first ruggedy draft

This commit is contained in:
2026-06-28 13:48:15 +02:00
parent c916f64fcb
commit 1053f6a2a3
78 changed files with 20905 additions and 2 deletions

View File

@@ -0,0 +1,644 @@
from __future__ import annotations
from datetime import UTC, date, datetime
from decimal import Decimal
from fastapi import APIRouter, Depends, HTTPException
from mousehold_domain import (
EventObservationRelation,
FinancialEventStatus,
FinancialEventType,
ImportConnectionStatus,
ObservationStatus,
)
from mousehold_matching import compute_normalized_transaction_fingerprint
from sqlalchemy import select
from sqlalchemy.orm import Session, selectinload
from .. import models
from ..auth import get_optional_local_session, require_household_access
from ..db import get_session
from ..schemas import (
EventRead,
ImportRunDetailRead,
ImportRunRead,
ManualObservationCreate,
ObservationAcceptRequest,
ObservationBatchRequest,
ObservationBatchResponse,
ObservationIgnoreRequest,
ObservationImportCreate,
ObservationLinkRequest,
ObservationReviewItem,
ObservationUpdateRequest,
RawObservationRead,
SyncObservationsRequest,
SyncObservationsResponse,
)
from ..services.ingestion import create_observation_with_candidate
from .events import _attach_event_tags
router = APIRouter(tags=["observations"])
def _get_observation(session: Session, observation_id: str) -> models.RawObservation:
observation = session.get(models.RawObservation, observation_id)
if not observation:
raise HTTPException(status_code=404, detail="Observation not found")
return observation
def _events_for_observation(
session: Session, observation_id: str
) -> list[models.FinancialEvent]:
events = list(
session.scalars(
select(models.FinancialEvent)
.join(
models.EventObservationLink,
models.EventObservationLink.event_id == models.FinancialEvent.id,
)
.where(models.EventObservationLink.observation_id == observation_id)
.options(selectinload(models.FinancialEvent.split_lines))
.order_by(models.FinancialEvent.created_at.desc())
).all()
)
_attach_event_tags(session, events)
return events
def _review_item(session: Session, observation: models.RawObservation) -> ObservationReviewItem:
events = _events_for_observation(session, observation.id)
if observation.status == ObservationStatus.DUPLICATE_SOURCE:
suggested_action = "duplicate"
elif observation.status == ObservationStatus.IGNORED:
suggested_action = "ignored"
elif events:
suggested_action = "review_event"
else:
suggested_action = "create_event"
action_required = observation.status not in {
ObservationStatus.IGNORED,
ObservationStatus.DUPLICATE_SOURCE,
}
return ObservationReviewItem(
observation=RawObservationRead.model_validate(observation),
events=[EventRead.model_validate(event) for event in events],
action_required=action_required,
suggested_action=suggested_action,
)
def _date_from_parsed(value: object, field: str) -> date | None:
if not value:
return None
try:
return date.fromisoformat(str(value))
except ValueError as exc:
raise HTTPException(status_code=400, detail=f"Observation has invalid {field}") from exc
def _create_event_from_observation(
session: Session,
observation: models.RawObservation,
event_status: FinancialEventStatus | None,
) -> models.FinancialEvent:
parsed = observation.parsed_json or {}
event_date = _date_from_parsed(parsed.get("event_date"), "event_date")
if not event_date:
raise HTTPException(status_code=400, detail="Observation does not contain an event date")
amount = parsed.get("amount")
if amount is None:
raise HTTPException(status_code=400, detail="Observation does not contain an amount")
event_type = FinancialEventType(str(parsed.get("event_type") or FinancialEventType.EXPENSE))
status = event_status or FinancialEventStatus(str(parsed.get("status") or FinancialEventStatus.CANDIDATE))
fingerprint = compute_normalized_transaction_fingerprint(
amount=Decimal(str(amount)),
currency=str(parsed.get("currency") or "EUR"),
event_date=event_date,
merchant=parsed.get("merchant"),
counterparty=parsed.get("counterparty"),
external_id=observation.provider_transaction_id,
)
event = models.FinancialEvent(
household_id=observation.household_id,
source_account_id=observation.source_account_id,
