Skip to content

Telegram Bot

The holmes-telegram process connects a Telegram bot to a running HolmesGPT HTTP server. It keeps per-chat conversation context in memory and forwards questions to /api/chat.

Setup

  1. Create a bot with BotFather and copy its token.
  2. Start the HolmesGPT HTTP server.
  3. Run the Telegram adapter:
export TELEGRAM_BOT_TOKEN="<bot token>"
export HOLMES_API_URL="http://localhost:8080"
export TELEGRAM_ALLOWED_CHAT_IDS="123456789,-1001234567890"
holmes-telegram

TELEGRAM_ALLOWED_CHAT_IDS is optional. When it is set, messages from other private chats and groups are rejected. Use /reset in Telegram to clear the in-memory conversation context for that chat.

The adapter uses Telegram long polling, so it does not require a public webhook endpoint. Run only one polling adapter for a bot token.

Kubernetes with Helm

Configure Telegram in Helm values in the same way as other messaging integrations: keep the bot token in a Kubernetes Secret and reference it from the chart.

kubectl create secret generic holmes-telegram \
  --from-literal=bot-token="<bot token>" \
  --namespace holmes

telegram:
  enabled: true
  existingSecret:
    name: holmes-telegram
    key: bot-token
  allowedChatIds:
    - 123456789
    - -1001234567890

The chart runs one Telegram polling replica independently of the Holmes API replica count. By default it connects to the in-cluster Holmes Service created by the same Helm release. Use telegram.holmesApiUrl only when the bot should connect to a different Holmes server.

Optional settings include pollTimeoutSeconds, requestTimeoutSeconds, historyMessages, additionalEnvVars, resources, nodeSelector, tolerations, affinity, and podAnnotations.