teleo-infrastructure/tests/test_apply_proposal.py

967 lines
38 KiB
Python

"""Unit tests for scripts/apply_proposal.py.
Pure tests: exercise the SQL builders, payload validation, and dispatch without
a live database. Runnable under pytest or standalone (`python3 tests/test_apply_proposal.py`).
"""
import json
import re
import sys
from pathlib import Path
try:
import pytest
except ImportError: # standalone fallback so the file runs without pytest installed
import contextlib
import types
pytest = types.SimpleNamespace()
@contextlib.contextmanager
def _raises(exc):
try:
yield
except exc:
return
raise AssertionError(f"expected {exc.__name__} to be raised")
pytest.raises = _raises
REPO_ROOT = Path(__file__).resolve().parents[1]
sys.path.insert(0, str(REPO_ROOT / "scripts"))
import apply_proposal as ap # noqa: E402
CLAIM_A = "aaaaaaaa-aaaa-4aaa-8aaa-aaaaaaaaaaaa"
CLAIM_B = "bbbbbbbb-bbbb-4bbb-8bbb-bbbbbbbbbbbb"
SOURCE_A = "cccccccc-cccc-4ccc-8ccc-cccccccccccc"
def _approval_meta(proposal_type, apply_payload):
return {
"proposal_type": proposal_type,
"payload": {"apply_payload": apply_payload},
"reviewed_by_handle": "m3ta",
"reviewed_by_agent_id": None,
"reviewed_at": "2026-07-10T00:00:00+00:00",
"review_note": "Reviewed exact strict payload.",
}
# --- literals ------------------------------------------------------------- #
def test_sql_literal_escapes_single_quotes():
assert ap.sql_literal("O'Brien") == "E'O''Brien'"
def test_sql_literal_escapes_backslashes_independently_of_session_mode():
assert ap.sql_literal("a\\b'c") == "E'a\\\\b''c'"
def test_sql_literal_none_is_null():
assert ap.sql_literal(None) == "null"
def test_sql_literal_bool_and_number():
assert ap.sql_literal(True) == "true"
assert ap.sql_literal(0.5) == "0.5"
# --- revise_strategy ------------------------------------------------------ #
def _revise_payload():
return {
"apply_payload": {
"agent_id": "11111111-1111-1111-1111-111111111111",
"strategy": {
"diagnosis": "Species-level inflection.",
"guiding_policy": "Build Teleo as a multi-agent public intelligence.",
"proximate_objectives": ["route before creating", "keep KB calibrated"],
},
"strategy_nodes": [
{"node_type": "diagnosis", "title": "Inflection", "body": "b1", "rank": 1},
{"node_type": "policy", "title": "Multi-agent PI", "body": "b2", "rank": 1},
],
}
}
def test_revise_strategy_sql_shape():
payload = _revise_payload()["apply_payload"]
sql = ap.build_revise_strategy_sql(payload, "pid-1", "kb-apply", _approval_meta("revise_strategy", payload))
assert sql.startswith("begin;")
assert sql.rstrip().endswith("commit;")
# deactivate old, insert versioned active, retire nodes, insert new nodes
assert "update public.strategies" in sql
assert "active = false" in sql
assert "coalesce(max(version), 0) + 1" in sql
assert "update public.strategy_nodes" in sql
assert "insert into public.strategy_nodes" in sql
# ledger + invariant
assert "kb_stage.finish_approved_proposal" in sql
assert "exactly one active strategy" in sql
# proximate_objectives rendered as jsonb
assert "::jsonb" in sql
def test_guarded_sql_canonicalizes_reviewed_timestamp_to_stable_utc_text():
payload = _revise_payload()["apply_payload"]
approval = _approval_meta("revise_strategy", payload)
approval["reviewed_at"] = "2026-07-10T02:00:00.1+02:00"
sql = ap.build_revise_strategy_sql(payload, "pid-1", "kb-apply", approval)
assert sql.count("2026-07-10T00:00:00.100000Z") == 2
assert "2026-07-10T02:00:00.1+02:00" not in sql
def test_guarded_sql_rejects_reviewed_timestamp_without_timezone():
payload = _revise_payload()["apply_payload"]
approval = _approval_meta("revise_strategy", payload)
approval["reviewed_at"] = "2026-07-10T00:00:00"
with pytest.raises(ValueError, match="timezone-aware ISO-8601"):
ap.build_revise_strategy_sql(payload, "pid-1", "kb-apply", approval)
def test_revise_strategy_requires_agent_id():
payload = _revise_payload()["apply_payload"]
del payload["agent_id"]
with pytest.raises(ValueError):
ap.build_revise_strategy_sql(payload, "pid-1", None, _approval_meta("revise_strategy", payload))
def test_revise_strategy_rejects_bad_node_type():
payload = _revise_payload()["apply_payload"]
