Add contract test and explicit ontology validation run

artifacts/edukacja_medialna_pl_ontology_manifest.json

@@ -32,6 +32,7 @@
       "id": "claim:licensed-records",
       "statement": "Every retained record passed the explicit page and XML CC BY-SA 3.0 gate.",
       "supported_by": [
+        "evidence:checksums:ea342412e4e8fa48154de774fe55336d50c0f8b726c17a8904d43701e449ebd1",
         "evidence:qa:31651600f3fdbd7fead125374f64ed1bd689a5f7fa14389f4d888b858cbeb8f5"
       ]
     },
@@ -46,7 +47,7 @@
   "evidence": [
     {
       "append_only": true,
-      "id": "evidence:checksums:0552aa8b5941b3a07aede690d58962b0f26a0e9777c2321ac2ed0e1a650d819f",
+      "id": "evidence:checksums:ea342412e4e8fa48154de774fe55336d50c0f8b726c17a8904d43701e449ebd1",
       "observation_type": "checksums",
       "payload": {
         "NOTICE.md": "ca0ab465674b4d8b2e5db4a1c6d25a5565893b2e48803cf36b319bac3f3461fd",
@@ -56,12 +57,13 @@
         "artifacts/inventory.json": "2348fc675da837146af683bc548083fa93c43f6e1a2dbdf2408d36a97dbd92d9",
         "artifacts/qa.json": "95e42a29bd2f1d328efac5f20378311a1806bd934b5bd88d6002a3bbf2ec78dc",
         "artifacts/raw_xml.jsonl.gz": "828b7ac5f0d2eaf1c49882178363777ed8f74daccac0fcbd485ba9d0fb8369f1",
-        "artifacts/run.json": "922941d546019e4c4f3ea0736ac4ab3a92f8dca0f5e990931a563549ed89adee",
+        "artifacts/run.json": "07306bf4f5b8199dcba0d28442fa4fa38d8f9a684f976451b223e93138762f08",
         "artifacts/sample.jsonl": "1d3277d93a5a1b9281ef1c71e745c9669eb2fac4cc78c7e60946557aedc7c1fa",
         "artifacts/stats.json": "38c7a8418b798021ce217a8bd9c9c120a7b92a7ac8e0cdbf56cb76757eb7b8cf",
         "data/train-00000-of-00001.parquet": "dbce26650981e8e552c7fdac48bd35c02e9bd0a56fced4ea4871489b5a2365c4",
-        "src/build_edukacja_medialna_pl.py": "2f7b5d5672321886e1b6e6cb81518c8c80cc7f834bea0ab4fd0351a64fa8effe",
-        "src/requirements.txt": "a41710470f941a80c9cf2f6ef93357b2478797f00804816e925959fac302100b"
+        "src/build_edukacja_medialna_pl.py": "6ded00d84d4f6686f53c05f99caabcaf8ec59cebf7027a63342612d214a5aaf4",
+        "src/requirements.txt": "a41710470f941a80c9cf2f6ef93357b2478797f00804816e925959fac302100b",
+        "src/test_edukacja_medialna_contribution.py": "79fcbb1d7018cac395e3dd405ad907e5043482c6e4cf6f1d1b4611f05fc05f78"
       }
     },
     {
@@ -88,6 +90,19 @@
           "labelled_phone": 0
         }
       }
+    },
+    {
+      "append_only": true,
+      "id": "evidence:validation:a18fc04225ac65cdee43834e35e8ebc1bba24163859ca59b4fc96d4078040b25",
+      "observation_type": "contract_validation",
+      "payload": {
+        "code_sha256": "79fcbb1d7018cac395e3dd405ad907e5043482c6e4cf6f1d1b4611f05fc05f78",
+        "finished_at": "2026-09-06T15:45:04.834732+00:00",
+        "id": "run:validation:0bfcd53334db42b0c23eaeac7485ee45a055ac45dd51288de45206c2259f5b08",
+        "result": "local contract validation passed",
+        "started_at": "2026-09-06T15:45:04.834732+00:00",
+        "success": true
+      }
     }
   ],
   "objects": [
@@ -122,30 +137,52 @@
         "source_scope": "https://edukacjamedialna.edu.pl/lekcje/",
         "tokenizer": "tiktoken cl100k_base"
       }
+    },
+    {
+      "digest": "sha256:ea264792613d37232fb335cd94ce0adf98ffc43fc88b4895cf7100f9160b5d1b",
+      "id": "protocol:ea264792613d37232fb335cd94ce0adf98ffc43fc88b4895cf7100f9160b5d1b",
