vkworkspace 1.2.1

Fully async framework for building VK Teams (VK Workspace) bots, inspired by aiogram 3

vkworkspace

Fully asynchronous framework for building bots on VK Teams (VK Workspace / VK SuperApp), inspired by aiogram 3.

A modern replacement for the official mail-ru-im/bot-python (mailru-im-bot) library.

README на русском языке

vkworkspace vs mailru-im-bot

The official mailru-im-bot library is synchronous, based on requests, and has not been updated in years. vkworkspace is a ground-up async rewrite with a modern developer experience.

Feature Comparison

	mailru-im-bot	vkworkspace
HTTP client	requests (sync)	httpx (async)
Models	raw dict	Pydantic v2
Router / Dispatcher	—	aiogram-style
Magic filters (F)	—	F.text, F.chat.type == "private"
FSM (state machine)	—	Memory + Redis backends
Middleware pipeline	—	inner/outer, per-event type
Inline keyboard builder	—	InlineKeyboardBuilder
Rate limiter	—	built-in token-bucket
Proxy support	—	proxy= parameter
Handler DI	—	auto-inject bot, state, command
Custom command prefixes	—	prefix=("/", "!", "")
Error handlers	—	@router.error()
Type hints	partial	full

Speed Benchmarks

Tested on Python 3.11, VK Teams corporate instance. 5 rounds per discipline, median values:

                          mailru-im-bot    vkworkspace     winner
                          (requests sync)  (httpx async)
 Bot Info (self/get)         34 ms            33 ms        ~tie
 Send Text                   82 ms           109 ms        mailru-im-bot
 Send Keyboard               97 ms            90 ms        vkworkspace
 Edit Message                57 ms            44 ms        vkworkspace
 Delete Message              33 ms            38 ms        mailru-im-bot
 Chat Info                   43 ms            42 ms        ~tie
 Send Actions                30 ms            28 ms        vkworkspace
 BURST (10 msgs)           1110 ms           335 ms        vkworkspace (3.3x!)
                          ───────────────────────────────────────────
 Score                        2                4           vkworkspace wins

Key takeaway: For single sequential requests, both libraries are within network noise (~1 ms difference). But when you need to send multiple messages or handle concurrent users, async wins decisively — 3.3x faster on burst operations via asyncio.gather().

The framework overhead is zero in real conditions. Network latency dominates 99%+ of total time.

Why async?

VK Teams bots spend 99% of their time waiting for the network. Synchronous frameworks (requests) block the entire process on every API call. vkworkspace uses httpx.AsyncClient and asyncio — while one request waits for a response, the bot handles other messages.

Features

• 100% async — httpx + asyncio, zero blocking calls, real concurrency
• Aiogram-like API — Router, Dispatcher, F magic filters, middleware, FSM
• Type-safe — Pydantic v2 models for all API types, full type hints
• Flexible filtering — Command, StateFilter, ChatTypeFilter, CallbackData, regex, magic filters
• Custom command prefixes — Command("start", prefix=("/", "!", "")) for any prefix style
• FSM — Finite State Machine with Memory and Redis storage backends
• Middleware — inner/outer middleware pipeline for logging, auth, throttling
• Text formatting — md, html helpers for MarkdownV2/HTML + split_text for long messages
• Keyboard builder — fluent API for inline keyboards with button styles
• Rate limiter — built-in request throttling (rate_limit=5 = max 5 req/sec)
• Proxy support — route API requests through corporate proxy
• Edited message routing — handle_edited_as_message=True to process edits as new messages
• Lifecycle hooks — on_startup / on_shutdown for init/cleanup logic
• Error handlers — @router.error() to catch exceptions from any handler
• Multi-bot polling — dp.start_polling(bot1, bot2) to poll multiple bots
• 9 event types — message, edited, deleted, pinned, unpinned, members join/leave, chat info, callbacks
• Python 3.11 — 3.14 support (free-threaded / no-GIL build not yet tested)

Installation

pip install vkworkspace

With Redis FSM storage:

pip install vkworkspace[redis]

Quick Start

import asyncio
from vkworkspace import Bot, Dispatcher, Router, F
from vkworkspace.filters import Command
from vkworkspace.types import Message

router = Router()

@router.message(Command("start"))
async def cmd_start(message: Message) -> None:
    await message.answer("Hello! I'm your bot.")