payload["strategy_nodes"][0]["node_type"] = "manifesto"
with pytest.raises(ValueError):
ap.build_revise_strategy_sql(payload, "pid-1", None, _approval_meta("revise_strategy", payload))
# --- add_edge ------------------------------------------------------------- #
def test_add_edge_sql_dedup_guard():
payload = {"from_claim": CLAIM_A, "to_claim": CLAIM_B, "edge_type": "supersedes", "weight": 0.9}
sql = ap.build_add_edge_sql(payload, "pid-2", None, _approval_meta("add_edge", payload))
assert "insert into public.claim_edges" in sql
assert "not exists" in sql
assert "::edge_type" in sql
assert "kb_stage.finish_approved_proposal" in sql
assert "pg_advisory_xact_lock" in sql
def test_add_edge_semantic_row_accepts_match_and_rejects_weight_mismatch():
payload = {"from_claim": CLAIM_A, "to_claim": CLAIM_B, "edge_type": "supports", "weight": 0.9}
sql = ap.build_add_edge_sql(payload, "pid-2", None, _approval_meta("add_edge", payload))
verify = sql[sql.index("do $verify$") : sql.index("kb_stage.finish_approved_proposal")]
assert "weight is not distinct from 0.9" in verify
assert "weight is distinct from 0.9" in verify
assert "claim_edge row does not match strict apply payload" in verify
assert sql.startswith("begin;")
assert sql.rstrip().endswith("commit;")
assert sql.index("raise exception 'apply_proposal: claim_edge row") < sql.index("kb_stage.finish_approved_proposal")
def test_add_edge_rejects_self_loop():
payload = {"from_claim": "same", "to_claim": "same", "edge_type": "supports"}
with pytest.raises(ValueError):
ap.build_add_edge_sql(payload, "pid-2", None, _approval_meta("add_edge", payload))
# --- attach_evidence ------------------------------------------------------ #
def test_attach_evidence_sql_with_source_id():
payload = {"evidence": [{"claim_id": CLAIM_A, "source_id": SOURCE_A, "role": "grounds", "weight": 0.78}]}
sql = ap.build_attach_evidence_sql(payload, "pid-3", None, _approval_meta("attach_evidence", payload))
assert "insert into public.claim_evidence" in sql
assert "public.claim_evidence (id," not in sql
assert "public.claim_evidence (claim_id, source_id, role, weight)" in sql
assert "::evidence_role" in sql
assert "not exists" in sql
def test_attach_evidence_semantic_row_accepts_match_and_rejects_weight_mismatch():
payload = {"evidence": [{"claim_id": CLAIM_A, "source_id": SOURCE_A, "role": "grounds", "weight": 0.78}]}
sql = ap.build_attach_evidence_sql(payload, "pid-3", None, _approval_meta("attach_evidence", payload))
verify = sql[sql.index("do $verify$") : sql.index("kb_stage.finish_approved_proposal")]
assert "weight is not distinct from 0.78" in verify
assert "weight is distinct from 0.78" in verify
assert "evidence row % does not match strict apply payload" in verify
assert sql.startswith("begin;")
assert sql.rstrip().endswith("commit;")
assert sql.index("raise exception 'apply_proposal: evidence row") < sql.index("kb_stage.finish_approved_proposal")
def test_attach_evidence_requires_source_id():
payload = {"evidence": [{"claim_id": "cccc"}]}
with pytest.raises(ValueError):
ap.build_attach_evidence_sql(payload, "pid-3", None, _approval_meta("attach_evidence", payload))
# --- approve_claim canonical bundle -------------------------------------- #
def _approve_claim_bundle():
claim_a = "aaaaaaaa-aaaa-4aaa-8aaa-aaaaaaaaaaaa"
claim_b = "bbbbbbbb-bbbb-4bbb-8bbb-bbbbbbbbbbbb"
source_a = "cccccccc-cccc-4ccc-8ccc-cccccccccccc"
source_b = "dddddddd-dddd-4ddd-8ddd-dddddddddddd"
return {
"contract_version": 2,
"agent_id": "11111111-1111-4111-8111-111111111111",
"claims": [
{
"id": claim_a,
"type": "structural",
"text": "Canonical claim A",
"status": "open",
"confidence": 0.8,
"tags": ["test", "working-leo"],
"created_by": None,
},
{
"id": claim_b,
"type": "concept",
"text": "Canonical claim B",
"status": "open",
"confidence": None,
"tags": [],
"created_by": None,
},
],
"sources": [
{
"id": source_a,
"source_type": "paper",
"url": "https://example.test/a",
"storage_path": None,
"excerpt": "Evidence A",
"hash": "hash-a",
"created_by": None,
},
{
"id": source_b,
"source_type": "observation",
"url": None,
"storage_path": None,
"excerpt": "Evidence B",
