Developer Builds Water-Tracking Telegram Bot Using Cloudflare's Serverless Stack

When faced with inconsistent water intake tracking, developer Prabhu Murthy opted for an unconventional solution: building a Telegram bot using entirely serverless technologies. The resulting Drinky bot demonstrates how Cloudflare's edge computing stack can power real-time applications with minimal infrastructure overhead.

Why Telegram + Serverless?

Murthy rejected traditional web/mobile apps due to login friction and maintenance concerns. Telegram's bot API offered immediate cross-platform access, while Cloudflare Workers provided globally distributed execution without server management. The critical innovation lies in using Durable Objects - Cloudflare's strongly consistent storage layer - to create isolated databases per user with sub-millisecond latency.

"Each user's hydration data lives in its own Durable Object instance," Murthy explained. "This eliminates cold starts compared to D1 SQLite and removes network hops since compute and storage colocate."

Technical Implementation Details

• grammY Framework: Chosen over Telegraf.js for active maintenance and comprehensive documentation
• Dynamic Command Registration: Implemented Telegram's setMyCommands API to avoid manual BotFather updates
• Timezone Handling: Used tz-lookup instead of overcomplicating with LLMs
• Testing Stack: Combined Vitest with Cloudflare's Miniflare for local worker testing

Key Challenges Solved

1. Webhook Configuration: Required precise setup of Telegram's webhook system using Cloudflare Tunnel for local development
2. Stateless Handling: Adapted grammY's serverless integration to work with Workers' request-based execution model
3. Data Isolation: Leveraged Durable Objects' automatic sharding to prevent user data cross-contamination

Broader Implications

This architecture demonstrates how serverless platforms now enable:

• Per-user database isolation at scale
• Zero-provisioning data persistence
• Global low-latency access without regional deployments

The complete source code showcases practical implementation of these cloud-native patterns, providing a template for other notification-based applications needing persistent user state.

"The stack costs pennies to operate while handling spikes gracefully," Murthy noted. "It's changed how I approach personal projects - no more overengineering before validating utility."

As serverless platforms mature, such experiments reveal how developers can leverage distributed systems primitives previously accessible only to large engineering teams.