teleo-infrastructure/scripts/kb_proposal_normalize.py
2026-07-09 03:01:40 +02:00

345 lines
13 KiB
Python

#!/usr/bin/env python3
"""Preview strict apply-payload normalization for staged KB proposals.
Leo and reviewers can approve rich proposal intent before it is mechanically
safe to apply. The apply worker intentionally accepts only strict contracts:
``add_edge``, ``attach_evidence``, and ``revise_strategy`` proposals that carry
``payload.apply_payload`` with canonical IDs.
This helper is read-only by default. It inspects a freeform proposal and reports:
* strict child proposals that can be emitted because all canonical IDs exist in
the proposal payload;
* blocked fragments that still require claim/source creation, source dedup, or a
schema decision before canonical apply is possible.
"""
from __future__ import annotations
import argparse
import hashlib
import json
import re
import sys
from pathlib import Path
from typing import Any
HERE = Path(__file__).resolve().parent
sys.path.insert(0, str(HERE))
import apply_proposal as ap # noqa: E402
import kb_proposal_review_packet as review_packet # noqa: E402
UUID_RE = re.compile(
r"^[0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{12}$"
)
def _payload_dict(payload: dict[str, Any], key: str) -> dict[str, Any]:
value = payload.get(key)
return value if isinstance(value, dict) else {}
def _payload_list(payload: dict[str, Any], key: str) -> list[Any]:
value = payload.get(key)
return value if isinstance(value, list) else []
def _is_uuid(value: Any) -> bool:
return isinstance(value, str) and bool(UUID_RE.match(value))
def _first_uuid(fragment: dict[str, Any], keys: tuple[str, ...]) -> str | None:
for key in keys:
value = fragment.get(key)
if _is_uuid(value):
return value
return None
def _fingerprint(value: Any) -> str:
raw = json.dumps(value, sort_keys=True, separators=(",", ":"))
return hashlib.sha256(raw.encode("utf-8")).hexdigest()[:16]
def _compact_fragment(fragment: Any) -> dict[str, Any]:
if not isinstance(fragment, dict):
return {"value_type": type(fragment).__name__, "fingerprint": _fingerprint(fragment)}
keys = [
"claim_key",
"source_key",
"headline",
"title",
"edge_type",
"target",
"target_claim_id",
"to_claim_id",
"to_existing_claim",
"from_claim_id",
"from_claim",
"claim_id",
"source_id",
]
compact = {key: fragment[key] for key in keys if key in fragment}
compact["fingerprint"] = _fingerprint(fragment)
return compact
def _blocked(kind: str, reason: str, missing: list[str], fragment: Any) -> dict[str, Any]:
return {
"kind": kind,
"reason": reason,
"missing": missing,
"fragment": _compact_fragment(fragment),
}
def _edge_candidate(fragment: dict[str, Any], default_from_claim: str | None = None) -> dict[str, Any] | None:
edge_type = fragment.get("edge_type")
from_claim = _first_uuid(fragment, ("from_claim", "from_claim_id", "from_claim_uuid")) or default_from_claim
to_claim = _first_uuid(
fragment,
(
"to_claim",
"to_claim_id",
"target_claim_id",
"target_claim",
"to_existing_claim",
"existing_claim_id",
),
)
if from_claim and to_claim and edge_type:
result: dict[str, Any] = {"from_claim": from_claim, "to_claim": to_claim, "edge_type": edge_type}
if fragment.get("weight") is not None:
result["weight"] = fragment.get("weight")
return result
return None
def _evidence_candidate(fragment: dict[str, Any], default_claim_id: str | None = None) -> dict[str, Any] | None:
claim_id = _first_uuid(fragment, ("claim_id", "target_claim_id", "to_claim_id")) or default_claim_id
source_id = _first_uuid(fragment, ("source_id", "source_uuid", "canonical_source_id"))
if claim_id and source_id:
result: dict[str, Any] = {"claim_id": claim_id, "source_id": source_id}
if fragment.get("role"):
result["role"] = fragment.get("role")
if fragment.get("weight") is not None:
result["weight"] = fragment.get("weight")
return result
return None