@router.message(F.text)
async def echo(message: Message) -> None:
    await message.answer(message.text)

async def main() -> None:
    bot = Bot(token="YOUR_TOKEN", api_url="https://myteam.mail.ru/bot/v1")
    dp = Dispatcher()
    dp.include_router(router)
    await dp.start_polling(bot)

asyncio.run(main())

API URL: Each company has its own VK Teams API endpoint. Check with your IT/integrator for the correct URL. Common examples:

• https://myteam.mail.ru/bot/v1 (Mail.ru default)
• https://api.teams.yourcompany.ru/bot/v1 (on-premise)
• https://agent.mail.ru/bot/v1 (SaaS variant)

Even SaaS deployments may have custom URLs — always verify with your administrator.

Bot Configuration

bot = Bot(
    token="YOUR_TOKEN",
    api_url="https://myteam.mail.ru/bot/v1",
    timeout=30.0,               # Request timeout (seconds)
    poll_time=60,               # Long-poll timeout (seconds)
    rate_limit=5.0,             # Max 5 requests/sec (None = unlimited)
    proxy="http://proxy:8080",  # HTTP proxy for corporate networks
    parse_mode="HTML",          # Default parse mode for all messages (None = plain text)
)

Rate Limiter

Built-in token-bucket rate limiter prevents API throttling:

# Max 10 requests per second — framework queues the rest automatically
bot = Bot(token="TOKEN", api_url="URL", rate_limit=10)

Proxy

Corporate networks often require a proxy. The proxy parameter applies only to API calls:

bot = Bot(
    token="TOKEN",
    api_url="https://api.internal.corp/bot/v1",
    proxy="http://corp-proxy.internal:3128",
)

Default Parse Mode

parse_mode can be set in two places — globally on Bot, or per-message on answer() / reply() / send_text():

from vkworkspace.enums import ParseMode

# ── 1. Global default on Bot ──────────────────────────────────────
# Applies to ALL send_text / answer / reply / edit_text / send_file
bot = Bot(
    token="TOKEN",
    api_url="https://myteam.mail.ru/bot/v1",
    parse_mode=ParseMode.HTML,  # or ParseMode.MARKDOWNV2
)

# parse_mode="HTML" is sent automatically — no need to pass it
await message.answer(f"{html.bold('Hello')}, world!")

# ── 2. Override per message ───────────────────────────────────────
await message.answer("*bold*", parse_mode=ParseMode.MARKDOWNV2)

# ── 3. Disable for a single message ──────────────────────────────
await message.answer("plain text", parse_mode=None)

Tip: IDE autocomplete works — type parse_mode=ParseMode. and your editor will suggest HTML and MARKDOWNV2.

Dispatcher

dp = Dispatcher(
    storage=MemoryStorage(),            # FSM storage (default: MemoryStorage)
    fsm_strategy="user_in_chat",        # FSM key strategy
    handle_edited_as_message=False,     # Route edits to @router.message handlers
)

# Lifecycle hooks
@dp.on_startup
async def on_start():
    print("Bot started!")

@dp.on_shutdown
async def on_stop():
    print("Bot stopped!")

# Include routers
dp.include_router(router)
dp.include_routers(admin_router, user_router)

# Start polling (supports multiple bots)
await dp.start_polling(bot)
await dp.start_polling(bot1, bot2, skip_updates=True)

Edited Message Routing

When handle_edited_as_message=True, edited messages are routed to @router.message handlers instead of @router.edited_message. Useful when you don't need to distinguish between new and edited messages:

dp = Dispatcher(handle_edited_as_message=True)

# This handler fires for BOTH new and edited messages
@router.message(F.text)
async def handle_text(message: Message) -> None:
    await message.answer(f"Got: {message.text}")

Router & Event Types

router = Router(name="my_router")

# All 9 supported event types:
@router.message(...)              # New message
@router.edited_message(...)       # Edited message
@router.deleted_message(...)      # Deleted message
@router.pinned_message(...)       # Pinned message
@router.unpinned_message(...)     # Unpinned message
@router.new_chat_members(...)     # Users joined
@router.left_chat_members(...)    # Users left
@router.changed_chat_info(...)    # Chat info changed
@router.callback_query(...)       # Button clicked

# Error handler — catches exceptions from any handler
@router.error()
async def on_error(event, error: Exception) -> None:
    print(f"Error: {error}")

# Sub-routers for modular code
admin_router = Router(name="admin")
user_router = Router(name="user")
main_router = Router()
main_router.include_routers(admin_router, user_router)

Filters

Command Filter

from vkworkspace.filters import Command

# Basic
@router.message(Command("start"))
@router.message(Command("help", "info"))     # Multiple commands

# Custom prefixes
@router.message(Command("start", prefix="/"))           # Only /start
@router.message(Command("menu", prefix=("/", "!", ""))) # /menu, !menu, menu