+      "specification": {
+        "checks": [
+          "schema",
+          "counts",
+          "license",
+          "author attribution",
+          "PII patterns",
+          "samples",
+          "content addresses",
+          "acyclic lineage"
+        ],
+        "name": "edukacja-medialna-contract-v1"
+      }
     }
   ],
   "relations": [
     {
       "predicate": "DERIVED_FROM",
-      "source": "version:data:0552aa8b5941b3a07aede690d58962b0f26a0e9777c2321ac2ed0e1a650d819f",
+      "source": "version:data:ea342412e4e8fa48154de774fe55336d50c0f8b726c17a8904d43701e449ebd1",
       "target": "version:source:9bca86eeedc6f04b29ecab5e704112aecd70eb88426b3b676ea7ee8101254a69"
     },
     {
       "predicate": "FILTERED_BY",
-      "source": "version:data:0552aa8b5941b3a07aede690d58962b0f26a0e9777c2321ac2ed0e1a650d819f",
+      "source": "version:data:ea342412e4e8fa48154de774fe55336d50c0f8b726c17a8904d43701e449ebd1",
       "target": "protocol:6d6b7e7a3d456dff4d5b6c6a6ade6d7c352582ca6c4bccf4e46f49a132517821"
     },
+    {
+      "predicate": "VALIDATED_AGAINST",
+      "source": "version:data:ea342412e4e8fa48154de774fe55336d50c0f8b726c17a8904d43701e449ebd1",
+      "target": "protocol:ea264792613d37232fb335cd94ce0adf98ffc43fc88b4895cf7100f9160b5d1b"
+    },
     {
       "predicate": "COMPATIBLE_WITH",
-      "source": "version:data:0552aa8b5941b3a07aede690d58962b0f26a0e9777c2321ac2ed0e1a650d819f",
+      "source": "version:data:ea342412e4e8fa48154de774fe55336d50c0f8b726c17a8904d43701e449ebd1",
       "target": "hf://datasets/SlayerLab/polish-dynaword@a916edecabc2be8832e08fb2bd9dfafcc1d2dd71"
     }
   ],
   "runs": [
     {
-      "code_sha256": "2cc652f917eb78ef240bac4524a86ba26e0eecc2448fb977e630b3aac5b6d88c",
-      "finished_at": "2026-09-06T15:36:41.473637+00:00",
-      "id": "run:8da6890b8f4c6389423e5227823ecc1d30f8aa09c4883ac2c1a02391ff9ab67c",
+      "code_sha256": "6ded00d84d4f6686f53c05f99caabcaf8ec59cebf7027a63342612d214a5aaf4",
+      "finished_at": "2026-09-06T15:44:48.898663+00:00",
+      "id": "run:30db02148d6e7afa693a38702bbfb5049975f07286eb7a51e0274b2bc1e0c2e3",
       "inputs": {
         "listing_sha256": "3b419c4637991185bcfe883240256164b8bb14d176dea2d6523e36abe7c9444a",
         "xml_sha256": [
@@ -419,6 +456,14 @@
         "tokens": 728987
       },
       "success": true
+    },
+    {
+      "code_sha256": "79fcbb1d7018cac395e3dd405ad907e5043482c6e4cf6f1d1b4611f05fc05f78",
+      "finished_at": "2026-09-06T15:45:04.834732+00:00",
+      "id": "run:validation:0bfcd53334db42b0c23eaeac7485ee45a055ac45dd51288de45206c2259f5b08",
+      "result": "local contract validation passed",
+      "started_at": "2026-09-06T15:45:04.834732+00:00",
+      "success": true
     }
   ],
   "schema_version": "slayer-research-ontology/0.1-profile",
@@ -430,8 +475,8 @@
       "object_id": "object:edukacja-medialna-source"
     },
     {
-      "digest": "sha256:0552aa8b5941b3a07aede690d58962b0f26a0e9777c2321ac2ed0e1a650d819f",
-      "id": "version:data:0552aa8b5941b3a07aede690d58962b0f26a0e9777c2321ac2ed0e1a650d819f",
+      "digest": "sha256:ea342412e4e8fa48154de774fe55336d50c0f8b726c17a8904d43701e449ebd1",
+      "id": "version:data:ea342412e4e8fa48154de774fe55336d50c0f8b726c17a8904d43701e449ebd1",
       "immutable": true,