payment_account_id=observation.source_account_id,
payer_user_id=observation.owner_user_id,
event_type=event_type,
status=status,
event_date=event_date,
booking_date=_date_from_parsed(parsed.get("booking_date"), "booking_date"),
due_date=_date_from_parsed(parsed.get("due_date"), "due_date"),
amount=Decimal(str(amount)),
currency=str(parsed.get("currency") or "EUR"),
merchant=parsed.get("merchant"),
counterparty=parsed.get("counterparty"),
description=parsed.get("description"),
created_by_user_id=observation.owner_user_id,
fingerprint=fingerprint,
)
session.add(event)
session.flush()
session.add(
models.EventObservationLink(
event_id=event.id,
observation_id=observation.id,
relation=EventObservationRelation.IMPORTED_FROM,
confidence=Decimal("1.00"),
)
)
session.flush()
return event
def _require_account(session: Session, household_id: str, account_id: str | None) -> None:
if not account_id:
return
account = session.get(models.Account, account_id)
if not account or account.household_id != household_id:
raise HTTPException(status_code=404, detail="Account not found")
def _require_user(session: Session, household_id: str, user_id: str | None) -> None:
if not user_id:
return
user = session.get(models.User, user_id)
if not user or user.household_id != household_id:
raise HTTPException(status_code=404, detail="User not found")
def _parsed_update_payload(payload: ObservationUpdateRequest) -> dict[str, object]:
parsed_fields = {
"amount",
"currency",
"event_date",
"booking_date",
"due_date",
"merchant",
"counterparty",
"description",
"event_type",
"status",
}
dumped = payload.model_dump(exclude_unset=True, mode="json")
return {key: value for key, value in dumped.items() if key in parsed_fields}
def _sync_event_from_observation(
event: models.FinancialEvent,
observation: models.RawObservation,
parsed_update: dict[str, object],
) -> None:
parsed = observation.parsed_json or {}
if "amount" in parsed_update:
event.amount = Decimal(str(parsed["amount"]))
if "currency" in parsed_update:
event.currency = str(parsed["currency"] or "EUR").upper()
if "event_date" in parsed_update:
parsed_date = _date_from_parsed(parsed.get("event_date"), "event_date")
if parsed_date:
event.event_date = parsed_date
if "booking_date" in parsed_update:
event.booking_date = _date_from_parsed(parsed.get("booking_date"), "booking_date")
if "due_date" in parsed_update:
event.due_date = _date_from_parsed(parsed.get("due_date"), "due_date")
if "merchant" in parsed_update:
event.merchant = parsed.get("merchant")
if "counterparty" in parsed_update:
event.counterparty = parsed.get("counterparty")
if "description" in parsed_update:
event.description = parsed.get("description")
if "event_type" in parsed_update and parsed.get("event_type"):
event.event_type = FinancialEventType(str(parsed["event_type"]))
if "status" in parsed_update and parsed.get("status"):
event.status = FinancialEventStatus(str(parsed["status"]))
event.fingerprint = compute_normalized_transaction_fingerprint(
amount=event.amount,
currency=event.currency,
event_date=event.event_date,
merchant=event.merchant,
counterparty=event.counterparty,
external_id=observation.provider_transaction_id,
)
def _apply_observation_update(
session: Session,
observation: models.RawObservation,
payload: ObservationUpdateRequest,
) -> None:
if "source_account_id" in payload.model_fields_set:
_require_account(session, observation.household_id, payload.source_account_id)
observation.source_account_id = payload.source_account_id
if "owner_user_id" in payload.model_fields_set:
_require_user(session, observation.household_id, payload.owner_user_id)
observation.owner_user_id = payload.owner_user_id
if "raw_text" in payload.model_fields_set:
observation.raw_text = payload.raw_text
parsed_update = _parsed_update_payload(payload)
if parsed_update:
parsed = dict(observation.parsed_json or {})
parsed.update(parsed_update)
if "currency" in parsed and parsed["currency"]:
parsed["currency"] = str(parsed["currency"]).upper()
observation.parsed_json = parsed
session.add(observation)
session.flush()
if payload.sync_linked_events:
for event in _events_for_observation(session, observation.id):