# Regex commands
import re
@router.message(Command(re.compile(r"cmd_\d+")))        # /cmd_1, /cmd_42, ...

# Access command arguments in handler
@router.message(Command("ban"))
async def ban_user(message: Message, command: CommandObject) -> None:
    user_to_ban = command.args  # Text after "/ban "
    await message.answer(f"Banned: {user_to_ban}")

CommandObject injected into handler:

• prefix — matched prefix ("/", "!", etc.)
• command — command name ("ban")
• args — arguments after command ("user123")
• raw_text — full message text
• match — re.Match if regex pattern was used

Magic Filter (F)

from vkworkspace import F

@router.message(F.text)                              # Has text
@router.message(F.text == "hello")                   # Exact match
@router.message(F.text.startswith("hi"))             # String methods
@router.message(F.from_user.user_id == "admin@co")   # Nested attrs
@router.message(F.chat.type == "private")            # Private chat
@router.message(F.chat.type.in_(["private", "group"]))
@router.callback_query(F.callback_data == "confirm")

State Filter

from vkworkspace.filters.state import StateFilter

@router.message(StateFilter(Form.name))     # Specific state
@router.message(StateFilter("*"))           # Any state (non-None)
@router.message(StateFilter(None))          # No state (default)

Other Filters

from vkworkspace.filters import CallbackData, ChatTypeFilter, RegexpFilter

# Callback data
@router.callback_query(CallbackData("confirm"))
@router.callback_query(CallbackData(re.compile(r"^action_\d+$")))

# Chat type
@router.message(ChatTypeFilter("private"))
@router.message(ChatTypeFilter(["private", "group"]))

# Regex on message text
@router.message(RegexpFilter(r"\d{4}"))

# Combine filters with &, |, ~
@router.message(ChatTypeFilter("private") & Command("secret"))

Custom Filters

from vkworkspace.filters.base import BaseFilter

class IsAdmin(BaseFilter):
    async def __call__(self, event, **kwargs) -> bool:
        admins = await event.bot.get_chat_admins(event.chat.chat_id)
        return any(a.user_id == event.from_user.user_id for a in admins)

@router.message(IsAdmin())
async def admin_only(message: Message) -> None:
    await message.answer("Admin panel")

Text Formatting

VK Teams supports MarkdownV2 and HTML formatting. Use the md and html helpers to build formatted messages safely:

from vkworkspace.utils.text import md, html, split_text

# ── MarkdownV2 ──
text = f"{md.bold('Status')}: {md.escape(user_input)}"
await message.answer(text, parse_mode="MarkdownV2")

md.bold("text")            # *text*
md.italic("text")          # _text_
md.underline("text")       # __text__
md.strikethrough("text")   # ~text~
md.code("x = 1")           # `x = 1`
md.pre("code", "python")   # ```python\ncode\n```
md.link("Click", "https://example.com")  # [Click](https://example.com)
md.quote("quoted text")    # >quoted text
md.mention("[e-mail]")  # @\[user@company\.ru\]
md.escape("price: $100")   # Escapes special chars

# ── HTML ──
text = f"{html.bold('Status')}: {html.escape(user_input)}"
await message.answer(text, parse_mode="HTML")

html.bold("text")           # <b>text</b>
html.italic("text")         # <i>text</i>
html.underline("text")      # <u>text</u>
html.strikethrough("text")  # <s>text</s>
html.code("x = 1")          # <code>x = 1</code>
html.pre("code", "python")  # <pre><code class="python">code</code></pre>
html.link("Click", "https://example.com")  # <a href="...">Click</a>
html.quote("quoted text")   # <blockquote>quoted text</blockquote>
html.mention("[e-mail]")  # @[[e-mail]]
html.ordered_list(["a", "b"])    # <ol><li>a</li><li>b</li></ol>
html.unordered_list(["a", "b"]) # <ul><li>a</li><li>b</li></ul>

# ── Split long text ──
# VK Teams may lag on messages > 4096 chars; split_text breaks them up
for chunk in split_text(long_text):
    await message.answer(chunk)

# Custom limit
for chunk in split_text(long_text, max_length=2000):
    await message.answer(chunk)

Text Builder (aiogram-style)

Composable formatting nodes with auto-escaping and automatic parse_mode. No need to think about which parse mode to use — as_kwargs() handles it:

from vkworkspace.utils.text import Text, Bold, Italic, Code, Link, Pre, Quote, Mention

# Compose text from nodes — strings are auto-escaped
content = Text(
    Bold("Order #42"), "\n",
    "Status: ", Italic("processing"), "\n",
    "Total: ", Code("$99.99"),
)
await message.answer(**content.as_kwargs())  # text + parse_mode="HTML" auto