"hash": "hash-b",
"created_by": None,
},
],
"evidence": [
{"claim_id": claim_a, "source_id": source_a, "role": "grounds", "weight": 0.9},
{"claim_id": claim_b, "source_id": source_b, "role": "illustrates", "weight": None},
],
"edges": [{"from_claim": claim_a, "to_claim": claim_b, "edge_type": "supports", "weight": 0.7}],
"reasoning_tools": [
{
"id": "eeeeeeee-eeee-4eee-8eee-eeeeeeeeeeee",
"agent_id": None,
"name": "Canonical test tool",
"description": "A strict reasoning-tool row.",
"category": "test",
}
],
}
def _approve_claim_v3_bundle():
agent_id = "11111111-1111-4111-8111-111111111111"
source_b = "dddddddd-dddd-4ddd-8ddd-dddddddddddd"
evidence_a = {
"id": "eeeeeeee-eeee-4eee-8eee-eeeeeeeeeeee",
"claim_id": CLAIM_A,
"source_id": SOURCE_A,
"polarity": "supports",
"excerpt": "The primary record supports claim A.",
"locator_json": {"section": "A", "paragraph": 1},
"source_content_hash": "a" * 64,
"rationale": "The cited paragraph directly supports the bounded proposition.",
"created_by": agent_id,
}
evidence_b = {
"id": "ffffffff-ffff-4fff-8fff-ffffffffffff",
"claim_id": CLAIM_B,
"source_id": source_b,
"polarity": "challenges",
"excerpt": "The primary record challenges claim B.",
"locator_json": {"section": "B", "paragraph": 2},
"source_content_hash": "b" * 64,
"rationale": "The cited paragraph directly challenges the bounded proposition.",
"created_by": agent_id,
}
evidence = [evidence_a, evidence_b]
return {
"contract_version": 3,
"agent_id": agent_id,
"claims": [
{
"id": CLAIM_A,
"type": "empirical",
"proposition": "Claim A has a reproducible evidence chain.",
"body_markdown": "Claim A is bounded to the cited primary record and exact locator.",
"scope": "V3 apply-writer disposable verification.",
"access_scope": "collective_internal",
"status": "active",
"revision_criteria": "Revise if the cited source or locator cannot be reproduced.",
"revision_kind": "original",
"supersedes_id": None,
"tags": ["v3", "apply-writer"],
"created_by": agent_id,
},
{
"id": CLAIM_B,
"type": "conceptual",
"proposition": "Claim B remains explicitly challengeable.",
"body_markdown": "Claim B records a conceptual object together with its challenge evidence.",
"scope": "V3 apply-writer disposable verification.",
"access_scope": "collective_internal",
"status": "contested",
"revision_criteria": "Revise when the challenge is resolved or materially changes.",
"revision_kind": "original",
"supersedes_id": None,
"tags": ["v3", "challenge"],
"created_by": agent_id,
},
],
"sources": [
{
"id": SOURCE_A,
"source_type": "paper",
"canonical_url": "https://example.test/v3/a",
"storage_uri": None,
"excerpt": "Primary source A.",
"content_hash": "a" * 64,
"access_scope": "collective_internal",
"ingestion_origin": "operator_upload",
"provenance_status": "verified",
"source_class": "primary_record",
"auto_promotion_policy": "human_required",
"captured_at": "2026-07-18T10:00:00+00:00",
"created_by": agent_id,
},
{
"id": source_b,
"source_type": "observation",
"canonical_url": None,
"storage_uri": "gs://livingip-test/v3/source-b",
"excerpt": "Primary source B.",
"content_hash": "b" * 64,
"access_scope": "collective_internal",
"ingestion_origin": "system_import",
"provenance_status": "verified",
"source_class": "primary_record",
"auto_promotion_policy": "human_required",
"captured_at": "2026-07-18T10:01:00+00:00",
"created_by": agent_id,
},
],
"evidence": evidence,
"edges": [
{
"id": "12121212-1212-4121-8121-121212121212",
"from_claim": CLAIM_A,
"to_claim": CLAIM_B,
"relation_type": "challenges",
"rationale": "Claim A supplies a bounded reason to challenge claim B.",
"scope_or_conditions": "Only inside the disposable V3 verification scope.",
"created_by": agent_id,
}
],
"assessments": [
{
"id": "13131313-1313-4131-8131-131313131313",
"claim_id": CLAIM_A,
"support_tier": "strong",
"challenge_tier": "none",
"overall_state": "support_dominant",
"rationale": "One exact primary record supports the bounded empirical claim.",
"evidence_set_hash": ap._v3_evidence_set_hash(CLAIM_A, evidence),
"supersedes_assessment_id": None,
},
{
"id": "14141414-1414-4141-8141-141414141414",
"claim_id": CLAIM_B,
"support_tier": None,