_sync_event_from_observation(event, observation, parsed_update)
if "source_account_id" in payload.model_fields_set:
event.source_account_id = observation.source_account_id
event.payment_account_id = observation.source_account_id
if "owner_user_id" in payload.model_fields_set:
event.payer_user_id = observation.owner_user_id
session.add(event)
session.flush()
def _accept_observation(
session: Session,
observation: models.RawObservation,
event_status: FinancialEventStatus | None,
create_event_if_missing: bool = True,
) -> None:
events = _events_for_observation(session, observation.id)
if not events and create_event_if_missing:
events = [_create_event_from_observation(session, observation, event_status)]
for event in events:
if event_status:
event.status = event_status
session.add(event)
if events and observation.status not in {
ObservationStatus.DUPLICATE_SOURCE,
ObservationStatus.IGNORED,
}:
observation.status = ObservationStatus.CANDIDATE_CREATED
session.add(observation)
session.flush()
def _ignore_observation(
session: Session,
observation: models.RawObservation,
ignore_linked_events: bool,
actor_user_id: str | None,
reason: str | None,
) -> None:
observation.status = ObservationStatus.IGNORED
session.add(observation)
if ignore_linked_events:
for event in _events_for_observation(session, observation.id):
event.status = FinancialEventStatus.IGNORED
session.add(event)
session.add(
models.AuditLog(
household_id=observation.household_id,
actor_user_id=actor_user_id,
action="observation.ignore",
entity_type="raw_observation",
entity_id=observation.id,
details_json={"reason": reason},
)
)
session.flush()
@router.post("/observations/manual", response_model=RawObservationRead)
def create_manual_observation(
payload: ManualObservationCreate,
session: Session = Depends(get_session),
local_session: models.LocalSession | None = Depends(get_optional_local_session),
) -> models.RawObservation:
require_household_access(payload.household_id, local_session)
observation, _event = create_observation_with_candidate(session, payload)
session.commit()
session.refresh(observation)
return observation
@router.post("/observations/import", response_model=RawObservationRead)
def import_observation(
payload: ObservationImportCreate,
session: Session = Depends(get_session),
local_session: models.LocalSession | None = Depends(get_optional_local_session),
) -> models.RawObservation:
require_household_access(payload.household_id, local_session)
observation, _event = create_observation_with_candidate(session, payload)
session.commit()
session.refresh(observation)
return observation
@router.post("/observations/sync", response_model=SyncObservationsResponse)
def sync_observations(
payload: SyncObservationsRequest,
session: Session = Depends(get_session),
local_session: models.LocalSession | None = Depends(get_optional_local_session),
) -> SyncObservationsResponse:
imported_ids: list[str] = []
event_ids: list[str] = []
duplicate_ids: list[str] = []
connection = session.get(models.ImportConnection, payload.connection_id)
household_id = (
payload.observations[0].household_id
if payload.observations
else connection.household_id if connection else None
)
if not household_id:
raise HTTPException(status_code=400, detail="No household available for import")
require_household_access(household_id, local_session)
run = models.ImportRun(
household_id=household_id,
connection_id=connection.id if connection else None,
actor_user_id=local_session.user_id if local_session else None,
source_type=str(payload.observations[0].source_type) if payload.observations else None,
status="running",
summary_json={"requested_count": len(payload.observations)},
)
session.add(run)
session.flush()
for observation_payload in payload.observations:
if observation_payload.household_id != household_id:
raise HTTPException(status_code=400, detail="Import payload mixes households")
observation, event = create_observation_with_candidate(session, observation_payload)
imported_ids.append(observation.id)
if observation.status == ObservationStatus.DUPLICATE_SOURCE:
duplicate_ids.append(observation.id)
if event:
event_ids.append(event.id)