# Nesting works
content = Bold(Italic("bold italic"))           # <b><i>bold italic</i></b>

# Operator chaining
content = "Hello, " + Bold("World") + "!"      # Text node
await message.answer(**content.as_kwargs())

# Render as MarkdownV2 instead
await message.answer(**content.as_kwargs("MarkdownV2"))

Available nodes: Text, Bold, Italic, Underline, Strikethrough, Code, Pre, Link, Mention, Quote, Raw

Warning: Do not mix string helpers (md.* / html.*) with node builder — raw strings inside nodes get auto-escaped, so Text(md.bold("x")) produces literal *x*, not bold. Use Bold("x") instead.

Inline Keyboards

from vkworkspace.utils.keyboard import InlineKeyboardBuilder
from vkworkspace.enums import ButtonStyle

builder = InlineKeyboardBuilder()
builder.button(text="Yes", callback_data="confirm", style=ButtonStyle.PRIMARY)
builder.button(text="No", callback_data="cancel", style=ButtonStyle.ATTENTION)
builder.button(text="Maybe", callback_data="maybe")
builder.adjust(2, 1)  # Row 1: 2 buttons, Row 2: 1 button

await message.answer("Are you sure?", inline_keyboard_markup=builder.as_markup())

# Copy & modify
builder2 = builder.copy()
builder2.button(text="Extra", callback_data="extra")

FSM (Finite State Machine)

from vkworkspace.fsm import StatesGroup, State, FSMContext
from vkworkspace.filters.state import StateFilter
from vkworkspace.fsm.storage.memory import MemoryStorage

class Form(StatesGroup):
    name = State()
    age = State()

@router.message(Command("start"))
async def start(message: Message, state: FSMContext) -> None:
    await state.set_state(Form.name)
    await message.answer("What is your name?")

@router.message(StateFilter(Form.name), F.text)
async def process_name(message: Message, state: FSMContext) -> None:
    await state.update_data(name=message.text)
    await state.set_state(Form.age)
    await message.answer("How old are you?")

@router.message(StateFilter(Form.age), F.text)
async def process_age(message: Message, state: FSMContext) -> None:
    data = await state.get_data()
    await message.answer(f"Name: {data['name']}, Age: {message.text}")
    await state.clear()

# Storage backends
dp = Dispatcher(storage=MemoryStorage())          # In-memory (dev)

from vkworkspace.fsm.storage.redis import RedisStorage
dp = Dispatcher(storage=RedisStorage())           # Redis (prod)

FSMContext methods:

• await state.get_state() — current state (or None)
• await state.set_state(Form.name) — transition to state
• await state.get_data() — get stored data dict
• await state.update_data(key=value) — merge into data
• await state.set_data({...}) — replace data entirely
• await state.clear() — clear state and data

Custom Middleware

from vkworkspace import BaseMiddleware

class LoggingMiddleware(BaseMiddleware):
    async def __call__(self, handler, event, data):
        print(f"Event from: {event.from_user.user_id}")
        result = await handler(event, data)
        print(f"Handled OK")
        return result

class ThrottleMiddleware(BaseMiddleware):
    def __init__(self, delay: float = 1.0):
        self.delay = delay
        self.last: dict[str, float] = {}

    async def __call__(self, handler, event, data):
        import time
        uid = getattr(event, "from_user", None)
        uid = uid.user_id if uid else "unknown"
        now = time.monotonic()
        if now - self.last.get(uid, 0) < self.delay:
            return None  # Skip handler — too fast
        self.last[uid] = now
        return await handler(event, data)

# Register on specific event observer
router.message.middleware.register(LoggingMiddleware())
router.message.middleware.register(ThrottleMiddleware(delay=0.5))
router.callback_query.middleware.register(LoggingMiddleware())

Message Methods

# In handler:
await message.answer("Hello!")                                # Send to same chat
await message.reply("Reply to you!")                          # Reply to this message
await message.edit_text("Updated text")                       # Edit this message
await message.delete()                                        # Delete this message
await message.pin()                                           # Pin this message
await message.unpin()                                         # Unpin this message
await message.answer_file(file=InputFile("photo.jpg"))        # Send file
await message.answer_voice(file=InputFile("audio.ogg"))       # Send voice

Bot API Methods

bot = Bot(token="TOKEN", api_url="URL")

