Files
clientflow_backend/app/blif_flow_v2_projection_service.py

216 lines
10 KiB
Python

"""Persistence scaffolding for the rebuildable BLIF Flow v2 projection.
The factual sources remain authoritative. This module writes only the additive
projection/audit tables introduced by migration 011 and never updates stages or
tasks.
"""
from __future__ import annotations
import hashlib
import json
import re
from datetime import datetime, timezone
from typing import Any, Iterable
from sqlalchemy import text
from app.config import settings
FLOW_VERSION = "blif-flow-v2-shadow-3"
WRITE_DATABASE_ALLOWLIST = frozenset({"clientflow_codex_test"})
def _jsonable(value: Any) -> Any:
if isinstance(value, datetime):
return value.isoformat()
if isinstance(value, dict):
return {str(key): _jsonable(item) for key, item in value.items()}
if isinstance(value, (list, tuple)):
return [_jsonable(item) for item in value]
return value
def _evidence_refs(row: dict[str, Any]) -> list[dict[str, Any]]:
evidence = row.get("evidence") or {}
refs: list[dict[str, Any]] = []
for name in ("latest_relevant_inbound", "latest_relevant_outbound"):
item = evidence.get(name)
if item and item.get("id"):
refs.append({"source": "message", "role": name, "id": item["id"], "at": item.get("at")})
for name, source in (("proforma", "jasmin_proforma"), ("invoice", "jasmin_invoice"),
("payment", "payment"), ("odoo", "odoo"),
("reconciliation", "reconciliation")):
for item in evidence.get(name) or []:
if item.get("id"):
refs.append({
"source": source, "id": item["id"],
"external_id": item.get("external_id"),
"document_number": item.get("document_number"),
"external_type": item.get("external_type"),
"status": item.get("status"),
})
return refs
def _projection_value(row: dict[str, Any], derived_at: datetime) -> dict[str, Any]:
raw = row["raw_v2"]
duplicate = row.get("canonical_process_id") != row.get("opportunity_id")
evidence_refs = _evidence_refs(row)
reason_code = str(raw.get("precedence") or "business_transition").upper()
stable = {
"opportunity_id": row["opportunity_id"],
"material_process_key": row["material_process_key"],
"canonical_opportunity_id": row["canonical_process_id"],
"is_duplicate_representation": duplicate,
"business_state": raw["business_state"],
"business_next_action": raw.get("business_next_action"),
"diagnostic_status": raw.get("diagnostic_status") or row.get("evidence", {}).get("diagnostic_status") or "clear",
"confidence": raw.get("confidence") or "low",
"reason_code": reason_code,
"reason_text": raw.get("reason") or "",
"evidence_refs": evidence_refs,
"flow_version": FLOW_VERSION,
}
fingerprint = hashlib.sha256(
json.dumps(_jsonable(stable), ensure_ascii=False, sort_keys=True, separators=(",", ":")).encode("utf-8")
).hexdigest()
return stable | {"source_fingerprint": fingerprint, "derived_at": derived_at}
def _derive_all() -> list[dict[str, Any]]:
# Reuse the validated shadow evidence adapter without making it authoritative.
from scripts.simulate_blif_flow_v2 import collect
report = collect(
expected_database="clientflow_codex_test",
expected_user="clientflow_codex_test",
# collect() still opens its factual read phase with BEGIN READ ONLY;
# the session default may be read-write in the isolated test database.
require_read_only=False,
)
return list(report["opportunities"])
def rebuild_blif_flow_v2_projection(
*,
mode: str | None = None,
derived_rows: Iterable[dict[str, Any]] | None = None,
allowed_databases: frozenset[str] = WRITE_DATABASE_ALLOWLIST,
target_schema: str = "public",
connection: Any | None = None,
) -> dict[str, Any]:
"""Idempotently rebuild projection rows and state-change transitions.
``off`` is a no-op. ``shadow`` writes only additive projection tables.
Compare/authoritative behavior is deliberately not implemented.