def _strict_child(kind: str, apply_payload: dict[str, Any], source_fragment: Any) -> dict[str, Any]:
return {
"proposal_type": kind,
"payload": {"apply_payload": apply_payload},
"source_fragment_fingerprint": _fingerprint(source_fragment),
}
def _normalize_edges(payload: dict[str, Any]) -> tuple[list[dict[str, Any]], list[dict[str, Any]]]:
strict: list[dict[str, Any]] = []
blocked: list[dict[str, Any]] = []
old_claim_id = _first_uuid(_payload_dict(payload, "old_claim"), ("id", "claim_id"))
edge_sources: list[tuple[dict[str, Any], str | None]] = []
for edge in _payload_list(payload, "proposed_supersession_edges"):
if isinstance(edge, dict):
edge_sources.append((edge, None))
for candidate in _payload_list(payload, "claim_candidates"):
if not isinstance(candidate, dict):
continue
candidate_claim_id = _first_uuid(candidate, ("claim_id", "id", "canonical_claim_id"))
for edge in _payload_list(candidate, "edges"):
if isinstance(edge, dict):
edge_sources.append((edge, candidate_claim_id))
if candidate_claim_id and old_claim_id:
edge_sources.append(
(
{
"from_claim": candidate_claim_id,
"to_claim": old_claim_id,
"edge_type": "supersedes_component_of",
},
None,
)
)
seen = set()
for fragment, default_from_claim in edge_sources:
candidate = _edge_candidate(fragment, default_from_claim)
if candidate:
key = json.dumps(candidate, sort_keys=True)
if key not in seen:
seen.add(key)
strict.append(_strict_child("add_edge", candidate, fragment))
continue
missing = []
if not fragment.get("edge_type"):
missing.append("edge_type")
if not (_first_uuid(fragment, ("from_claim", "from_claim_id", "from_claim_uuid")) or default_from_claim):
missing.append("canonical from_claim")
if not _first_uuid(
fragment,
("to_claim", "to_claim_id", "target_claim_id", "target_claim", "to_existing_claim", "existing_claim_id"),
):
missing.append("canonical to_claim")
blocked.append(_blocked("edge", "edge fragment is not a strict add_edge contract", missing, fragment))
return strict, blocked
def _normalize_evidence(payload: dict[str, Any]) -> tuple[list[dict[str, Any]], list[dict[str, Any]]]:
evidence_rows: list[dict[str, Any]] = []
blocked: list[dict[str, Any]] = []
evidence_sources: list[tuple[dict[str, Any], str | None]] = []
for evidence in _payload_list(payload, "evidence"):
if isinstance(evidence, dict):
evidence_sources.append((evidence, None))
for evidence in _payload_list(payload, "evidence_candidates"):
if isinstance(evidence, dict):
evidence_sources.append((evidence, None))
for candidate in _payload_list(payload, "claim_candidates"):
if not isinstance(candidate, dict):
continue
candidate_claim_id = _first_uuid(candidate, ("claim_id", "id", "canonical_claim_id"))
for evidence in _payload_list(candidate, "evidence"):
if isinstance(evidence, dict):
evidence_sources.append((evidence, candidate_claim_id))
seen = set()
for fragment, default_claim_id in evidence_sources:
candidate = _evidence_candidate(fragment, default_claim_id)
if candidate:
key = json.dumps(candidate, sort_keys=True)
if key not in seen:
seen.add(key)
evidence_rows.append(candidate)
continue
missing = []
if not (_first_uuid(fragment, ("claim_id", "target_claim_id", "to_claim_id")) or default_claim_id):
missing.append("canonical claim_id")
if not _first_uuid(fragment, ("source_id", "source_uuid", "canonical_source_id")):
missing.append("canonical source_id")
blocked.append(_blocked("evidence", "evidence fragment is not a strict attach_evidence contract", missing, fragment))
strict = [_strict_child("attach_evidence", {"evidence": evidence_rows}, evidence_rows)] if evidence_rows else []
return strict, blocked
def _normalize_strategy(proposal: dict[str, Any], payload: dict[str, Any]) -> tuple[list[dict[str, Any]], list[dict[str, Any]]]:
if proposal.get("proposal_type") != "revise_strategy":
return [], []
strategy_payload = payload.get("apply_payload") or {
key: payload.get(key) for key in ("agent_id", "strategy", "strategy_nodes") if key in payload
