WeCom — Company AI Assistant

WeCom — Company AI Assistant

If your company lives inside WeCom (Enterprise WeChat), you probably spend all day in it — chatting with colleagues, sharing files, and coordinating projects. Wouldn’t it be nice if your AI assistant lived right there too?

Good news: Hermes now connects directly to WeCom, so you can talk to your AI from the same app you already use for work. No extra tabs, no switching contexts. Just open a chat and ask.

What’s New in This Version

The WeCom adapter got a major upgrade. Here’s what changed:

• Scan-to-create setup — Hermes can now create your bot automatically. Just run one command and scan a QR code with your phone.
• Native streaming — replies appear token-by-token in a single bubble, with a typing indicator, just like chatting with a human.
• Per-group sender allowlists — you can now control exactly who can talk to the bot inside each group.
• Stream keepalive — for long-running tasks, Hermes sends keepalive frames so the reply stream doesn’t time out.

Why WeCom?

WeCom is Tencent’s enterprise messaging platform — think of it as Slack or Teams, but deeply integrated with WeChat. The best part? Hermes connects through WeCom’s AI Bot WebSocket gateway, which means:

• No public endpoint needed — no opening ports, no webhook configuration, no firewall headaches.
• Real-time bidirectional communication — messages flow both ways instantly over a persistent connection.
• Auto-reconnect — if the connection drops, Hermes backs off and reconnects automatically.

Getting Started

Step 1: Create Your AI Bot (The Easy Way)

Run this command:

hermes gateway setup

Select WeCom, and Hermes will display a QR code in your terminal. Scan it with your WeCom mobile app, and the wizard will:

1. Create a bot application with the correct permissions
2. Automatically grab the Bot ID and Secret
3. Guide you through access control setup

That’s it. No clicking through admin consoles.

Step 2: Configure Hermes

If you prefer manual setup, add this to ~/.hermes/.env:

WECOM_BOT_ID=your-bot-id
WECOM_SECRET=your-secret

# Optional: restrict who can DM the bot
WECOM_ALLOWED_USERS=user_id_1,user_id_2

# Optional: home channel for cron/notifications
WECOM_HOME_CHANNEL=chat_id

Step 3: Start the Gateway

hermes gateway

Done. Your AI is now live in WeCom.

Access Control That Makes Sense

You probably don’t want everyone in your company pinging the bot. Hermes gives you three levels of control:

DM Policy — who can send direct messages:

Value	What it means
open	Anyone can DM (default)
allowlist	Only users in allow_from
disabled	No DMs at all
pairing	Setup mode

Group Policy — which groups the bot responds in:

Value	What it means
open	All groups (default)
allowlist	Only groups in group_allow_from
disabled	No groups

Per-Group Sender Allowlists — the new fine-grained control. You can now say “this bot works in the engineering group, but only Alice and Bob can talk to it”:

platforms:
  wecom:
    enabled: true
    extra:
      bot_id: "your-bot-id"
      secret: "your-secret"
      group_policy: "allowlist"
      group_allow_from:
        - "group_id_1"
        - "group_id_2"
      groups:
        group_id_1:
          allow_from:
            - "user_alice"
            - "user_bob"
        "*":
          allow_from:
            - "user_admin"

What the Bot Can Do

Once connected, your WeCom AI can:

• Send and receive images, files, voice, and video — with automatic AES decryption for inbound attachments
• Render Markdown — rich text responses, not plain text walls
• Preserve reply threads — quote context keeps conversations coherent
• Stream responses natively — the thinking bubble appears immediately, and the reply renders token-by-token

Message Limits

Keep message length in mind when sending:

Action	What it does	Limit
send	Markdown text messages	4000 chars per message; longer text is split across messages

So you don’t have to worry about trimming long replies yourself — Hermes breaks them into multiple messages automatically once they pass 4000 characters.

Streaming and Typing Indicators

Native streaming follows the global streaming switch, which is off by default. Turn it on with streaming.enabled: true in config.yaml. WeCom’s per-platform display.platforms.wecom.streaming (default true) only applies while the global switch is on; set it to false to keep single-shot delivery on WeCom.

A Note on Python Dependencies

Python dependency commands on this page use a PM-prepared source checkout. After a dependency change, reactivate the checkout and restart Hermes.

WeCom needs a couple of Python packages. If something is missing, hermes pm repair usually fixes it in one shot:

Symptom	Fix
WeCom startup failed: aiohttp not installed	Install aiohttp: python -c "import pm; pm.sync_venv(['messaging'], explicit=True)"
WeCom startup failed: httpx not installed	Install httpx: hermes pm repair
Media decryption fails	Install cryptography: hermes pm repair
cryptography is required for WeCom media decryption	The inbound media is AES-encrypted. Install: hermes pm repair

One Thing to Keep in Mind

Keep your Bot Secret private. Anyone with it can impersonate your bot. Treat it like a password — don’t commit it to git, don’t share it in chats.

Ready to Try?

If your team lives in WeCom, this is the easiest way to bring AI into your daily workflow. Run hermes gateway setup, scan the QR code, and start chatting with your company AI assistant in minutes.