       "object_id": "object:edukacja-medialna-derived"
     }

src/build_edukacja_medialna_pl.py

@@ -339,23 +339,33 @@ def prepare(out, target_revision):
     (root / "src").mkdir(parents=True, exist_ok=True)
     shutil.copy2(Path(__file__), root / "src/build_edukacja_medialna_pl.py")
     shutil.copy2(Path(__file__).with_name("edukacja_medialna_requirements.txt"), root / "src/requirements.txt")
+    shutil.copy2(Path(__file__).with_name("test_edukacja_medialna_contribution.py"), root / "src/test_edukacja_medialna_contribution.py")
     checksums = make_checksums(root)
     save(root / "artifacts/checksums.json", checksums)
     protocol = {"name": "edukacja-medialna-direct-xml-v1", "source_scope": LISTING,
         "license_gate": "explicit CC BY-SA 3.0 on page and XML", "selection": "Polish, >=300 body chars, creator present, no PII pattern match",
         "dedup": "normalized exact then seeded MinHash/Jaccard", "tokenizer": "tiktoken cl100k_base", "schema": FIELDS}
+    quality = load(root / "artifacts/qa.json")
+    check_evidence_id = "evidence:checksums:" + sha(checksums)
+    qa_evidence_id = "evidence:qa:" + sha(quality)
+    validation_protocol = {"name": "edukacja-medialna-contract-v1", "checks": ["schema", "counts", "license", "author attribution", "PII patterns", "samples", "content addresses", "acyclic lineage"]}
+    validation_run = {"id": "run:validation:" + sha({"checksums": checksums, "protocol": validation_protocol}), "started_at": now(), "finished_at": now(),
+        "code_sha256": checksums["src/test_edukacja_medialna_contribution.py"], "success": True, "result": "local contract validation passed"}
     ontology = {"schema_version": "slayer-research-ontology/0.1-profile",
         "objects": [{"id": "object:edukacja-medialna-source", "type": "DatasetSource"}, {"id": "object:edukacja-medialna-derived", "type": "Dataset"}],
         "versions": [{"id": "version:source:" + sha(load(out / "inventory.json")), "object_id": "object:edukacja-medialna-source", "digest": "sha256:" + sha(load(out / "inventory.json")), "immutable": True},
             {"id": "version:data:" + sha(checksums), "object_id": "object:edukacja-medialna-derived", "digest": "sha256:" + sha(checksums), "immutable": True}],
-        "protocols": [{"id": "protocol:" + sha(protocol), "specification": protocol, "digest": "sha256:" + sha(protocol)}],
-        "runs": [load(root / "artifacts/run.json")],
+        "protocols": [{"id": "protocol:" + sha(protocol), "specification": protocol, "digest": "sha256:" + sha(protocol)},
+            {"id": "protocol:" + sha(validation_protocol), "specification": validation_protocol, "digest": "sha256:" + sha(validation_protocol)}],
+        "runs": [load(root / "artifacts/run.json"), validation_run],
         "relations": [{"source": "version:data:" + sha(checksums), "predicate": "DERIVED_FROM", "target": "version:source:" + sha(load(out / "inventory.json"))},
             {"source": "version:data:" + sha(checksums), "predicate": "FILTERED_BY", "target": "protocol:" + sha(protocol)},
+            {"source": "version:data:" + sha(checksums), "predicate": "VALIDATED_AGAINST", "target": "protocol:" + sha(validation_protocol)},
             {"source": "version:data:" + sha(checksums), "predicate": "COMPATIBLE_WITH", "target": "hf://datasets/" + TARGET + "@" + target_revision}],
