Installation

Requires Python 3.11 or later.

$ uv add pureskillgg-dsdk

The s3_xgboost model type needs the xgboost extra:

$ uv add "pureskillgg-dsdk[xgboost]"

Reading tomes into polars (get_dataframe(library="polars") and scan()) needs the polars extra:

$ uv add "pureskillgg-dsdk[polars]"

Usage

Read CSDS channels into DataFrames:

from pureskillgg_dsdk import DsReaderS3, GameDsLoader

reader = DsReaderS3(bucket="my-csds-bucket", manifest_key="path/to/match/csds")
loader = GameDsLoader(reader=reader)

deaths = loader.get_channel(
    {"channel": "player_death", "columns": ["round", "tick", "attacker_id"]}
)
data = loader.get_channels([{"channel": "header"}, {"channel": "round_end"}])

DsReaderFs(root_path=..., manifest_key=...) reads the same layout from disk. Leave out columns to read every column.

Consume an SQS queue:

import structlog
from pureskillgg_dsdk import SqsConsumer

async def handler(content, metadata):
    # content is the message's parsed JSON body.
    return True  # Delete the message. False leaves it for the redrive policy.

consumer = SqsConsumer(
    queue="my-queue", handler=handler, log=structlog.get_logger(), concurrency=4
)
consumer.run()  # Blocks. consumer.start() schedules it on a running loop.

Modules

pureskillgg-csgo-dsdk uses this package in its tests only.

Documentation

• docs/sqs-consumer.md: the SqsConsumer handler contract, concurrency, shutdown and error routing.

• docs/tome-data-model.md: how tomes are laid out, paged and resumed.

Development

You need Python 3, uv and Git LFS (the test fixtures are in LFS).

$ git clone https://github.com/pureskillgg/dsdk.git
$ cd dsdk
$ git lfs install
$ git lfs pull
$ uv sync

The tasks are in the Makefile: make lint, make test, make watch (tests on every change) and make format.

License

MIT. See LICENSE.txt.