README.md

---
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"
---

# AgentLogs

<a href="https://huggingface.co/datasets/risenlab/agentlogs" target="_blank" rel="noopener noreferrer"><img src="https://img.shields.io/badge/Version-v0.2-blue?logo=huggingface&amp;logoColor=whitesmoke" alt="Version"></a>
<a href="https://pypi.org/project/risenlab-agentlogs/" target="_blank" rel="noopener noreferrer"><img src="https://img.shields.io/pypi/v/risenlab-agentlogs?label=Package&amp;logo=pypi&amp;logoColor=whitesmoke" alt="Package"></a>
<a href="https://github.com/risenlab/agentlogs/tree/main" target="_blank" rel="noopener noreferrer"><img src="https://img.shields.io/badge/Code-v0.2-181717?logo=github&amp;logoColor=whitesmoke" alt="Code"></a>
<a href="https://arxiv.org/abs/2608.29204" target="_blank" rel="noopener noreferrer"><img src="https://img.shields.io/badge/Preprint-2608.29204-b31b1b.svg?logo=arxiv&amp;logoColor=whitesmoke" alt="Preprint"></a>

AgentLogs is a dataset of activity related to the [GitHub agents functionality](https://github.com/features/copilot/agents): 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](https://arxiv.org/abs/2608.29204) (preprint).

| | # Records | Size | Table | Content |
| --- | ---: | ---: | --- | --- |
| **Repositories**<br><small>−&nbsp;1.98% with agent tasks</small> | 1,812,362<br><small>-&nbsp;35,810</small> | 395.6&nbsp;MB | [`repositories`](docs/schema/repository.md) | Public GitHub repositories with over 10 stars (metadata including name, license, language, stars, forks, timestamps, labels, topics). |
| **Agent tasks**<br><small>−&nbsp;99.90% found</small><br><small>−&nbsp;97.00% with sessions</small> | 307,416<br><small>-&nbsp;307,108</small><br><small>-&nbsp;298,188</small> | 77.1&nbsp;MB | [`agent_tasks`](docs/schema/agent_task.md) | Agent assignment on a repository (metadata including name, request, state, creator, timestamps, branch/PR identifiers). |
| **Agent sessions** | 549,239 | 225.1&nbsp;MB | [`agent_sessions`](docs/schema/agent_session.md) | Agent runs within a task (metadata including model, prompt, outcome, usage, and branch/PR identifier for that session). |
| **Log entries**<br><small>−&nbsp;>99.99% parsed</small> | 64,255,174<br><small>-&nbsp;64,254,936</small> | 56.0&nbsp;GB | [`agent_session_logs`](docs/schema/agent_session_log_entry.md) | Session log events (including messages, usage details, tool calls for file edits, git, and GitHub issues, PRs, comments, CI). |
| **Users**<br><small>−&nbsp;99.96% found</small> | 33,573<br><small>-&nbsp;33,561</small> | 39.7&nbsp;MB | [`users`](docs/schema/user.md) | Users related to the agent tasks and sessions (only GitHub id and username). |
| **Total** | **66,957,764** | **56.7&nbsp;GB** | | |

See the [schema reference](docs/schema/README.md) for field-level documentation of each table.

<p align="center">
  <img src="docs/figures/relationships.png" alt="Entity relationship diagram" width="300">
  <br>
  <em>Entity relationship diagram for the AgentLogs dataset. Non-foreign key fields are omitted.</em>
</p>

Example analysis code and a small sample: [risenlab/agentlogs](https://github.com/risenlab/agentlogs).

## Installation

```bash
pip install risenlab-agentlogs huggingface_hub datasets
```

## Load

### Snapshot (recommended)

Download parquet once, then query with DuckDB, Polars, or the local streaming notebooks. Prefer this for `agent_session_logs`.

```python
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="v0.2",
    local_dir=dataset_path,
)
assert_dataset_version(dataset_path)
```

### `load_dataset`

Convenient for the smaller tables.

```python
from datasets import load_dataset
from agentlogs.schema import assert_version

assert_version("v0.2")

repositories = load_dataset("risenlab/agentlogs", "repositories", split="train", revision="v0.2")
tasks = load_dataset("risenlab/agentlogs", "agent_tasks", split="train", revision="v0.2")
sessions = load_dataset("risenlab/agentlogs", "agent_sessions", split="train", revision="v0.2")
users = load_dataset("risenlab/agentlogs", "users", split="train", revision="v0.2")
```

`agent_session_logs` is large. Prefer a snapshot. Do not stream that config with `load_dataset` (full rows, including payloads, are downloaded to your machine).

## Citation

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

```bibtex
@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](DATA_LICENSE). Suggested attribution:

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

Repository sampling uses [GitHub Search](https://seart-ghs.si.usi.ch/)
([GitHub](https://github.com/seart-group/ghs), [Zenodo](https://doi.org/10.5281/zenodo.4588464))
&copy; SEART Research Group and Contributors, used under the
[MIT License](https://github.com/seart-group/ghs/blob/master/LICENSE).

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