if connection:
connection.status = ImportConnectionStatus.ACTIVE
connection.last_sync_at = datetime.now(UTC)
connection.last_error = None
config = dict(connection.connector_config_json or {})
config["last_imported_count"] = len(imported_ids) - len(duplicate_ids)
config["last_duplicate_count"] = len(duplicate_ids)
config["last_created_event_count"] = len(event_ids)
config["last_import_run_id"] = run.id
connection.connector_config_json = config
session.add(connection)
run.status = "done"
run.imported_count = len(imported_ids) - len(duplicate_ids)
run.duplicate_count = len(duplicate_ids)
run.created_event_count = len(event_ids)
run.finished_at = datetime.now(UTC)
run.summary_json = {
"requested_count": len(payload.observations),
"imported_observation_ids": imported_ids,
"created_event_ids": event_ids,
"duplicate_observation_ids": duplicate_ids,
}
session.add(run)
session.commit()
session.refresh(run)
return SyncObservationsResponse(
imported_observation_ids=imported_ids,
created_event_ids=event_ids,
duplicate_observation_ids=duplicate_ids,
connection_id=connection.id if connection else None,
imported_count=len(imported_ids) - len(duplicate_ids),
duplicate_count=len(duplicate_ids),
created_event_count=len(event_ids),
run=ImportRunRead.model_validate(run),
)
@router.get("/observations/inbox", response_model=list[RawObservationRead])
def get_observation_inbox(
household_id: str,
status: ObservationStatus | None = None,
session: Session = Depends(get_session),
local_session: models.LocalSession | None = Depends(get_optional_local_session),
) -> list[models.RawObservation]:
require_household_access(household_id, local_session)
query = (
select(models.RawObservation)
.where(models.RawObservation.household_id == household_id)
.order_by(models.RawObservation.received_at.desc())
)
if status:
query = query.where(models.RawObservation.status == status)
return list(session.scalars(query).all())
@router.get("/observations/review", response_model=list[ObservationReviewItem])
def get_observation_review_items(
household_id: str,
status: ObservationStatus | None = None,
limit: int = 100,
session: Session = Depends(get_session),
local_session: models.LocalSession | None = Depends(get_optional_local_session),
) -> list[ObservationReviewItem]:
require_household_access(household_id, local_session)
query = (
select(models.RawObservation)
.where(models.RawObservation.household_id == household_id)
.order_by(models.RawObservation.received_at.desc())
.limit(limit)
)
if status:
query = query.where(models.RawObservation.status == status)
return [_review_item(session, observation) for observation in session.scalars(query).all()]
@router.patch("/observations/{observation_id}", response_model=ObservationReviewItem)
def update_observation(
observation_id: str,
payload: ObservationUpdateRequest,
session: Session = Depends(get_session),
local_session: models.LocalSession | None = Depends(get_optional_local_session),
) -> ObservationReviewItem:
observation = _get_observation(session, observation_id)
require_household_access(observation.household_id, local_session)
_apply_observation_update(session, observation, payload)
session.commit()
session.refresh(observation)
return _review_item(session, observation)
@router.post("/observations/{observation_id}/accept", response_model=ObservationReviewItem)
def accept_observation(
observation_id: str,
payload: ObservationAcceptRequest,
session: Session = Depends(get_session),
local_session: models.LocalSession | None = Depends(get_optional_local_session),
) -> ObservationReviewItem:
observation = _get_observation(session, observation_id)
require_household_access(observation.household_id, local_session)
_accept_observation(
session,
observation,
event_status=payload.event_status,
create_event_if_missing=payload.create_event_if_missing,
)
session.commit()
session.refresh(observation)
return _review_item(session, observation)
@router.post("/observations/{observation_id}/ignore", response_model=ObservationReviewItem)
def ignore_observation(
observation_id: str,
payload: ObservationIgnoreRequest,
session: Session = Depends(get_session),
local_session: models.LocalSession | None = Depends(get_optional_local_session),