# Self
await bot.get_me()

# Messages
await bot.send_text(chat_id, "Hello!", parse_mode="HTML")
await bot.edit_text(chat_id, msg_id, "Updated")
await bot.delete_messages(chat_id, msg_id)
await bot.send_file(chat_id, file=InputFile("photo.jpg"), caption="Look!")
await bot.send_voice(chat_id, file=InputFile("voice.ogg"))
await bot.answer_callback_query(query_id, "Done!", show_alert=True)

# Chat management
await bot.get_chat_info(chat_id)
await bot.get_chat_admins(chat_id)
await bot.get_chat_members(chat_id)
await bot.get_blocked_users(chat_id)
await bot.get_pending_users(chat_id)
await bot.set_chat_title(chat_id, "New Title")
await bot.set_chat_about(chat_id, "Description")
await bot.set_chat_rules(chat_id, "Rules")
await bot.set_chat_avatar(chat_id, file=InputFile("avatar.png"))
await bot.block_user(chat_id, user_id, del_last_messages=True)
await bot.unblock_user(chat_id, user_id)
await bot.resolve_pending(chat_id, approve=True, user_id=uid)
await bot.delete_chat_members(chat_id, members=[uid1, uid2])
await bot.pin_message(chat_id, msg_id)
await bot.unpin_message(chat_id, msg_id)
await bot.send_actions(chat_id, "typing")

# Files & Threads
await bot.get_file_info(file_id)
await bot.threads_add(chat_id, msg_id)
await bot.threads_get_subscribers(thread_id)
await bot.threads_autosubscribe(chat_id, enable=True)

Handler Dependency Injection

Handlers receive only the parameters they declare. The framework inspects the signature and injects available values:

@router.message(Command("info"))
async def full_handler(
    message: Message,              # The message object
    bot: Bot,                      # Bot instance
    state: FSMContext,             # FSM context (if storage configured)
    command: CommandObject,        # Parsed command (from Command filter)
    raw_event: dict,               # Raw VK Teams event dict
    event_type: str,               # "message", "callback_query", etc.
) -> None:
    ...

# Minimal — only take what you need
@router.message(F.text)
async def simple(message: Message) -> None:
    await message.answer(message.text)

Mini-Apps + Bots

VK Teams mini-apps (WebView apps inside the client) cannot send notifications or messages to users. The official recommendation is to use a bot as the notification channel.

Typical pattern:

1. Bot sends a message with a mini-app link (with parameters for deep-linking)
2. Mini-app opens, identifies the user via GetSelfId / GetAuth (JS Bridge)
3. Mini-app communicates with its own backend (can be the same server as the bot)
4. For notifications back to the user — the bot sends messages via the Bot API

# Bot sends a deep-link to a mini-app
miniapp_url = "https://u.myteam.mail.ru/miniapp/my-app-id?order=42"
await message.answer(
    f"Open your order: {miniapp_url}",
)

Mini-apps are registered via Metabot (/newapp) — the same bot used for bot management (/newbot).

Examples

Example	Description
echo_bot.py	Basic commands + text echo
keyboard_bot.py	Inline keyboards + callback handling
fsm_bot.py	Multi-step dialog with FSM
middleware_bot.py	Custom middleware (logging, access control)
proxy_bot.py	Corporate proxy + rate limiter
error_handling_bot.py	Error handlers, lifecycle hooks, edited message routing
custom_prefix_bot.py	Custom command prefixes, regex commands, argument parsing
multi_router_bot.py	Modular sub-routers, chat events (join/leave/pin)
formatting_bot.py	MarkdownV2/HTML formatting + split_text for long messages

Project Structure

vkworkspace/
├── client/          # Async HTTP client (httpx), rate limiter, proxy
├── types/           # Pydantic v2 models (Message, Chat, User, CallbackQuery, ...)
├── enums/           # EventType, ChatType, ParseMode, ButtonStyle, ...
├── dispatcher/      # Dispatcher, Router, EventObserver, Middleware pipeline
├── filters/         # Command, StateFilter, CallbackData, ChatType, Regexp
├── fsm/             # StatesGroup, State, FSMContext, Memory/Redis storage
└── utils/           # InlineKeyboardBuilder, magic filter (F), text formatting (md, html, split_text)

Requirements

• Python 3.11+
• httpx >= 0.28.1
• pydantic >= 2.6
• magic-filter >= 1.0.12

License

MIT LICENSE.

---

_{Keywords: VK Teams bot, VK Workspace bot, VK WorkSpace, VK SuperApp, vkteams, workspace.vk.ru, workspacevk, vkworkspace, myteam.mail.ru, agent.mail.ru, mail-ru-im-bot, mailru-im-bot, bot-python, VK Teams API, VK Teams framework, async bot framework, Python bot VK Teams, aiogram VK Teams, httpx bot, Pydantic bot framework}