"""
selected_mode = str(mode or settings.blif_flow_v2_mode or "off").strip().lower()
if selected_mode == "off":
return {"mode": "off", "projection_count": 0, "transitions_written": 0, "disabled": True}
if selected_mode != "shadow":
raise RuntimeError(f"BLIF Flow v2 mode {selected_mode!r} is not implemented; only off/shadow are safe")
if not re.fullmatch(r"[a-z_][a-z0-9_]*", target_schema):
raise ValueError("invalid target_schema")
rows = list(derived_rows) if derived_rows is not None else _derive_all()
derived_at = datetime.now(timezone.utc)
values = [_projection_value(row, derived_at) for row in rows]
if len({value["opportunity_id"] for value in values}) != len(values):
raise RuntimeError("Flow v2 projection requires exactly one derived row per opportunity")
owns_connection = connection is None
if connection is None:
from app.db import engine
conn = engine.connect().execution_options(isolation_level="AUTOCOMMIT")
else:
conn = connection
try:
identity = conn.execute(text(
"SELECT current_database(), current_user, current_setting('transaction_read_only')"
)).one()
if identity[0] not in allowed_databases:
raise RuntimeError(f"refusing Flow v2 projection write to database {identity[0]!r}")
if owns_connection:
conn.exec_driver_sql("BEGIN READ WRITE")
try:
if conn.execute(text("SELECT current_setting('transaction_read_only')")).scalar_one() != "off":
raise RuntimeError("Flow v2 projection rebuild requires an explicit READ WRITE transaction")
conn.exec_driver_sql(f'SET LOCAL search_path TO "{target_schema}"')
existing = {
str(row["opportunity_id"]): dict(row)
for row in conn.execute(text("""
SELECT opportunity_id::text, business_state, source_fingerprint
FROM opportunity_flow_state_v2
""")).mappings()
}
transitions_written = 0
for value in values:
previous = existing.get(value["opportunity_id"])
if previous is None or previous["business_state"] != value["business_state"]:
result = conn.execute(text("""
INSERT INTO opportunity_flow_transitions (
opportunity_id, from_state, to_state, reason_code, reason_text,
evidence_refs, flow_version, source_fingerprint
) VALUES (
CAST(:opportunity_id AS UUID), :from_state, :to_state, :reason_code,
:reason_text, CAST(:evidence_refs AS JSONB), :flow_version, :source_fingerprint
)
ON CONFLICT (opportunity_id, from_state, to_state, flow_version, source_fingerprint)
DO NOTHING
"""), {
**value, "from_state": previous["business_state"] if previous else None,
"to_state": value["business_state"],
"evidence_refs": json.dumps(_jsonable(value["evidence_refs"]), ensure_ascii=False),
})
transitions_written += result.rowcount
conn.execute(text("""
INSERT INTO opportunity_flow_state_v2 (
opportunity_id, material_process_key, canonical_opportunity_id,
is_duplicate_representation, business_state, business_next_action,
diagnostic_status, confidence, reason_code, reason_text, evidence_refs,
flow_version, source_fingerprint, derived_at, updated_at
) VALUES (
CAST(:opportunity_id AS UUID), :material_process_key,
CAST(:canonical_opportunity_id AS UUID), :is_duplicate_representation,
:business_state, :business_next_action, :diagnostic_status, :confidence,
:reason_code, :reason_text, CAST(:evidence_refs_json AS JSONB), :flow_version,
:source_fingerprint, :derived_at, now()
)
ON CONFLICT (opportunity_id) DO UPDATE SET
material_process_key=EXCLUDED.material_process_key,
canonical_opportunity_id=EXCLUDED.canonical_opportunity_id,
is_duplicate_representation=EXCLUDED.is_duplicate_representation,
business_state=EXCLUDED.business_state,
business_next_action=EXCLUDED.business_next_action,
diagnostic_status=EXCLUDED.diagnostic_status,
confidence=EXCLUDED.confidence,
reason_code=EXCLUDED.reason_code,
reason_text=EXCLUDED.reason_text,
evidence_refs=EXCLUDED.evidence_refs,
flow_version=EXCLUDED.flow_version,
source_fingerprint=EXCLUDED.source_fingerprint,
derived_at=EXCLUDED.derived_at,
updated_at=CASE
WHEN opportunity_flow_state_v2.source_fingerprint IS DISTINCT FROM EXCLUDED.source_fingerprint
THEN now() ELSE opportunity_flow_state_v2.updated_at END
"""), value | {
"evidence_refs_json": json.dumps(_jsonable(value["evidence_refs"]), ensure_ascii=False)
})
if owns_connection:
conn.exec_driver_sql("COMMIT")
except Exception:
if owns_connection:
conn.exec_driver_sql("ROLLBACK")
raise
finally:
if owns_connection:
conn.close()
return {
"mode": "shadow", "projection_count": len(values),
"canonical_count": sum(not value["is_duplicate_representation"] for value in values),
"duplicate_count": sum(value["is_duplicate_representation"] for value in values),
"transitions_written": transitions_written, "disabled": False,
}