Lobster Gateway

Current Configuration

🤖 LLM Provider

The assistant uses jengaKit's built-in LLM proxy automatically. You can also provide your own API key.

🔒 HTTPS Tunnel

Telegram and WhatsApp webhooks require a public HTTPS URL. jengaKIT's nip.io URL is HTTP only. Configure a tunnel to get HTTPS automatically at deploy time.

✈️ Telegram Integration

Connect the assistant to Telegram so it can reply to messages in DMs or groups.

1. Open Telegram and message @BotFather
2. Send /newbot and follow the prompts
3. Copy the token (looks like 123456:ABC-DEF...)
4. In your jengaKit deployment form, paste it into "Telegram Bot Token"
5. Set the webhook URL to https://YOUR-APP-URL/webhook/telegram

🎮 Discord Integration

Connect to Discord via a bot application.

1. Go to discord.com/developers/applications → New Application
2. Go to Bot → Reset Token, copy it
3. Invite the bot to your server with Message Content Intent enabled
4. Paste the token as "Discord Bot Token" in the deploy form

💬 WhatsApp Integration

Connect via the Meta WhatsApp Cloud API. Requires a Meta Business account, a registered phone number, and a public HTTPS webhook URL.

1. Go to developers.facebook.com → My Apps → Create App → Business.
2. Add the WhatsApp product. Under API Setup, note your Phone Number ID and generate a temporary (or permanent) Access Token.
3. In the jengaKIT deploy form, paste:
   • WhatsApp Access Token — from step 2
   • WhatsApp Phone Number ID — from step 2
   • WhatsApp Webhook Verify Token — any string you choose (e.g. lobster-wa-verify)
4. After deployment, register the webhook in Meta → WhatsApp → Configuration:
   Callback URL: https://YOUR-APP-URL/webhook/whatsapp

   Verify token: <your LOBSTER_WHATSAPP_VERIFY_TOKEN>

   Subscribe to: messages
5. Send a WhatsApp message to your test number — the bot will reply within seconds.

📡 Signal Integration

Connect the assistant to Signal using a self-hosted signal-cli-rest-api instance. No HTTPS required — lobster polls for new messages every 2 seconds.

1. Run signal-cli-rest-api on any host reachable from the cluster:
   docker run -d --name signal-api -p 8080:8080 \
     -v $(pwd)/signal-config:/home/.local/share/signal-cli \
     bbernhard/signal-cli-rest-api
2. Register or link a phone number (one-time):
   # Option A — new number (SMS verification):

   curl -X POST 'http://localhost:8080/v1/register/+14155552671'

   curl -X POST 'http://localhost:8080/v1/register/+14155552671/verify/123456'

   # Option B — link existing Signal account:

   curl 'http://localhost:8080/v1/qrcodelink?device_name=lobster'

   # Scan QR in Signal app → Settings → Linked Devices
3. In the jengaKIT deploy form set Signal API URL (e.g. http://192.168.2.177:8080) and Signal Phone Number in E.164 format.
4. Set Signal Allowed Senders to your personal number to restrict access.

🧠 Personality & Memory

Customize the assistant's personality in the deploy prompt, or via env vars.

Memory entries added in the Memory tab are automatically included in every LLM prompt as context.

🤖 Runtime LLM Override

Override the LLM used for chat without redeploying. Useful for switching to a local Ollama server after a Gemini deploy. Stored in the pod database — survives restarts.

🔑 Pod SSH Public Key

This pod has its own ED25519 keypair. Add the public key below to ~/.ssh/authorized_keys on any server you want the assistant to access.

🖥️ SSH Hosts

Add hosts the assistant can SSH into. Reference them by label in the Chat tab — e.g. "What is the disk usage on homelab?"