discord.py

A modern, easy to use, feature-rich, and async ready API wrapper for Discord written in Python.

Key Features

• Modern Pythonic API using async and await.
• Proper rate limit handling.
• Optimised in both speed and memory.

Installing

Python 3.8 or higher is required

To install the library without full voice support, you can just run the following command:

Note

A Virtual Environment is recommended to install the library, especially on Linux where the system Python is externally managed and restricts which packages you can install on it.

# Linux/macOS
python3 -m pip install -U discord.py

# Windows
py -3 -m pip install -U discord.py

Otherwise to get voice support you should run the following command:

# Linux/macOS
python3 -m pip install -U "discord.py[voice]"

# Windows
py -3 -m pip install -U discord.py[voice]

To install the development version, do the following:

$ git clone https://github.com/Rapptz/discord.py
$ cd discord.py
$ python3 -m pip install -U .[voice]

Optional Packages

• PyNaCl (for voice support)

Please note that when installing voice support on Linux, you must install the following packages via your favourite package manager (e.g. apt, dnf, etc) before running the above commands:

• libffi-dev (or libffi-devel on some systems)
• python-dev (e.g. python3.8-dev for Python 3.8)

Quick Example

import discord

class MyClient(discord.Client):
    async def on_ready(self):
        print('Logged on as', self.user)

    async def on_message(self, message):
        # don't respond to ourselves
        if message.author == self.user:
            return

        if message.content == 'ping':
            await message.channel.send('pong')

intents = discord.Intents.default()
intents.message_content = True
client = MyClient(intents=intents)
client.run('token')

Bot Example

import discord
from discord.ext import commands

intents = discord.Intents.default()
intents.message_content = True
bot = commands.Bot(command_prefix='>', intents=intents)

@bot.command()
async def ping(ctx):
    await ctx.send('pong')

bot.run('token')

You can find more examples in the examples directory.

Fork changes: caching

This fork adds cache eviction that upstream discord.py does not have. All of it lives in discord/cache.py; the upstream files only carry a few hook lines marked # Maki fork: cache layer. Without the cache option the library behaves exactly like upstream.

What changed

• cache=discord.CacheSettings(...) is accepted by Client, AutoShardedClient, commands.Bot and commands.AutoShardedBot.
• Members inside a guild are evicted after member_ttl seconds without activity, or beyond member_max entries per guild. The bot's own member is never evicted.
• The message cache is keyed by id (O(1) lookups) and evicts messages after message_ttl seconds. max_messages still bounds its size and on a busy bot is usually what evicts first.
• A background sweeper runs every sweep_interval seconds (default 300).

Nothing else is bounded. Guild, channel, role, emoji, sticker, thread and voice state caches behave as upstream. Users are held weakly by upstream already and die with their last member or message.

Setup

import discord
from discord.ext import commands

intents = discord.Intents.default()
intents.members = True

bot = commands.Bot(
    command_prefix='!',
    intents=intents,
    chunk_guilds_at_startup=False,
    max_messages=50_000,        # size this to the delete log window you want
    cache=discord.CacheSettings(
        member_ttl=24 * 3600,   # members untouched for a day are dropped
        member_max=10_000,      # and at most 10000 members per guild
        message_ttl=6 * 3600,
    ),
)

Caveats

• An evicted member or message behaves like one that was never cached: the raw event still fires, the richer one does not. on_member_update, on_member_remove, on_message_edit and on_message_delete only fire for objects still in memory.
• max_messages is a count shared by every guild on the process, not a duration. Set it from your MESSAGE_CREATE rate times the window you want; message_ttl only trims quiet processes.
• Guild.chunked compares member_count with the cached member count and is not meaningful once members are evicted. Use chunk_guilds_at_startup=False with member_ttl.
• Message objects keep their Member/User alive until the message itself is evicted.
• Nothing persists across restarts; the cache starts cold like upstream.

Updating from upstream

Where the fork lives

• All logic: discord/cache.py and tests/test_cache.py. Upstream never touches these.
• Hook lines inside upstream files, every one marked # Maki fork: cache layer: discord/state.py (7), discord/client.py (3), discord/__init__.py (1). 11 in total; tests/test_cache.py asserts those numbers.
• discord/guild.py, discord/gateway.py and discord/ext/commands/bot.py are identical to upstream and must stay so.
• Also fork-only: this README section and the CacheSettings entry in docs/api.rst.

Procedure

Merge, never rebase.

git remote add upstream https://github.com/Rapptz/discord.py.git   # once
git fetch upstream
git checkout master
git merge upstream/master

Resolving a conflict: keep upstream's version of the surrounding code, then put the marked line(s) back. Never drop a marked line to make a conflict go away. Each hook is a single call into self._cache.

Checklist after every merge

git diff upstream/master -- discord/guild.py discord/gateway.py discord/ext/commands/bot.py   # must print nothing
python -m pytest -q
python -m pyright discord/cache.py discord/state.py discord/client.py
ruff format --check

test_fork_marker_count fails if a hook line was dropped. test_message_cache_supports_every_state_usage fails if upstream started calling something on self._messages that MessageCache does not provide; add the method to MessageCache when that happens. When a hook is intentionally added or removed, update the counts above and in the test in the same commit.

Links

• Documentation
• Official Discord Server
• Discord API