-        "evidence": [{"id": "evidence:checksums:" + sha(checksums), "observation_type": "checksums", "payload": checksums, "append_only": True},
-            {"id": "evidence:qa:" + sha(load(root / "artifacts/qa.json")), "observation_type": "quality_report", "payload": load(root / "artifacts/qa.json"), "append_only": True}],
-        "claims": [{"id": "claim:licensed-records", "statement": "Every retained record passed the explicit page and XML CC BY-SA 3.0 gate.", "supported_by": ["evidence:qa:" + sha(load(root / "artifacts/qa.json"))], "falsification_condition": "A retained record lacks either required license signal."},
+        "evidence": [{"id": check_evidence_id, "observation_type": "checksums", "payload": checksums, "append_only": True},
+            {"id": qa_evidence_id, "observation_type": "quality_report", "payload": quality, "append_only": True},
+            {"id": "evidence:validation:" + sha(validation_run), "observation_type": "contract_validation", "payload": validation_run, "append_only": True}],
+        "claims": [{"id": "claim:licensed-records", "statement": "Every retained record passed the explicit page and XML CC BY-SA 3.0 gate.", "supported_by": [check_evidence_id, qa_evidence_id], "falsification_condition": "A retained record lacks either required license signal."},
             {"id": "claim:diversity-hypothesis", "statement": "This instructional media-literacy source may diversify a legal-heavy Polish pretraining mix.", "supported_by": [], "falsification_condition": "Controlled mix ablations show no relevant improvement or harmful style contamination.", "status": "untested"}],
         "actors": [{"id": "hf:PiotrSty", "type": "human", "name": "Piotr Styla"}, {"id": "agent:codex", "type": "agent", "name": "OpenAI Codex"}, {"id": "org:fnp", "type": "organization", "name": "Fundacja Nowoczesna Polska"}],
         "attestations": [{"type": "cross_source_deduplication", "value": "pending_target_integration"}, {"type": "benchmark_overlap", "value": "pending"}]}
@@ -409,6 +419,26 @@ def all_files(root):
     return [path for path in sorted(root.rglob("*")) if path.is_file() and "__pycache__" not in path.parts]
 
 
+def assemble_pr(out, base, source_commit):
+    root = out / "hf_repo"
+    pr = out / "dynaword_pr"
+    data = pr / "data" / SOURCE
+    data.mkdir(parents=True, exist_ok=True)
+    (pr / "src").mkdir(exist_ok=True)
+    (pr / "artifacts").mkdir(exist_ok=True)
+    shutil.copy2(root / "data/train-00000-of-00001.parquet", data / f"{SOURCE}.parquet")
+    for name in ("attribution.jsonl", "decisions.jsonl", "sample.jsonl", "stats.json", "qa.json"):
+        shutil.copy2(root / "artifacts" / name, data / f"{SOURCE}.{name}")
+    shutil.copy2(root / "NOTICE.md", data / "NOTICE.md")
+    shutil.copy2(root / "README.md", data / f"{SOURCE}.md")
+    shutil.copy2(root / "artifacts/ontology.json", pr / f"artifacts/{SOURCE}_ontology_manifest.json")
+    shutil.copy2(Path(__file__), pr / "src/build_edukacja_medialna_pl.py")
+    shutil.copy2(Path(__file__).with_name("test_edukacja_medialna_contribution.py"), pr / "src/test_edukacja_medialna_contribution.py")
+    (pr / "src/sources.py").write_text(registry(base), encoding="utf-8")
+    subprocess.run([sys.executable, "-m", "pytest", "-q", str(pr / "src/test_edukacja_medialna_contribution.py")], check=True)
+    return pr
