{
  "encoding": "leangpt-parquet-v2",
  "columns": {
    "schema_version": "int64",
    "example_id": "string",
    "source_dataset": "string",
    "source_url": "string",
    "repo_commit": "string",
    "file_path": "string",
    "lean_version": "string",
    "mathlib_version": "string",
    "extractor_version": "string",
    "pantograph_version": "string",
    "theorem_name": "string",
    "theorem_statement_raw": "string",
    "theorem_source_raw": "string",
    "namespace": "string",
    "imports": "list<item: string>",
    "proof_id": "string",
    "step_index": "int64",
    "state_before_raw": "string",
    "state_before_structured": "string",
    "state_before_internal": "string",
    "tactic_raw": "string",
    "atomic_tactics": "list<item: string>",
    "tactic_internal": "string",
    "state_after_raw": "string",
    "state_after_structured": "string",
    "state_after_internal": "string",
    "terminal": "bool",
    "terminal_marker": "string",
    "terminal_raw": "string",
    "used_premises": "list<item: string>",
    "available_context": "list<item: string>",
    "source_start": "string",
    "source_end": "string",
    "theorem_fingerprint": "string",
    "state_fingerprint": "string",
    "transition_fingerprint": "string",
    "provenance": "string",
    "replay_status": "string",
    "replay_error": "string",
    "split": "string"
  },
  "json_columns": [
    "provenance",
    "source_end",
    "source_start",
    "state_after_internal",
    "state_after_structured",
    "state_before_internal",
    "state_before_structured",
    "tactic_internal"
  ]
}