"challenge_tier": None,
"overall_state": "challenge_dominant",
"rationale": "The conceptual claim has an explicit challenge without numeric evidence tiers.",
"evidence_set_hash": ap._v3_evidence_set_hash(CLAIM_B, evidence),
"supersedes_assessment_id": None,
},
],
"reasoning_tools": [],
}
def test_approve_claim_bundle_sql_is_atomic_and_row_verified():
payload = _approve_claim_bundle()
sql = ap.build_approve_claim_sql(
payload,
"99999999-9999-4999-8999-999999999999",
None,
_approval_meta("approve_claim", payload),
)
assert sql.startswith("begin;")
assert sql.rstrip().endswith("commit;")
assert "insert into public.claims" in sql
assert "insert into public.sources" in sql
assert "insert into public.claim_evidence" in sql
assert "public.claim_evidence (id," not in sql
assert "public.claim_evidence (claim_id, source_id, role, weight, created_by)" in sql
assert "insert into public.claim_edges" in sql
assert "insert into public.reasoning_tools" in sql
assert "source hash collision" in sql
assert "claim row does not match strict apply payload" in sql
assert "source row does not match strict apply payload" in sql
assert "reasoning tool row does not match strict apply payload" in sql
assert "kb_stage.finish_approved_proposal" in sql
def test_v3_approve_claim_sql_writes_complete_epistemic_bundle() -> None:
payload = _approve_claim_v3_bundle()
proposal_id = "99999999-9999-4999-8999-999999999999"
sql = ap.build_approve_claim_sql(
payload,
proposal_id,
None,
_approval_meta("approve_claim", payload),
)
shared_lock = f"select pg_catalog.pg_advisory_lock_shared({ap.V3_CATALOG_DDL_ADVISORY_LOCK_KEY});"
shared_unlock = f"select pg_catalog.pg_advisory_unlock_shared({ap.V3_CATALOG_DDL_ADVISORY_LOCK_KEY});"
assert sql.startswith(f"{shared_lock}\nbegin;")
assert "set transaction isolation level serializable" in sql
assert sql.rstrip().endswith(shared_unlock)
assert sql.count(shared_lock) == 1
assert sql.count(shared_unlock) == 1
assert "insert into public.claims" in sql
assert "body_markdown" in sql
assert "accepted_by_proposal_id" in sql
assert "insert into public.sources" in sql
assert "content_hash" in sql
assert "insert into public.claim_evidence (" in sql
assert "locator_json" in sql
assert "insert into public.claim_edges (" in sql
assert "relation_type" in sql
assert "insert into public.claim_evidence_assessments" in sql
assert "evidence_set_hash" in sql
assert "insert into public.reasoning_tools" not in sql
expected_inserts = sum(len(payload[key]) for key in ("sources", "claims", "evidence", "edges", "assessments"))
assert sql.lower().count("on conflict do nothing") == expected_inserts
assert sql.index("do $v3_contract_guard$") < sql.index("kb_stage.assert_approved_proposal")
relation_locks = [
f"select kb_stage.{ap.V3_CATALOG_LOCK_FUNCTION}();",
"lock table public.claim_edges in row exclusive mode;",
"lock table public.claim_evidence in row exclusive mode;",
"lock table public.claim_evidence_assessments in row exclusive mode;",
"lock table public.claims in row exclusive mode;",
"lock table public.sources in row exclusive mode;",
]
ordered_catalog_guard = [
shared_lock,
"begin;",
"set transaction isolation level serializable;",
*relation_locks,
"do $v3_contract_guard$",
"commit;",
shared_unlock,
]
assert [sql.index(statement) for statement in ordered_catalog_guard] == sorted(
sql.index(statement) for statement in ordered_catalog_guard
)
assert sql.index("kb_stage.assert_approved_proposal") < sql.index("insert into public.sources")
assert sql.index("insert into public.claim_evidence_assessments") < sql.index("kb_stage.finish_approved_proposal")
assert sql.index("on conflict do nothing") < sql.index("do $verify$")
assert "set local timezone = 'UTC'" in sql
assert "apply_proposal: V3 catalog contract is %s; refusing mutating transaction" in sql
assert "and superseded_by is null" in sql
expected_creator_checks = sum(len(payload[key]) for key in ("sources", "claims", "evidence", "edges"))
assert sql.count("created_by is not distinct from") == expected_creator_checks
def test_v3_catalog_guard_requires_the_full_hardened_contract() -> None:
sql = ap.build_v3_contract_state_query()
