tabellio.validation.json

{
  "$schema": "tabellio-validation/v0.2",
  "acceptance": {
    "id": "backintel-capability-platform-v1",
    "source": "Product owner correction 2026-09-26: all 16 capabilities using synthetic sources with actual Jev, actual CatBoost and actual TabICLv2; docs/roadmap/v0.1.0-development-roadmap.md is the active acceptance source.",
    "risk": "high",
    "outcomes": [
      "Validation binds to the approved exact candidate commit and writes durable evidence for passed, failed and blocked outcomes.",
      "Both unrelated synthetic scenarios preserve source revisions, immutable lineage, typed findings and correction history through shared configurable contracts.",
      "Real Jev interprets relevant synthetic text; actual outputs and request/model identities are preserved. Simulated responses only validate development controls.",
      "Real CatBoost and real TabICLv2 run comparable facts-only and facts-plus-real-Jev evaluations alongside a baseline, using time-safe features and known synthetic outcomes.",
      "Actual predictions flow through further analysis, durable attention and useful audience-scoped reports, with controlled selection, updates, fallback and rollback.",
      "Generic background API work, persistent schedules, retries, cancellation, restart and concurrent replay preserve one accepted result and local delivery.",
      "Generated artifact code runs inside an actual sandbox that rejects prohibited reads, network access and resource use.",
      "One documented local command runs the full demonstration; required workflow, rendering, safety, budget and recovery checks pass; obsolete goal-owned scaffolding has a safe disposition.",
      "Blocked real provider/model execution prevents completion. Synthetic-data results establish neither real-world accuracy, time savings nor representative business value."
    ],
    "invariants": [
      "Missing or untrustworthy required evidence is blocked, never passed.",
      "An uncommitted or dirty candidate cannot receive exact-commit validation status passed.",
      "Development validators use deterministic fixtures with zero provider calls. Final real-model evidence is separately bounded and requires applicable paid-inference/model-use approval.",
      "No paid inference, external notifications, production writes, publication or live-stack replacement without its separate authorization.",
      "Google TabFM and TabICLv2 are distinct model choices; do not claim Google TabFM execution from a TabICLv2 run."
    ],
    "forbiddenOutcomes": [
      "Do not describe structural checks as complete product validation.",
      "Do not treat a local working-tree result as evidence for a different commit.",
      "Do not report unknown validation cost as zero.",
      "Do not commit raw Kaggle source files or represent a review as attributable to every seller in a multi-seller order.",
      "Do not join order items, payments, and reviews at their raw grains to calculate totals.",
      "Do not call retrospective final-extract metrics a live or month-end snapshot, attribute multi-seller order feedback to each seller, or claim human time savings without a measured baseline.",
      "Do not substitute invented or prerecorded model responses for final Jev/predictor execution, or claim real-world quality or business value from synthetic data."
    ],
    "requiredValidatorTypes": [
      "static",
      "schema",
      "semantic",
      "workflow",
      "visual",
      "operational",
      "security"
    ]
  },
  "validators": [
    {
      "id": "manifest-schema",
      "type": "schema",
      "argv": [
        "node",
        "scripts/validation/check-contract.mjs",
        "--manifest",
        "tabellio.validation.json",
        "--check",
        "schema"
      ],
      "cwd": ".",
      "timeout": 30,
      "required": true,
      "evidencePath": "manifest-schema.json",
      "metricThresholds": {},
      "costPolicy": {
        "requiredTelemetry": true,
        "maxUsd": 0
      }
    },
    {
      "id": "validation-policy",
      "type": "static",
      "argv": [
        "node",
        "scripts/validation/check-contract.mjs",
        "--manifest",
        "tabellio.validation.json",
        "--check",
        "policy"
      ],
      "cwd": ".",
      "timeout": 30,
      "required": true,
      "evidencePath": "validation-policy.json",
      "metricThresholds": {},
      "costPolicy": {
        "requiredTelemetry": true,
        "maxUsd": 0
      }
    },
    {
      "id": "status-self-test",
      "type": "static",
      "argv": [
        "node",
        "scripts/validation/test-runner.mjs"
      ],
      "cwd": ".",
      "timeout": 30,
      "required": true,
      "evidencePath": "status-self-test.json",
      "metricThresholds": {},
      "costPolicy": {
        "requiredTelemetry": true,
        "maxUsd": 0
      }
    },
    {
      "id": "olist-profile-contract",
      "type": "semantic",
      "argv": [
        "python3",
        "scripts/data/check_olist_profile.py"
      ],
      "cwd": ".",
      "timeout": 30,
      "required": true,
      "evidencePath": "olist-profile-contract.json",
      "metricThresholds": {},
      "costPolicy": {
        "requiredTelemetry": true,
        "maxUsd": 0
      }
    },
    {
      "id": "olist-business-facts",
      "type": "semantic",
      "argv": [
        "python3",
        "-m",
        "unittest",
        "discover",
        "-s",
        "tests",
        "-p",
        "test_olist_facts.py",
        "-v"
      ],
      "cwd": ".",
      "timeout": 180,
      "required": true,
      "evidencePath": "olist-business-facts.json",
      "metricThresholds": {},
      "costPolicy": {
        "requiredTelemetry": true,
        "maxUsd": 0
      }
    },
    {
      "id": "runtime-contracts",
      "type": "semantic",
      "argv": [
        "python3",
        "-m",
        "scripts.validation.check_runtime"
      ],
      "cwd": ".",
      "timeout": 120,
      "required": true,
      "evidencePath": "runtime-contracts.json",
      "metricThresholds": {},
      "costPolicy": {
        "requiredTelemetry": true,
        "maxUsd": 0
      }
    },
    {
      "id": "capability-simulation",
      "type": "workflow",
      "argv": [
        "python3",
        "-m",
        "unittest",
        "discover",
        "-s",
        "tests",
        "-p",
        "test_simulation.py",
        "-v"
      ],
      "cwd": ".",
      "timeout": 60,
      "required": true,
      "evidencePath": "capability-simulation.json",
      "metricThresholds": {},
      "costPolicy": {
        "requiredTelemetry": true,
        "maxUsd": 0
      }
    }
  ]
}