}
has_contract = all(strategy_payload.get(key) for key in ("agent_id", "strategy", "strategy_nodes"))
if has_contract:
return [_strict_child("revise_strategy", strategy_payload, payload)], []
return [], [
_blocked(
"strategy",
"strategy revision needs agent_id, strategy, and strategy_nodes",
[key for key in ("agent_id", "strategy", "strategy_nodes") if not strategy_payload.get(key)],
payload,
)
]
def _dedupe_children(children: list[dict[str, Any]]) -> list[dict[str, Any]]:
seen = set()
out = []
for child in children:
key = json.dumps({"proposal_type": child["proposal_type"], "payload": child["payload"]}, sort_keys=True)
if key in seen:
continue
seen.add(key)
out.append(child)
return out
def normalize_proposal(proposal: dict[str, Any]) -> dict[str, Any]:
payload = proposal.get("payload") or {}
if not isinstance(payload, dict):
payload = {}
if payload.get("apply_payload") and proposal.get("proposal_type") in ap.APPLYABLE_TYPES:
return {
"proposal_id": proposal.get("id"),
"normalization_state": "already_strict_apply_payload",
"strict_child_proposals": [],
"blocked_fragments": [],
"blocked_count": 0,
"strict_child_count": 0,
"next_normalization_action": "Proposal already carries a strict apply_payload; use dry-run/apply review.",
}
strict_children: list[dict[str, Any]] = []
blocked_fragments: list[dict[str, Any]] = []
for strict, blocked in (
_normalize_edges(payload),
_normalize_evidence(payload),
_normalize_strategy(proposal, payload),
):
strict_children.extend(strict)
blocked_fragments.extend(blocked)
strict_children = _dedupe_children(strict_children)
if strict_children and not blocked_fragments:
state = "strict_children_ready"
next_action = "Review generated strict child proposal JSON, then stage as approved apply_payload proposals."
elif strict_children:
state = "partial_strict_children_ready"
next_action = "Stage only the strict children after review; resolve blocked fragments separately."
elif blocked_fragments:
state = "blocked_missing_canonical_ids"
next_action = "Create or identify canonical claim/source rows, then rerun normalization."
else:
state = "no_supported_apply_fragments"
next_action = "Rewrite as a supported add_edge, attach_evidence, or revise_strategy apply contract."
return {
"proposal_id": proposal.get("id"),
"normalization_state": state,
"strict_child_proposals": strict_children,
"blocked_fragments": blocked_fragments,
"blocked_count": len(blocked_fragments),
"strict_child_count": len(strict_children),
"next_normalization_action": next_action,
}
def load_from_input(path: str) -> list[dict[str, Any]]:
return review_packet.load_from_input(path)
def load_from_db(args: argparse.Namespace) -> list[dict[str, Any]]:
return review_packet.load_from_db(args)
def parse_args(argv: list[str]) -> argparse.Namespace:
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("--input-json", help="normalize proposal JSON from a file")
parser.add_argument("--proposal-id", help="load one live proposal from Postgres")
parser.add_argument("--status", default="approved", help="load live proposals by status")
parser.add_argument("--limit", type=int, default=20)
parser.add_argument("--secrets-file", default=ap.DEFAULT_SECRETS_FILE)
parser.add_argument("--container", default=ap.DEFAULT_CONTAINER)
parser.add_argument("--db", default=ap.DEFAULT_DB)
parser.add_argument("--host", default=ap.DEFAULT_HOST)
parser.add_argument("--role", default=ap.DEFAULT_ROLE)
return parser.parse_args(argv)
def main(argv: list[str] | None = None) -> int:
args = parse_args(sys.argv[1:] if argv is None else argv)
proposals = load_from_input(args.input_json) if args.input_json else load_from_db(args)
previews = [normalize_proposal(proposal) for proposal in proposals]
print(json.dumps({"normalization_previews": previews}, indent=2, sort_keys=True))
return 0
if __name__ == "__main__":
raise SystemExit(main())