assert ap.V3_CATALOG_DDL_ADVISORY_LOCK_KEY == 6072343533146485505
assert ap.V3_PROTECTED_RELATIONS == (
("kb_stage", "kb_proposals"),
("public", "claim_edges"),
("public", "claim_evidence"),
("public", "claim_evidence_assessments"),
("public", "claims"),
("public", "sources"),
)
assert (
tuple(sorted({(row["schema_name"], row["table_name"]) for row in ap.V3_EXPECTED_TRIGGER_CONTRACT}))
== ap.V3_PROTECTED_RELATIONS
)
assert len(ap.V3_EXPECTED_TRIGGER_CONTRACT) == 24
assert len({row["trigger_name"] for row in ap.V3_EXPECTED_TRIGGER_CONTRACT}) == 24
assert all(row["when_expression"] is None for row in ap.V3_EXPECTED_TRIGGER_CONTRACT)
assert "when not artifacts_present then 'V2'" in sql
assert "when functions_valid and catalog_lock_function_valid and triggers_valid" in sql
assert "and owner_role_valid and memberships_valid then 'V3'" in sql
assert "else 'INVALID'" in sql
assert "actual.tgenabled = 'O'" in sql
assert "actual.tgtype = expected.trigger_type" in sql
assert "actual.update_columns = expected.update_columns" in sql
assert "actual.when_expression is not distinct from expected.when_expression" in sql
assert "actual.argument_hex = expected.argument_hex" in sql
assert "actual.tgconstraint = 0::pg_catalog.oid" in sql
assert "actual.tgoldtable is null" in sql
assert "actual.tgnewtable is null" in sql
assert "actual.tgfoid" in sql
assert "owner_role.rolname = E'kb_gate_owner'" in sql
assert "actual.function_security_definer = expected.function_security_definer" in sql
assert "actual.function_source_sha256 = expected.function_source_sha256" in sql
assert f"procedure.proname = E'{ap.V3_CATALOG_LOCK_FUNCTION}'" in sql
assert ap.V3_CATALOG_LOCK_FUNCTION_SOURCE_SHA256 in sql
assert "and prorettype = 'pg_catalog.void'::pg_catalog.regtype" in sql
assert "and grantee.rolname = 'kb_apply'" in sql
assert "and execute_acl_exact" in sql
assert "acl.grantee <> procedure.proowner" in sql
assert "from pg_catalog.pg_auth_members membership" in sql
def test_migration_trigger_contract_and_catalog_queries_match_worker_builder() -> None:
migration = (REPO_ROOT / "ops" / "teleo_v3_epistemic_contract.sql").read_text(encoding="utf-8")
contract_match = re.search(r"\$trigger_contract\$\s*(\[.*?\])\s*\$trigger_contract\$", migration, re.DOTALL)
assert contract_match is not None
assert json.loads(contract_match.group(1)) == list(ap.V3_EXPECTED_TRIGGER_CONTRACT)
exclusive_lock = f"select pg_catalog.pg_advisory_xact_lock({ap.V3_CATALOG_DDL_ADVISORY_LOCK_KEY});"
assert migration.count(exclusive_lock) == 1
assert migration.index(exclusive_lock) < migration.index(
f"create or replace function kb_stage.{ap.V3_CATALOG_LOCK_FUNCTION}()"
)
assert migration.index(exclusive_lock) < migration.index(
"create or replace function kb_stage.teleo_v3_enforce_apply_contract_cutover()"
)
assert migration.index(exclusive_lock) < migration.index("create or replace trigger")
expected_query = ap.build_v3_contract_state_query(
into="v_contract_state",
expected_contract_expression="pg_catalog.current_setting('teleo.v3_expected_trigger_contract')",
)
for marker in ("STATE", "POSTFLIGHT"):
query_match = re.search(
rf"TELEO_V3_CONTRACT_{marker}_QUERY_BEGIN\n(.*?)\n\s*-- TELEO_V3_CONTRACT_{marker}_QUERY_END",
migration,
re.DOTALL,
)
assert query_match is not None
assert query_match.group(1).strip() == expected_query + ";"
@pytest.mark.parametrize(
("mutation", "error"),
[
(lambda payload: payload["claims"][0].update(body_markdown=""), "body_markdown"),
(lambda payload: payload["claims"][0].update(revision_kind="semantic"), "original claims only"),
(lambda payload: payload["sources"][0].update(content_hash="not-a-hash"), "SHA-256"),
(lambda payload: payload["evidence"][0].update(source_content_hash="b" * 64), "does not match"),
(lambda payload: payload["assessments"][0].update(evidence_set_hash="c" * 64), "evidence_set_hash"),
(
lambda payload: payload["assessments"][0].update(
supersedes_assessment_id="15151515-1515-4151-8151-151515151515"
),
"assessment supersession",
),
(lambda payload: payload.update(reasoning_tools=[{"id": CLAIM_A}]), "reasoning_tools"),
],
)
def test_v3_approve_claim_rejects_incomplete_or_unbound_payloads(mutation, error) -> None:
