rho-cli 0.1.25

# `rho data`

## Purpose

`rho data` handles data registration and mock-data generation.

This is the simplified successor to the old Twin idea.

## Responsibilities

- inspect source data
- infer schema and shape
- register local private data
- generate mock or synthetic shareable data
- track released outputs

## Preserve From The Old System

- explicit mock/private split
- local-only private data by default
- released outputs as a separate class from both mock and private data

## Simplify

- use a clear model:
  - private
  - mock
- use some kind of unique path like rho://datasetname/file.path or something to allow substitution in prompt templates when applying real file paths to tool calls

## Replace Old Functionality With

- simple manifests for registered datasets
- deterministic mock-generation pipeline where possible
- agent-assisted generation only as a helper, not as the source of truth

## Leave Off In V1

- full Twin session runtime
- many domain-specific format plugins unless needed immediately
- secure-compute-aware data state transitions

## Suggested Files

```text
.rho/data/registry.yaml
.rho/data/mock/<dataset-id>/
.rho/data/private/<dataset-id>/
.rho/data/released/<dataset-id>/
```

## Important Rule

Mock generation should preserve useful shape and semantics while minimizing leakage from the private source.
Human review may still be required before publishing mock derivatives.