UPLOAD_INSTRUCTIONS.md

Upload Instructions

Target dataset name:

agent-runtime-telemetry-small

Recommended repo id:

Lightcap/agent-runtime-telemetry-small

Current uploaded URL:

https://huggingface.co/datasets/Lightcap/agent-runtime-telemetry-small

Files To Upload

Upload the full contents of this folder:

data/huggingface_exports/agent-runtime-telemetry-small/

Do not upload the original runtime SQLite files. This folder already contains the viewer-friendly Parquet export and the Hugging Face dataset card.

CLI Upload

From the repository root:

python3 - <<'PY'
from pathlib import Path
import os

from dotenv import load_dotenv
from huggingface_hub import HfApi, create_repo, upload_folder

root = Path("data/huggingface_exports/agent-runtime-telemetry-small").resolve()
load_dotenv(".env")

token = (
    os.getenv("HF_TOKEN")
    or os.getenv("HUGGINGFACE_HUB_TOKEN")
    or os.getenv("HUGGING_FACE_HUB_TOKEN")
)
if not token:
    raise SystemExit("Missing HF_TOKEN or HUGGINGFACE_HUB_TOKEN.")

api = HfApi(token=token)
username = api.whoami(token=token)["name"]
repo_id = f"{username}/agent-runtime-telemetry-small"

create_repo(repo_id=repo_id, repo_type="dataset", token=token, exist_ok=True, private=False)
upload_folder(
    repo_id=repo_id,
    repo_type="dataset",
    folder_path=str(root),
    token=token,
    commit_message="Add small agent runtime telemetry dataset",
)

print(f"https://huggingface.co/datasets/{repo_id}")
PY

Manual Web Upload

1. Create a new Hugging Face Dataset named agent-runtime-telemetry-small.
2. Upload README.md, export_manifest.json, and the full data/ directory from this folder.
3. Wait for the Dataset Viewer to process the Parquet files.
4. Confirm the configs appear as operations, operation_events, artifact_records, audit_records, tool_summary, artifact_summary, daily_activity, and dataset_overview.

Validation After Upload

from datasets import load_dataset

repo_id = "Lightcap/agent-runtime-telemetry-small"
for config in ["operations", "operation_events", "artifact_records", "audit_records", "tool_summary"]:
    ds = load_dataset(repo_id, config)
    print(config, ds["train"].num_rows, ds["train"].column_names[:8])

Export Policy

This folder is a sanitized export:

• no source SQLite database files
• no raw nested payload_json bodies
• no absolute local paths
• no secret-like token strings
• Parquet tables optimized for Hugging Face Dataset Viewer

If you regenerate from newer runtime state, keep the same policy so the dataset remains useful and safe to browse.