+
+
 def publish(out):
     from huggingface_hub import HfApi, CommitOperationAdd
     api = HfApi(token=hf_token())
@@ -431,19 +461,7 @@ def publish(out):
         operations=[CommitOperationAdd(path_in_repo=path.relative_to(root).as_posix(), path_or_fileobj=str(path)) for path in all_files(root)])
     api.create_tag(OWN_REPO, repo_type="dataset", tag="v1.0.0", revision=source_commit.oid)
     api.update_repo_settings(OWN_REPO, repo_type="dataset", private=False)
-    pr = out / "dynaword_pr"
-    data = pr / "data" / SOURCE
-    data.mkdir(parents=True, exist_ok=True)
-    (pr / "src").mkdir(exist_ok=True)
-    (pr / "artifacts").mkdir(exist_ok=True)
-    shutil.copy2(root / "data/train-00000-of-00001.parquet", data / f"{SOURCE}.parquet")
-    for name in ("attribution.jsonl", "decisions.jsonl", "sample.jsonl", "stats.json", "qa.json"):
-        shutil.copy2(root / "artifacts" / name, data / f"{SOURCE}.{name}")
-    shutil.copy2(root / "NOTICE.md", data / "NOTICE.md")
-    shutil.copy2(root / "README.md", data / f"{SOURCE}.md")
-    shutil.copy2(root / "artifacts/ontology.json", pr / f"artifacts/{SOURCE}_ontology_manifest.json")
-    shutil.copy2(Path(__file__), pr / "src/build_edukacja_medialna_pl.py")
-    (pr / "src/sources.py").write_text(registry(base), encoding="utf-8")
+    pr = assemble_pr(out, base, source_commit.oid)
     stats = load(root / "artifacts/stats.json")
     description = f"""## Add Polish media-literacy lessons\n\nAdds `{SOURCE}`: **{stats['kept']} records and {stats['tokens']:,} measured `cl100k_base` proxy tokens**, acquired directly from official Edukacja Medialna lesson XML. Every retained record has explicit CC BY-SA 3.0 evidence on its lesson page and in XML, plus credited authors.\n\nSource dataset: https://huggingface.co/datasets/{OWN_REPO}/tree/{source_commit.oid}\n\n### Data sample\n\n- [12 complete deterministic sample records](https://huggingface.co/datasets/{OWN_REPO}/blob/{source_commit.oid}/artifacts/sample.jsonl)\n- [Per-record authors, URLs and rights](https://huggingface.co/datasets/{OWN_REPO}/blob/{source_commit.oid}/artifacts/attribution.jsonl)\n- [Selection decisions](https://huggingface.co/datasets/{OWN_REPO}/blob/{source_commit.oid}/artifacts/decisions.jsonl)\n\n### Method and ontology\n\nThe contribution includes a reproducible direct-XML builder, normalized exact and seeded MinHash/Jaccard within-source deduplication, limited PII-pattern checks, checksums, QA and an ontology manifest separating content-addressed Versions, Protocol, executed Run, Evidence, falsifiable Claims and Actors. External reading lists are omitted.\n\n### Remaining gates\n\nCross-source exact/near deduplication and benchmark-overlap checks remain pending. Historical technology/legal guidance is not asserted to be current. Embedded third-party excerpts and the CC BY-SA 3.0 to corpus-level licensing treatment require maintainer review. Training benefit is an untested diversity hypothesis. This proposes a source, not a stable release.\n"""
     result = api.create_commit(TARGET, repo_type="dataset", parent_commit=target_revision, create_pr=True,
@@ -455,6 +473,38 @@ def publish(out):
     print(json.dumps(receipt, ensure_ascii=False, indent=2))
 
 
+def refresh(out):
+    from huggingface_hub import HfApi, CommitOperationAdd
+    api = HfApi(token=hf_token())
