quality Assurance

Smoke Test Suite

#!/bin/bash
smoke-test.sh — OpenClaw deployment verification
echo "🔍 OpenClaw Smoke Tests"
echo "========================"
Test 1: Gateway health
echo -n "1. Gateway health... "
STATUS=$(curl -sf http://localhost:18789/health | jq -r '.status' 2>/dev/null)
[ "$STATUS" = "ok" ] && echo "✅" || echo "❌ (status: $STATUS)"
Test 2: LLM provider connectivity
echo -n "2. LLM provider... "
RESPONSE=$(curl -sf http://localhost:18789/api/chat \
  -H "Content-Type: application/json" \
  -d '{"message":"ping","model":"chutes/zai-org/GLM-5-Turbo"}' 2>/dev/null)
[ -n "$RESPONSE" ] && echo "✅" || echo "❌"
Test 3: Skill loading
echo -n "3. Skill count... "
COUNT=$(curl -sf http://localhost:18789/api/skills 2>/dev/null | jq 'length')
[ "$COUNT" -gt 0 ] && echo "✅ ($COUNT skills)" || echo "❌"
Test 4: Redis connectivity
echo -n "4. Redis... "
docker exec openclaw-redis redis-cli ping 2>/dev/null | grep -q PONG && echo "✅" || echo "❌"
Test 5: Nginx SSL
echo -n "5. SSL termination... "
curl -sf https://openclaw.munaf.oc.s4s.host/health >/dev/null 2>&1 && echo "✅" || echo "❌"
Test 6: Browser automation
echo -n "6. CamoFox... "
curl -sf http://localhost:9377/health >/dev/null 2>&1 && echo "✅" || echo "❌"
echo "========================"
echo "Smoke tests complete"

Load Testing

# Basic load test with hey (HTTP load generator)
hey -n 100 -c 10 -m POST \
  -H "Content-Type: application/json" \
  -d '{"message":"Hello, how are you?","model":"chutes/zai-org/GLM-5-Turbo"}' \
  http://localhost:18789/api/chat
Expected results:
- p95 latency < 5s (for simple queries)
- 0% error rate
- Throughput > 10 req/s