payload = _approve_claim_v3_bundle()
mutation(payload)
with pytest.raises(ValueError, match=error):
ap.build_approve_claim_sql(
payload,
"99999999-9999-4999-8999-999999999999",
None,
_approval_meta("approve_claim", payload),
)
def test_v3_approve_claim_preserves_exact_reviewed_text_for_replay_receipts() -> None:
payload = _approve_claim_v3_bundle()
payload["claims"][0]["body_markdown"] = " Exact reviewed body with intentional spacing. "
normalized = ap._normalize_v3_approve_claim(payload)
assert normalized["claims"][0]["body_markdown"] == payload["claims"][0]["body_markdown"]
@pytest.mark.parametrize(
("captured_at", "expected"),
[
("2026-07-18T10:00:00Z", "2026-07-18T10:00:00.000000Z"),
("2026-07-18T12:00:00+02:00", "2026-07-18T10:00:00.000000Z"),
("2026-07-18T12:00:00.123400+02:00", "2026-07-18T10:00:00.123400Z"),
("2026-07-18T10:00:00.1Z", "2026-07-18T10:00:00.100000Z"),
],
)
def test_v3_approve_claim_normalizes_capture_timestamp_to_fixed_utc(
captured_at: str,
expected: str,
) -> None:
payload = _approve_claim_v3_bundle()
payload["sources"][0]["captured_at"] = captured_at
normalized = ap._normalize_v3_approve_claim(payload)
assert normalized["sources"][0]["captured_at"] == expected
@pytest.mark.parametrize(
"captured_at",
[
"2026-07-18 10:00:00+00:00",
"2026-07-18T10:00:00",
"2026-07-18T10:00:00.1234567Z",
"not-a-timestamp",
],
)
def test_v3_approve_claim_rejects_untyped_capture_timestamp(captured_at: str) -> None:
payload = _approve_claim_v3_bundle()
payload["sources"][0]["captured_at"] = captured_at
with pytest.raises(ValueError, match="timezone-aware ISO-8601"):
ap._normalize_v3_approve_claim(payload)
@pytest.mark.parametrize(
"spelling",
[
CLAIM_A.upper(),
"{" + CLAIM_A + "}",
CLAIM_A.replace("-", ""),
],
)
def test_v3_approve_claim_rejects_noncanonical_uuid_spellings(spelling: str) -> None:
payload = _approve_claim_v3_bundle()
payload["claims"][0]["id"] = spelling
with pytest.raises(ValueError, match="canonical lowercase hyphenated UUID"):
ap._normalize_v3_approve_claim(payload)
def test_approve_claim_semantic_rows_accept_matches_and_reject_payload_mismatches():
payload = _approve_claim_bundle()
sql = ap.build_approve_claim_sql(
payload,
"99999999-9999-4999-8999-999999999999",
None,
_approval_meta("approve_claim", payload),
)
verify = sql[sql.index("do $verify$") : sql.index("kb_stage.finish_approved_proposal")]
assert "weight is not distinct from 0.9" in verify
assert "weight is distinct from 0.9" in verify
assert "weight is not distinct from 0.7" in verify
assert "weight is distinct from 0.7" in verify
assert "weight is not distinct from null" in verify
assert "weight is distinct from null" in verify
expected_creator = "E'11111111-1111-4111-8111-111111111111'::uuid"
assert f"created_by is not distinct from {expected_creator}" in verify
assert f"created_by is distinct from {expected_creator}" in verify
assert "approve_claim: evidence row does not match strict apply payload" in verify
assert "approve_claim: edge row does not match strict apply payload" in verify
assert sql.startswith("begin;")
assert sql.rstrip().endswith("commit;")
assert sql.index("raise exception 'approve_claim: evidence row") < sql.index("kb_stage.finish_approved_proposal")
def test_approve_claim_bundle_rejects_duplicate_source_hashes():
payload = _approve_claim_bundle()
payload["sources"][1]["hash"] = payload["sources"][0]["hash"]
with pytest.raises(ValueError):
ap.build_approve_claim_sql(
payload,
"99999999-9999-4999-8999-999999999999",
None,
_approval_meta("approve_claim", payload),
)
def test_approve_claim_bundle_rejects_invalid_enum_and_self_edge():
payload = _approve_claim_bundle()
payload["claims"][0]["type"] = "vibes"
with pytest.raises(ValueError):
ap.build_approve_claim_sql(
payload,
"99999999-9999-4999-8999-999999999999",
None,
_approval_meta("approve_claim", payload),
)
payload = _approve_claim_bundle()
payload["edges"][0]["to_claim"] = payload["edges"][0]["from_claim"]
with pytest.raises(ValueError):
ap.build_approve_claim_sql(
payload,
"99999999-9999-4999-8999-999999999999",
None,
_approval_meta("approve_claim", payload),