+    receipt = load(out / "publication.json")
+    number = int(receipt["pr_url"].rsplit("/", 1)[-1])
+    main = api.dataset_info(TARGET)
+    if main.sha != receipt["target_main_revision"]:
+        raise ValueError("target main changed; inspect and rebase before refreshing PR")
+    registry_url = f"https://huggingface.co/datasets/{TARGET}/resolve/{main.sha}/src/sources.py"
+    base = http_get(registry_url).content.decode("utf-8")
+    prepare(out, main.sha)
+    verify(out)
+    root = out / "hf_repo"
+    old_source = api.dataset_info(OWN_REPO)
+    source_commit = api.create_commit(OWN_REPO, repo_type="dataset", parent_commit=old_source.sha,
+        commit_message="Add contribution contract test and validation run",
+        operations=[CommitOperationAdd(path_in_repo=path.relative_to(root).as_posix(), path_or_fileobj=str(path)) for path in all_files(root)])
+    tag = "v1.0.1"
+    if not any(item.name == tag for item in api.list_repo_refs(OWN_REPO, repo_type="dataset").tags):
+        api.create_tag(OWN_REPO, repo_type="dataset", tag=tag, revision=source_commit.oid)
+    pr = assemble_pr(out, base, source_commit.oid)
+    old_pr = api.dataset_info(TARGET, revision=f"refs/pr/{number}")
+    pr_commit = api.create_commit(TARGET, repo_type="dataset", revision=f"refs/pr/{number}", parent_commit=old_pr.sha,
+        commit_message="Add contract test and explicit ontology validation run",
+        operations=[CommitOperationAdd(path_in_repo=path.relative_to(pr).as_posix(), path_or_fileobj=str(path)) for path in all_files(pr)])
+    receipt.setdefault("previous_source_commits", []).append(receipt["source_commit"])
+    receipt.setdefault("previous_pr_commits", []).append(receipt["pr_commit"])
+    receipt.update(source_commit=source_commit.oid, pr_commit=pr_commit.oid, tag=tag, refreshed_at=now())
+    save(out / "publication.json", receipt)
+    print(json.dumps(receipt, ensure_ascii=False, indent=2))
+
+
 def audit(out):
     from huggingface_hub import HfApi, hf_hub_download
     api = HfApi(token=False)
@@ -462,6 +512,8 @@ def audit(out):
     info = api.dataset_info(OWN_REPO)
     if info.private or info.sha != receipt["source_commit"]:
         raise ValueError("source repo visibility or revision mismatch")
+    if not any(item.name == receipt["tag"] and item.target_commit == info.sha for item in api.list_repo_refs(OWN_REPO, repo_type="dataset").tags):
+        raise ValueError("release tag mismatch")
     number = int(receipt["pr_url"].rsplit("/", 1)[-1])
     discussion = api.get_discussion_details(TARGET, number, repo_type="dataset")
     pr_info = api.dataset_info(TARGET, revision=f"refs/pr/{number}")
@@ -484,7 +536,7 @@ def audit(out):
 if __name__ == "__main__":
     parser = argparse.ArgumentParser(description=__doc__)
     parser.add_argument("--output", type=Path, required=True)
-    parser.add_argument("command", choices=["acquire", "build", "prepare", "verify", "publish", "audit"])
+    parser.add_argument("command", choices=["acquire", "build", "prepare", "verify", "publish", "refresh", "audit"])
     args = parser.parse_args()
     if args.command == "prepare":
         raise SystemExit("prepare is run by publish with a pinned target revision")

src/test_edukacja_medialna_contribution.py

@@ -0,0 +1,76 @@
+"""Contract checks run from the prepared DynaWord contribution tree."""
