Version Snapshot at Export

Reproducibility is not only about methods. It is also about execution context: app version, ruleset version, stats backend, model version, and when the figure was produced. Licklider carries this provenance metadata alongside export workflows so future review starts from the right environment assumptions.

STEP 1 - The Pitfall: Environment-Blind Results

The same analysis request can behave differently across software versions, ruleset updates, or model updates. If that context is missing, it becomes harder to explain why a figure passed one review path, triggered a later guard, or generated different draft text at a later date.

• Ruleset drift affects guard outcomes. A figure that once passed export review may surface different disclosure requirements after a ruleset update.
• Stats backend updates affect numerical context. Even small numerical differences matter when results are close to a decision threshold.
• Model updates affect generated drafts. Methods and legend text can change when the underlying model changes, even if the figure inputs do not.

STEP 2 - Journal Requirement

Journal and reproducibility standards commonly expect software versions to be disclosed when they are material to the analysis or the generated output. At minimum, reviewers and future readers need enough provenance to understand which system state produced the exported figure.

STEP 3 - Licklider's Solution

Input

• No extra user input is required. Provenance metadata is assembled automatically from the export-time generation state.
• When a reproducibility bundle is created, companion artifacts such as preprocess_report.json can also be referenced through bundle metadata.

Output

• Generation provenance bundle. Export-facing metadata includes app version, ruleset version, stats backend version, model version, and execution timestamp.
• Companion artifact linkage. Repro bundle metadata can point to reproducibility artifacts such as preprocess_report.json and figure patch references when they exist.
• Researcher-visible provenance note. The same provenance is surfaced in export drafts so the researcher can review the environment context alongside Methods and legend text.

Guard

STEP 4 - Export Example

Example: provenance bundle

{
  "app_version": "2.4.1",
  "ruleset_version": "figsafe-44",
  "stats_backend_version": "python-stats@2.5.0",
  "model_version": "openai:gpt-5.4",
  "executed_at": "2026-04-05T10:22:31.000Z"
}

Example: repro manifest excerpt

{
  "meta": {
    "preprocess": {
      "preprocess_version": "2",
      "status": "ok",
      "preprocess_report_path": "preprocess_report.json",
      "preprocess_report_hash": "4f65b1b9c7f4f9a0b73d2d2a9f0f0ce1f65a7c9b4a8d2e6f3c0b1f1a2d3e4f5a"
    }
  }
}

Related Features