State & Storage

Bots often need persistent state:

• User preferences.
• Game progress.
• Karma counts.
• Custom commands.
• Scheduled tasks.

Options:

In-memory:

• dict, list.
• Lost on restart.
• Fine for development.

Local file (JSON, SQLite):

• Simple. Survives restart.
• Single instance only.
• Good for small bots.

Database (Postgres, MongoDB, Redis):

• Multiple instances supported.
• Robust persistence.
• Standard for production.

Conversation state:

• Multi-step flows: bot asks → user replies → bot asks more.
• State machine per user.
• Track current step + context.

Modern: state machines (Telegram Bot frameworks do this), or LLM-based context management.

Caching:

• API calls expensive (rate limited).
• Cache user info, channel info.
• TTL: ~5 minutes typical.

Background tasks:

• Scheduled: every X minutes (cron-like).
• Examples: clean stale data, post weekly stats.

Run as separate task in event loop.

Distributed bots:

• Multiple bot instances for HA.
• Shared state: database.
• Coordination: lock primitives, leader election.
• Most bots run single instance; sharding only for huge.