Chatwoot self-hosted: customer service platform with AI

IRC-Mania Die Seite rund um Sicherheit und Programmierung

Internet, Sicherheit , Internet Relay Chat

Gesponsert

What this article covers

• What Chatwoot is: open-source customer service with omnichannel support.
• Docker setup and channel configuration (website, email, WhatsApp).
• AI integration: Ollama answers tickets automatically or drafts responses.
• Agent workflows, automation, labels, and routing.
• Comparison to Rocket.Chat Omnichannel and Intercom.

Introduction

Chatwoot is the open-source alternative to Intercom and Zendesk: a customer service platform that consolidates all communication channels into a single inbox. Website chat widget, email, WhatsApp Business, Telegram, Facebook, even phone integrations. Agents handle everything from one interface.

For AI self-hosters, Chatwoot is the ideal frontend for AI-powered support: messages arrive, Ollama drafts a response, an agent reviews and sends it (or AI responds automatically for simple cases). Human-in-the-loop by design.

Typical use cases

• AI-first-level support: AI answers standard questions, humans take complex cases.
• Omnichannel inbox: website, WhatsApp, email all land in one queue.
• Draft responses: agents receive AI-generated drafts for review (copilot mode).
• Auto-categorization: AI tags tickets (bug, billing, inquiry) and routes them.
• Bot handoff: Typebot or your own bot escalates to Chatwoot.

Installation: Docker

Chatwoot is a Rails app with Postgres and Redis:

services:
  chatwoot:
    image: chatwoot/chatwoot:latest
    ports: ["127.0.0.1:3000:3000"]
    environment:
      - SECRET_KEY_BASE=langer-zufaelliger-key
      - FRONTEND_URL=https://support.deine-domain.de
      - POSTGRES_HOST=postgres
      - POSTGRES_USERNAME=chatwoot
      - POSTGRES_PASSWORD=geheim
      - POSTGRES_DATABASE=chatwoot
      - REDIS_URL=redis://redis:6379
      - MAILER_SENDER_EMAIL=support@deine-domain.de
      - SMTP_ADDRESS=smtp.dein-server.de
      - SMTP_PORT=587
      - SMTP_USERNAME=...
      - SMTP_PASSWORD=...
    volumes: [chatwoot_data:/app/storage]
    depends_on: [postgres, redis]
    restart: always

  chatwoot-worker:
    image: chatwoot/chatwoot:latest
    command: bundle exec sidekiq -C config/sidekiq.yml
    environment: *same-env-vars*
    volumes: [chatwoot_data:/app/storage]
    depends_on: [postgres, redis]
    restart: always

  postgres:
    image: postgres:16
    environment:
      POSTGRES_USER: chatwoot
      POSTGRES_PASSWORD: geheim
      POSTGRES_DB: chatwoot
    volumes: [pg_data:/var/lib/postgresql/data]
    restart: always

  redis:
    image: redis:7
    volumes: [redis_data:/data]
    restart: always

volumes:
  chatwoot_data:
  pg_data:
  redis_data:

After startup, run docker exec chatwoot bundle exec rails db:prepare for database setup, then walk through the setup wizard.

Setting up channels

• Website live chat: Inbox → Website → embed the widget snippet on your site (like Rocket.Chat Livechat, but as a dedicated support platform).
• Email: incoming mail to support@ becomes tickets.
• WhatsApp Business: via Twilio, 360dialog, or directly through WhatsApp Cloud API.
• Telegram: enter a bot token; Telegram messages become tickets.
• API channel: custom integrations via the Chatwoot API.

AI integration: Ollama as support agent

Chatwoot offers webhooks (on each new message) and a REST API (post responses). The AI agent sits in the middle:

from flask import Flask, request
import requests
import ollama

app = Flask(__name__)
CW = "https://support.deine-domain.de"
CW_TOKEN = "dein-api-access-token"
client = ollama.Client(host="http://ollama:11434")

@app.route("/webhook", methods=["POST"])
def webhook():
    data = request.json
    if data.get("event") != "message_created":
        return {"ok": True}
    if data.get("message_type") != "incoming":   # only customer messages
        return {"ok": True}

    convo_id = data["conversation"]["id"]
    text = data.get("content", "")

    # Load conversation context from Chatwoot
    msgs = requests.get(
        f"{CW}/api/v1/accounts/1/conversations/{convo_id}/messages",
        headers={"api_access_token": CW_TOKEN}).json()
    history = [{"role": "user" if m["message_type"] == 0 else "assistant",
                "content": m["content"]} for m in msgs["payload"][-10:]]

    resp = client.chat(model="llama3.1:8b", messages=[
        {"role": "system", "content":
         "Du bist Support-Agent. Antworte kurz, hilfreich. "
         "Wenn Du unsicher bist, sag 'Ich leite an einen Kollegen weiter'."}
    ] + history)
    answer = resp["message"]["content"]

    # Post response as agent
    requests.post(
        f"{CW}/api/v1/accounts/1/conversations/{convo_id}/messages",
        headers={"api_access_token": CW_TOKEN},
        json={"content": answer})
    return {"ok": True}

Copilot mode instead of auto-reply

Safer for business: AI posts a draft as a private note (visible only internally), and the agent reviews and sends it:

# private: True → visible only to staff
requests.post(url, json={"content": f"[KI-Entwurf]\n{answer}",
                         "private": True})

Automation within Chatwoot

Built-in automation rules (no code needed):

• label assignment by keywords
• auto-assign to agent teams
• SLA tracking and escalation
• canned responses (pre-written templates)

Extensible with AI webhooks: AI tags → automation routes.

Security and data protection

• GDPR: on-premise means customer data stays with you, the main advantage over Intercom and Zendesk.
• Customer data: support chat contains personal information. Use HTTPS and restrict agent access.
• API tokens: use separate tokens per integration.
• AI transparency: customers should know if they’re talking to AI or a human. Make this clear in the bot name or introduction.

Chatwoot vs. alternatives

Further resources

• IRC-Coding.de: programming tutorials on webhooks and API integration.
• Typebot: bot escalation to Chatwoot.
• Rocket.Chat: alternative with omnichannel.
• WhatsApp Bots: WhatsApp channel in detail.
• n8n: workflow automation.

Key takeaways:

• Chatwoot = open-source Intercom: omnichannel support (website, email, WhatsApp) in one inbox.
• Webhooks + API mean Ollama can answer tickets or draft responses for agents.
• Human-in-the-loop: copilot mode (AI draft, agent review) is the safe approach.
• Self-hosted means customer data stays with you (GDPR compliance).
• The best open-source choice for real customer service with AI.

FAQ

Sources and further reading

• Chatwoot: GitHub.
• Chatwoot Docs: Documentation.
• IRC-Coding.de: Programming tutorials.

Premium VPS Hosting

Schnelle und sichere VPS Server

Gesponsert