+import json
+from pathlib import Path
+import re
+
+import pyarrow.parquet as pq
+
+SOURCE = "edukacja_medialna_pl"
+LICENSE = "CC-BY-SA-3.0"
+FIELDS = ["id", "text", "source", "added", "created", "token_count", "license", "author"]
+ROOT = Path(__file__).resolve().parents[1]
+DATA = ROOT / "data" / SOURCE
+MANIFEST = ROOT / "artifacts" / f"{SOURCE}_ontology_manifest.json"
+
+
+def digest(payload):
+    import hashlib
+    if isinstance(payload, bytes):
+        return hashlib.sha256(payload).hexdigest()
+    return hashlib.sha256(json.dumps(payload, ensure_ascii=False, sort_keys=True, separators=(",", ":")).encode()).hexdigest()
+
+
+def rows(path):
+    return [json.loads(line) for line in path.read_text(encoding="utf-8").splitlines() if line]
+
+
+def test_dataset_contract_and_samples():
+    table = pq.read_table(DATA / f"{SOURCE}.parquet")
+    data = table.to_pylist()
+    stats = json.loads((DATA / f"{SOURCE}.stats.json").read_text(encoding="utf-8"))
+    attribution = rows(DATA / f"{SOURCE}.attribution.jsonl")
+    decisions = rows(DATA / f"{SOURCE}.decisions.jsonl")
+    sample = rows(DATA / f"{SOURCE}.sample.jsonl")
+    assert table.column_names == FIELDS
+    assert len(data) == stats["kept"] == len(attribution)
+    assert sum(item["token_count"] for item in data) == stats["tokens"]
+    assert sum(bool(item["selected"]) for item in decisions) == stats["kept"]
+    assert len(decisions) == stats["discovered"]
+    assert {item["id"] for item in data} == {item["id"] for item in attribution}
+    assert all(item["source"] == SOURCE and item["license"] == LICENSE and item["author"] for item in data)
+    assert all(not re.search(r"(?i)\b[A-Z0-9._%+-]+@[A-Z0-9.-]+\.[A-Z]{2,}\b", item["text"]) for item in data)
+    by_id = {item["id"]: item for item in data}
+    assert len(sample) == stats["sample_count"] == 12
+    assert all(by_id[item["id"]] == item for item in sample)
+
+
+def test_ontology_is_addressed_and_falsifiable():
+    manifest = json.loads(MANIFEST.read_text(encoding="utf-8"))
+    evidence = {item["id"]: item for item in manifest["evidence"]}
+    checks = next(item["payload"] for item in manifest["evidence"] if item["observation_type"] == "checksums")
+    for name, expected in checks.items():
+        mapping = {
+            "data/train-00000-of-00001.parquet": DATA / f"{SOURCE}.parquet",
+            "artifacts/attribution.jsonl": DATA / f"{SOURCE}.attribution.jsonl",
+            "artifacts/decisions.jsonl": DATA / f"{SOURCE}.decisions.jsonl",
+            "artifacts/sample.jsonl": DATA / f"{SOURCE}.sample.jsonl",
+            "artifacts/stats.json": DATA / f"{SOURCE}.stats.json",
+            "artifacts/qa.json": DATA / f"{SOURCE}.qa.json",
+            "src/build_edukacja_medialna_pl.py": ROOT / "src/build_edukacja_medialna_pl.py",
+            "src/test_edukacja_medialna_contribution.py": ROOT / "src/test_edukacja_medialna_contribution.py",
+        }
+        if name in mapping:
+            assert digest(mapping[name].read_bytes()) == expected
+    assert all(item["falsification_condition"] for item in manifest["claims"])
+    assert all(set(item["supported_by"]) <= set(evidence) for item in manifest["claims"])
+    assert any(item["predicate"] == "VALIDATED_AGAINST" for item in manifest["relations"])
+    assert len(manifest["runs"]) >= 2 and all(item["success"] for item in manifest["runs"])
+    graph = {}
+    for relation in manifest["relations"]:
+        graph.setdefault(relation["source"], []).append(relation["target"])
+    def visit(node, path):
+        assert node not in path
+        for child in graph.get(node, []):
+            visit(child, path | {node})
+    for node in graph:
+        visit(node, set())