New ratelimiting implementation

[davfsa]

Hey!

This pull request adds support for nirn to utilize a sliding window technique with regards to ratelimits, which allows for a higher throughput and (optionally, which can be enabled through ALLOW_CONCURRENT_REQUESTS=true) the ability to perform concurrent requests to the same bucket!

I also took the liberty to update all the dependencies to their latest version, as of writing, and to update the usage of some deprecated apis.

[davfsa]

Some examples:

2025-06-22.23-49-58.mp4

Script

from __future__ import annotations

import asyncio
import os

import dotenv

import hikari

dotenv.load_dotenv()

CHANNEL_ID = 

bot = hikari.GatewayBot(
    logs={
        "version": 1,
        "incremental": True,
        "loggers": {"": {"level": "INFO"}, "hikari.ratelimits": {"level": "DEBUG"}},
    },
    rest_url="http://localhost:8080/api/v10",
    max_retries=0,
    cache_settings=hikari.impl.CacheSettings(
        components=hikari.impl.CacheComponents.NONE
    ),
    token=os.environ["TOKEN"],
)


async def spam() -> None:
    while True:
        await bot.rest.create_message(CHANNEL_ID, "Hello!")


@bot.listen()
async def started(_: hikari.StartedEvent) -> None:
    await asyncio.gather(*(spam() for _ in range(5)))


bot.run()

---

2025-06-23.01-13-09.mp4

Script

from __future__ import annotations

import asyncio
import os

import dotenv

import hikari

dotenv.load_dotenv()

bot = hikari.GatewayBot(
    logs={
        "version": 1,
        "incremental": True,
        "loggers": {"": {"level": "INFO"}, "hikari.ratelimits": {"level": "DEBUG"}},
    },
    rest_url="http://localhost:8080/api/v10",
    max_retries=0,
    cache_settings=hikari.impl.CacheSettings(
        components=hikari.impl.CacheComponents.NONE
    ),
    token=os.environ["HIKARI_TOKEN"],
)

CHANNEL_ID =
BOT_ID =
message_id = 0


async def spam() -> None:
    while True:
        await bot.rest.add_reaction(CHANNEL_ID, message_id, "\u2764\ufe0f")
        await bot.rest.delete_my_reaction(CHANNEL_ID, message_id, "\u2764\ufe0f")


@bot.listen()
async def test(event: hikari.ReactionAddEvent) -> None:
    if event.message_id != message_id or event.user_id == BOT_ID:
        return

    if event.emoji_id:
        await bot.rest.delete_reaction(
            event.channel_id,
            event.message_id,
            event.user_id,
            emoji=event.emoji_name,
            emoji_id=event.emoji_id,
        )
    else:
        await bot.rest.delete_reaction(
            event.channel_id, event.message_id, event.user_id, emoji=event.emoji_name
        )


@bot.listen()
async def started(_: hikari.StartedEvent) -> None:
    message = await bot.rest.create_message(
        CHANNEL_ID, "Spam reactions on me, I will delete them :D"
    )
    await message.add_reaction("\u2764")

    global message_id
    message_id = message.id

    asyncio.create_task(spam())


bot.run()

---

Note:, hikari was patched to comment out these two lines https://github.com/hikari-py/hikari/blob/f07f4187b9e6d284e334cb913625b2988217a13f/hikari/impl/rest.py#L802-L803 to allow nirn to fully take over handling ratelimiting