Chroneme · RWE workspace

Describe. Export. Run locally. Refine.

Build a reproducible study with your agent. Upload descriptions and reviewed examples; keep datasets and execution in your environment.

Workspace

First release: local CSV profiling and feedback. Clinical evaluation, OMOP database adapters, and final research reports are planned. A generated package is not a completed study.

1. Describe your source

Provide schema, field meanings, units, time semantics, and approved format examples. The example below is synthetic. Load a description JSON or edit it here.

{
  "title": "ICU trajectory feasibility",
  "question": "Can the recording frequency support a reviewed trajectory rule?",
  "dataset": "Synthetic example — replace with your source",
  "release": "Describe the actual release",
  "source_description": "Describe population, capture processes, known omissions, and timestamp semantics.",
  "format": "csv",
  "examples_reviewed": true,
  "tables": [
    {
      "name": "vitals",
      "entity_columns": [
        "stay_id"
      ],
      "time_column": "offset_minutes",
      "time_format": "relative_minutes",
      "time_meaning": "Elapsed minutes from ICU admission; result availability is not yet established.",
      "columns": [
        {
          "name": "stay_id",
          "kind": "text",
          "description": "ICU episode key",
          "examples": [
            "SYNTHETIC_STAY"
          ],
          "missing_values": [
            "",
            "NULL"
          ]
        },
        {
          "name": "offset_minutes",
          "kind": "number",
          "description": "Time relative to ICU admission",
          "unit": "min",
          "examples": [
            "0",
            "30"
          ],
          "missing_values": [
            "",
            "NULL"
          ]
        },
        {
          "name": "heart_rate",
          "kind": "number",
          "description": "Recorded heart rate; verify measurement channel",
          "unit": "beats/min",
          "examples": [
            "80",
            "<40"
          ],
          "missing_values": [
            "",
            "NULL",
            "NA"
          ]
        }
      ]
    }
  ]
}

I reviewed these descriptions and examples for sharing with the product and its model service.

2. Export and run locally

Generate from the saved description. Each ZIP includes readable Python, a schema, local configuration example, and run instructions. No LLM is needed locally.

3. Return local feedback

Load feedback.preview.json from a local run. The accepted format contains check states and package IDs, without raw values or logs. Review it before sending.

I reviewed this exact feedback file and approve sharing it.

4. Continue with your agent

Discuss the description or returned diagnostics. The agent can create a profiling package. Its private instructions stay on the server; this study conversation is retained.