Anystat

English | Русский

Lightweight, privacy-first analytics for Telegram bots built with aiogram 3.

Add two lines of code and see how people actually use your bot: /start sources and deep links, command usage, button clicks, blocks and unblocks — all in your Anystat dashboard.

anystat = Anystat(api_key="YOUR_API_KEY")
setup_anystat(dp, anystat)

Why Anystat

• Two-line integration. One middleware, zero changes to your handlers.
• Privacy by default. Message text is not collected unless you explicitly opt in. No user profile data is collected.
• Nothing hidden. Turn on debug=True and the SDK logs every event it captures, every payload it sends, and everything it skips — verify it yourself.
• Fire and forget. Events are batched, retried with backoff, and never block your handlers. If analytics fails, your bot keeps working.
• Fully typed. Ships with py.typed — autocomplete and type checking out of the box.

Installation

pip install anystat

or with uv:

uv add anystat

Requires Python 3.12+ and aiogram 3.

Quickstart

import asyncio
from aiogram import Bot, Dispatcher
from anystat import Anystat, setup_anystat


async def main():
    bot = Bot(token="YOUR_BOT_TOKEN")
    dp = Dispatcher()

    anystat = Anystat(api_key="YOUR_ANYSTAT_API_KEY")   # 1
    setup_anystat(dp, anystat)                          # 2
    dp.shutdown.register(anystat.close)  # flush pending events on shutdown

    # ... register your handlers as usual ...

    await dp.start_polling(bot)


asyncio.run(main())

That's it. Auto-tracking starts immediately — no changes to your handlers required.

You can also provide the API key via the ANYSTAT_API_KEY environment variable and create the client with just Anystat().

What gets tracked

Every event includes the user ID, a timestamp, and how long your handler took to process the update (in ms).

What is NOT collected

• Message text — unless you explicitly set track_messages=True.
• User profile data (username, name, language) — profile collection is disabled in this version.
• Anything else. Don't take our word for it — run with debug=True and watch every byte that leaves your bot.

Custom events

Track anything that matters to your bot with track():

from aiogram import types
from aiogram.filters import Command


@dp.message(Command("buy"))
async def buy_handler(message: types.Message):
    # ... your logic ...
    await anystat.track(
        "purchase",
        user_id=message.from_user.id,
        amount=99,
        currency="USD",
    )

The first argument is the event name, the second is the user ID (or None for system events). Any extra keyword arguments become event properties — pass anything JSON-serializable.

Configuration

Pass options directly:

anystat = Anystat(
    api_key="YOUR_API_KEY",
    track_messages=True,
    debug=True,
)

or group them in a config object:

from anystat import Anystat, AnystatConfig

config = AnystatConfig(track_messages=True, debug=True)
anystat = Anystat(api_key="YOUR_API_KEY", config=config)

Options passed directly to Anystat(...) override the config.

Debug mode: see exactly what leaves your bot

Trust, but verify. With debug=True the SDK logs every event it captures, every event it skips (and why), and the exact payload of every request:

[anystat] endpoint: https://api.anystat.me | api key: any_…7f2c
[anystat] auto-tracking: /start=on commands=on callbacks=on messages=off
[anystat] message text is NOT collected (track_messages=off)
[anystat] capture start_command via AnystatMiddleware:
{
  "event_type": "start_command",
  "user_id": 12345,
  "received_at": 1767225600,
  "duration": 12,
  "message_id": 42,
  "start_param": "campaign_x"
}
[anystat] skip message from user=12345 (track_messages=off, text not collected)
[anystat] → POST /v1/collect/events (2 events)
[anystat] ← 200 in 143 ms

Debug output goes through the standard logging module under the "anystat" logger. If you've already configured logging in your app, Anystat respects your handlers and formatting; otherwise it sets up a minimal [anystat] console handler for you.

Reliability

Analytics should never be the reason your bot breaks:

• Batching. Events are buffered in memory and sent in batches (up to 30 events or every 60 seconds), not one HTTP request per update.
• Retries. Network errors and retryable status codes (429, 5xx, …) are retried with exponential backoff and jitter.
• Non-blocking. Tracking runs after your handler finishes, so it never delays replies to users.
• Fail-safe. If the Anystat API is unreachable after retries, events are dropped with a warning in the logs — your bot keeps running.

Call await anystat.close() on shutdown (or register it as shown in the quickstart) to flush any buffered events before the process exits.

Requirements

• Python 3.12+
• aiogram >= 3.28

License

MIT