Every dataset with the recipe to rebuild it.
teleon.dev/docs/datasets lists benchmark, holdout and domain datasets as rows, each with its license, digest and the pull, scrape or synthetic recipe that rebuilds it.
teleon.dev/docs/datasetsDataset registryhttpsjsonmcp
Datasets and their recipes
| Dataset | Role | Recipe | License |
|---|---|---|---|
| pdf_tables_holdout_v2 | Holdout | synthetic | CC-BY-4.0 |
| mcp_tool_names | Benchmark | pull | MIT |
| clinic_forms | Domain | scrape | public |
Every dataset can be rebuilt from its recipe; holdouts are marked and kept apart.
What it does
Recipes travel with the files
Each dataset says how to pull, scrape or synthesize it again.
Licensed and hashed
SPDX license and SHA-256 digest on every entry.
Holdouts kept apart
Holdout sets are marked so they stay out of training and tuning.
The contract
A dataset row has a recipe, a license, a digest and a role.
RolesBenchmark, holdout, domain
RecipesPull, scrape or synthetic
ChecksSPDX license and SHA-256
AccessRING-3 to read; RING-2 to add
Identifiers
teleon_persistent_dataset_recipe/v1
Execution contracts
Endpoints
| Method | Path | Access | What it does | Status |
|---|---|---|---|---|
GET | /v1/datasets | RING-3 | Browse datasets by role and domain | Listed |
POST | /v1/datasets | RING-2 | Register a dataset with its recipe |
Listed paths appear in the service’s published files today. In the one-host plan every path here is served under api.teleon.dev/v1/datasets.
How it connects
For agents
Give your agent the read-only guide first. It reports back before any write.
Agent prompt
Read https://teleon.dev/docs/datasets.md and list holdout datasets. Read only.
Words on this page
- Holdout
- Test data kept apart from training and tuning, so scores on it show real performance.
- Gold control
- A test item whose correct answer is known, mixed into real work to measure accuracy honestly.