Hugging Face's logo

Datasets:
risenlab
/
agentlogs

Modalities:
Tabular
Text
Formats:
parquet
Size:
10M - 100M
ArXiv:
Libraries:
Datasets
Dask
Polars
License:
Dataset card Data Studio Files
xet
 Community
2
agentlogs / README.md
jonan-richards's picture
jonan-richards
Update dataset card
128c249 verified
preview code
|
Raw History Blame Contribute Delete
7.01 kB
metadata 
license: cc-by-4.0
pretty_name: AgentLogs
configs:
  - config_name: repositories
    data_files: repositories/*.parquet
  - config_name: agent_tasks
    data_files: agent_tasks/*.parquet
  - config_name: agent_sessions
    data_files: agent_sessions/*.parquet
    default: true
  - config_name: agent_session_logs
    data_files: agent_session_logs/*.parquet
  - config_name: users
    data_files: users/*.parquet
  - config_name: pull_requests
    data_files: pull_requests/*.parquet
  - config_name: pull_request_timelines
    data_files: pull_request_timelines/*.parquet
  - config_name: workflow_runs
    data_files: workflow_runs/*.parquet

AgentLogs

GitHub risenlab/agentlogs PyPI risenlab-agentlogs arXiv 2608.29204

The latest dataset release is 1.0-20261008, which only works with the schema v1.0. It is recommended to pin the Python package with ~=1.0.0

Dataset releases | Changelog

AgentLogs is a dataset of activity related to the GitHub agents functionality: repository metadata, agent tasks, sessions, session logs (messages, tool calls, usage details, etc.), and user records.

This dataset is described in:

📄 Jonan Richards, Kosei Horikawa, Youmei Fan, Yutaro Kashiwa, and Mairieli Wessel (2026), AgentLogs: A Dataset for Opening the Black Box of GitHub's Cloud Agent. arXiv: 2608.29204 (preprint).

Dataset

# Records Size Table Content
Repositories
− 2.18% with agent tasks		 1,861,300
- 40,565		 462.7 MB repositories Public GitHub repositories with over 10 stars (metadata including name, license, language, stars, forks, timestamps, labels, topics).
Agent tasks
− 97.20% with sessions		 363,624
- 353,432		 83.7 MB agent_tasks Agent assignment on a repository (metadata including name, request, state, creator, timestamps, branch/PR identifiers).
Agent sessions
− 98.24% with session logs		 635,886
- 624,668		 243.1 MB agent_sessions Agent runs within a task (metadata including model, prompt, outcome, usage, and branch/PR identifier for that session).
Log entries
− >99.99% parsed		 75,631,136
- 75,630,896		 63.0 GB agent_session_logs Session log events (including messages, usage details, tool calls for file edits, git, and GitHub issues, PRs, comments, CI).
Users
− 4.40% with tasks or sessions		 857,428
- 37,714		 211.5 MB users Users, organizations, and bots related to the other tables (GitHub id, login slug, and creation date).
Pull requests
− 99.82% with timeline events		 317,878
- 317,297		 483.9 MB pull_requests Pull requests worked on in agent tasks and sessions (GitHub id, number, state, and branches).
Pull request timeline events 9,070,939 327.4 MB pull_request_timelines Timeline events for collected pull requests (actor, label, commit, and message).
Workflow runs 554,496 83.3 MB workflow_runs GitHub Actions runs that executed agent sessions (status, head commit, and triggering user/bot).
Total 89,292,687 64.9 GB

See the schema reference for field-level documentation of each table.

Example analysis code and a small sample: risenlab/agentlogs.

Releases

Schema versionDataset releaseRelease dateCutoff date
v1.01.0-202610088 October, 202630 September, 2026
v0.20.229 August, 202617 July, 2026

Installation 

pip install 'risenlab-agentlogs~=1.0.0' huggingface_hub

Load

Download parquet once, then query with DuckDB, Polars, or the local streaming notebook.

from pathlib import Path
from huggingface_hub import snapshot_download
from agentlogs.schema import assert_dataset_version

dataset_path = Path("data") / "dataset"
snapshot_download(
    repo_id="risenlab/agentlogs",
    repo_type="dataset",
    revision="1.0-20261008",
    local_dir=dataset_path,
)
assert_dataset_version(dataset_path)

Citation

If you use AgentLogs in an academic publication, please cite this preprint:

@misc{richards2026AgentLogsDatasetOpening,
  title = {{AgentLogs: A Dataset for Opening the Black Box of GitHub's Cloud Agent}},
  author = {Richards, Jonan and Horikawa, Kosei and Fan, Youmei and Kashiwa, Yutaro and Wessel, Mairieli},
  year = 2026,
  month = aug,
  eprint = {2608.29204},
  primaryclass = {cs.SE},
  doi = {10.48550/arXiv.2608.29204},
  archiveprefix = {arXiv},
  url = {https://arxiv.org/abs/2608.29204}
}

License

CC BY 4.0. Suggested attribution:

AgentLogs dataset by RISEN Lab, licensed under CC BY 4.0.

Repository sampling uses GitHub Search (GitHub, Zenodo) © SEART Research Group and Contributors, used under the MIT License.

The GitHub Search seed CSV is not distributed with this dataset. The repositories table is built from that CSV.