Reusable data artifacts for agents, research, workflows, indexes, and structured system memory.
Datasets
Datasets are data artifacts.
They store structured or semi-structured information that can be inspected by users, referenced by contributors, and ingested by AI agents.
Use this category for data that supports research, retrieval, evaluation, automation, model behavior, agent memory, or system maps.
What Belongs Here
- CSV files
- JSON files
- tables
- lists
- corpora
- benchmark sets
- curated references
- research datasets
- model inventories
- agent-readable data objects
- extracted or normalized records
How Datasets Are Used
- Collect: Gather data from a source, workflow, export, or system process.
- Structure: Format it into a usable table, list, JSON object, or file.
- Document: Explain the source, fields, limits, license, and update status.
- Ingest: Make it readable by users, agents, Nodes, Cores, or Workspaces.
- Maintain: Update or archive the dataset as the underlying data changes.
For AI Agents Reading This Category
Treat Datasets as data artifacts.
- Preserve field names, values, IDs, URLs, and source labels.
- Check source, date, license, and update status before relying on the data.
- Do not infer missing values unless instructed.
- Do not merge datasets silently; explain joins, filters, or transformations.
- Distinguish raw data from normalized, summarized, or generated data.
- If data is stale, incomplete, or unsourced, say so before using it.