)
def test_approve_claim_rejects_unknown_keys_and_contract_versions():
payload = _approve_claim_bundle()
payload["contract_version"] = 1
with pytest.raises(ValueError):
ap.build_approve_claim_sql(
payload,
"99999999-9999-4999-8999-999999999999",
None,
_approval_meta("approve_claim", payload),
)
payload = _approve_claim_bundle()
payload["contract_version"] = 2
payload["claimz"] = payload.pop("claims")
with pytest.raises(ValueError):
ap.build_approve_claim_sql(
payload,
"99999999-9999-4999-8999-999999999999",
None,
_approval_meta("approve_claim", payload),
)
def test_approve_claim_verifies_superseded_by_and_pins_string_mode():
payload = _approve_claim_bundle()
payload["contract_version"] = 2
payload["claims"][0]["superseded_by"] = None
sql = ap.build_approve_claim_sql(
payload,
"99999999-9999-4999-8999-999999999999",
None,
_approval_meta("approve_claim", payload),
)
assert "superseded_by" in sql
assert "set local standard_conforming_strings = on" in sql.lower()
def test_apply_sql_locks_and_binds_approval_before_canonical_write():
payload = _approve_claim_bundle()
payload["contract_version"] = 2
proposal = {
"id": "99999999-9999-4999-8999-999999999999",
"proposal_type": "approve_claim",
"status": "approved",
"payload": {"apply_payload": payload},
"reviewed_by_handle": "m3ta",
"reviewed_by_agent_id": None,
"reviewed_at": "2026-07-10T00:00:00+00:00",
"review_note": "Reviewed exact strict payload.",
}
sql = ap.build_apply_sql(proposal, "kb-apply")
assert "kb_stage.assert_approved_proposal" in sql
assert "kb_stage.finish_approved_proposal" in sql
assert sql.index("kb_stage.assert_approved_proposal") < sql.index("insert into public.claims")
assert sql.index("insert into public.claims") < sql.index("kb_stage.finish_approved_proposal")
assert "Reviewed exact strict payload." in sql
def test_fresh_owned_apply_uses_strict_inserts_and_binds_preflight_hash():
payload = _approve_claim_bundle()
proposal = {
"id": "99999999-9999-4999-8999-999999999999",
"proposal_type": "approve_claim",
"status": "approved",
"payload": {"apply_payload": payload},
"reviewed_by_handle": "m3ta",
"reviewed_by_agent_id": None,
"reviewed_at": "2026-07-10T00:00:00+00:00",
"review_note": "Reviewed exact strict payload.",
}
preflight_sha256 = "a" * 64
sql = ap.build_apply_sql(
proposal,
"kb-apply",
fresh_owned=True,
fresh_preflight_sha256=preflight_sha256,
)
assert preflight_sha256 in sql
assert "set transaction isolation level serializable" in sql.lower()
assert "on conflict" not in sql.lower()
assert "where not exists" not in sql.lower()
assert sql.index("set transaction isolation level serializable") < sql.index("kb_stage.assert_approved_proposal")
def test_build_apply_sql_requires_review_provenance():
proposal = {
"id": "99999999-9999-4999-8999-999999999999",
"proposal_type": "approve_claim",
"status": "approved",
"payload": {"apply_payload": _approve_claim_bundle()},
}
with pytest.raises(ValueError):
ap.build_apply_sql(proposal, "kb-apply")
def test_approve_claim_dispatch_and_prerequisite_grants():
proposal = {
"id": "99999999-9999-4999-8999-999999999999",
"proposal_type": "approve_claim",
"payload": {"apply_payload": _approve_claim_bundle()},
}
proposal.update(_approval_meta("approve_claim", proposal["payload"]["apply_payload"]))
sql = ap.build_apply_sql(proposal, None)
assert "insert into public.claims" in sql
assert "approve_claim" in ap.APPLYABLE_TYPES
prereqs = (REPO_ROOT / "scripts" / "kb_apply_prereqs.sql").read_text(encoding="utf-8")
assert "grant select, insert on public.claims to kb_apply" in prereqs.lower()
assert "grant select, insert on public.sources to kb_apply" in prereqs.lower()
assert "grant select, insert on public.reasoning_tools to kb_apply" in prereqs.lower()
# --- dispatch + guards ---------------------------------------------------- #
def test_build_apply_sql_requires_apply_payload():
proposal = {"id": "pid", "proposal_type": "add_edge", "payload": {"rationale": "x"}}
with pytest.raises(ValueError):
ap.build_apply_sql(proposal, None)
@pytest.mark.parametrize(
("proposal", "error"),
[
({"proposal_type": "approve_claim", "payload": []}, "payload must be a JSON object"),
(
{"proposal_type": "approve_claim", "payload": {"apply_payload": []}},
"payload.apply_payload must be a JSON object",
),
(
{
"proposal_type": "approve_claim",
"payload": {"apply_payload": {"contract_version": "3"}},
},
"must be integer 2 or 3",
),
(
{
"proposal_type": "approve_claim",
"payload": {"apply_payload": {"contract_version": 3.0}},
},
"must be integer 2 or 3",
),
(
{
"proposal_type": "approve_claim",
"payload": {"apply_payload": {"contract_version": True}},
},
"must be integer 2 or 3",
),
(
{
"proposal_type": "add_edge",
"payload": {"apply_payload": {"contract_version": 3}},
},
"must omit contract_version",
),
(
{
"proposal_type": "add_edge",
"payload": {"apply_payload": {"contract_version": None}},
},
"must omit contract_version",
),
],
)
def test_exact_contract_gate_rejects_malformed_and_hybrid_proposals(
proposal: dict,
error: str,
) -> None:
with pytest.raises(ValueError, match=error):
ap.proposal_contract_version(proposal)
def test_mutating_subprocess_contract_expectation_is_reasserted() -> None:
payload = _approve_claim_v3_bundle()
proposal = {
"id": "99999999-9999-4999-8999-999999999999",
"proposal_type": "approve_claim",
"payload": {"apply_payload": payload},
}
assert ap.assert_expected_contract_version(proposal, 3) == 3
with pytest.raises(ValueError, match="does not match expected version 2"):
ap.assert_expected_contract_version(proposal, 2)
def test_build_apply_sql_rejects_unsupported_type():
proposal = {"id": "pid", "proposal_type": "reject_claim", "payload": {"apply_payload": {}}}
with pytest.raises(ValueError):
ap.build_apply_sql(proposal, None)
def test_assert_applyable_blocks_pending():
with pytest.raises(SystemExit):
ap.assert_applyable({"id": "pid", "status": "pending_review"})
def test_assert_applyable_blocks_already_applied():
with pytest.raises(SystemExit):
ap.assert_applyable({"id": "pid", "status": "applied"})
def test_assert_applyable_allows_approved():
ap.assert_applyable({"id": "pid", "status": "approved"}) # no raise
# --- ledger flip: concurrency guard + FK stamp --------------------------- #
def test_ledger_transition_uses_payload_bound_security_functions():
payload = {"from_claim": CLAIM_A, "to_claim": CLAIM_B, "edge_type": "supports"}
sql = ap.build_add_edge_sql(payload, "pid", None, _approval_meta("add_edge", payload))
assert "kb_stage.assert_approved_proposal" in sql
assert "kb_stage.finish_approved_proposal" in sql
assert sql.index("kb_stage.assert_approved_proposal") < sql.index("insert into public.claim_edges")
assert sql.index("insert into public.claim_edges") < sql.index("kb_stage.finish_approved_proposal")
prereqs = (REPO_ROOT / "scripts" / "kb_apply_prereqs.sql").read_text(encoding="utf-8").lower()
assert "status = 'approved'" in prereqs
assert "for update" in prereqs
assert "payload = p_payload" in prereqs
def test_ledger_flip_stamps_agent_fk():
payload = {"from_claim": CLAIM_A, "to_claim": CLAIM_B, "edge_type": "supports"}
sql = ap.build_add_edge_sql(payload, "pid", "kb-apply", _approval_meta("add_edge", payload))
# Hard resolve into a variable + NOT-NULL assert, never a silent inline
# subselect that would stamp NULL on an unresolved handle.
assert "kb_stage.finish_approved_proposal" in sql
assert "'kb-apply'" in sql
def test_build_apply_sql_defaults_applied_by_to_service_agent():
payload = {"from_claim": CLAIM_A, "to_claim": CLAIM_B, "edge_type": "supports"}
proposal = {
"id": "pid",
"proposal_type": "add_edge",
"payload": {"apply_payload": payload},
}
proposal.update(_approval_meta("add_edge", payload))
sql = ap.build_apply_sql(proposal, None)
assert f"E'{ap.SERVICE_AGENT_HANDLE}'" in sql
def test_load_password_accepts_live_db_password_key(tmp_path):
secrets_file = tmp_path / "kb-apply.env"
secrets_file.write_text("KB_APPLY_DB_PASSWORD=live-secret\n", encoding="utf-8")
assert ap.load_password(str(secrets_file)) == "live-secret"
if __name__ == "__main__":
import traceback
failures = 0
for name, fn in sorted(globals().items()):
if name.startswith("test_") and callable(fn):
try:
fn()
print(f"PASS {name}")
except Exception:
failures += 1
print(f"FAIL {name}")
traceback.print_exc()
sys.exit(1 if failures else 0)