) -> ObservationReviewItem:
observation = _get_observation(session, observation_id)
require_household_access(observation.household_id, local_session)
_ignore_observation(
session,
observation,
ignore_linked_events=payload.ignore_linked_events,
actor_user_id=local_session.user_id if local_session else None,
reason=payload.reason,
)
session.commit()
session.refresh(observation)
return _review_item(session, observation)
@router.post("/observations/{observation_id}/link", response_model=ObservationReviewItem)
def link_observation(
observation_id: str,
payload: ObservationLinkRequest,
session: Session = Depends(get_session),
local_session: models.LocalSession | None = Depends(get_optional_local_session),
) -> ObservationReviewItem:
observation = _get_observation(session, observation_id)
require_household_access(observation.household_id, local_session)
event = session.get(models.FinancialEvent, payload.event_id)
if not event or event.household_id != observation.household_id:
raise HTTPException(status_code=404, detail="Event not found")
exists = session.scalars(
select(models.EventObservationLink)
.where(models.EventObservationLink.event_id == event.id)
.where(models.EventObservationLink.observation_id == observation.id)
.where(models.EventObservationLink.relation == payload.relation)
.limit(1)
).first()
if not exists:
session.add(
models.EventObservationLink(
event_id=event.id,
observation_id=observation.id,
relation=payload.relation,
confidence=Decimal("1.00"),
)
)
if observation.status == ObservationStatus.NEW:
observation.status = ObservationStatus.PARSED
session.add(observation)
session.commit()
session.refresh(observation)
return _review_item(session, observation)
@router.post("/observations/batch", response_model=ObservationBatchResponse)
def batch_review_observations(
payload: ObservationBatchRequest,
session: Session = Depends(get_session),
local_session: models.LocalSession | None = Depends(get_optional_local_session),
) -> ObservationBatchResponse:
if payload.action not in {"accept", "confirm", "ignore"}:
raise HTTPException(status_code=400, detail="Unsupported batch action")
if not payload.observation_ids:
return ObservationBatchResponse(updated_count=0, items=[])
items: list[ObservationReviewItem] = []
household_id: str | None = None
for observation_id in payload.observation_ids:
observation = _get_observation(session, observation_id)
if household_id is None:
household_id = observation.household_id
require_household_access(household_id, local_session)
elif observation.household_id != household_id:
raise HTTPException(status_code=400, detail="Batch payload mixes households")
if payload.action == "ignore":
_ignore_observation(
session,
observation,
ignore_linked_events=payload.ignore_linked_events,
actor_user_id=local_session.user_id if local_session else None,
reason="Batch review",
)
else:
_accept_observation(
session,
observation,
event_status=FinancialEventStatus.CONFIRMED
if payload.action == "confirm"
else payload.event_status,
create_event_if_missing=True,
)
items.append(_review_item(session, observation))
session.commit()
return ObservationBatchResponse(updated_count=len(items), items=items)
@router.get("/import-runs", response_model=list[ImportRunRead])
def list_import_runs(
household_id: str,
connection_id: str | None = None,
limit: int = 25,
session: Session = Depends(get_session),
local_session: models.LocalSession | None = Depends(get_optional_local_session),
) -> list[models.ImportRun]:
require_household_access(household_id, local_session)
query = (
select(models.ImportRun)
.where(models.ImportRun.household_id == household_id)
.order_by(models.ImportRun.started_at.desc())
.limit(limit)
)
if connection_id:
query = query.where(models.ImportRun.connection_id == connection_id)
return list(session.scalars(query).all())
@router.get("/import-runs/{run_id}", response_model=ImportRunDetailRead)
def get_import_run_detail(
run_id: str,
session: Session = Depends(get_session),
local_session: models.LocalSession | None = Depends(get_optional_local_session),
) -> ImportRunDetailRead:
run = session.get(models.ImportRun, run_id)
if not run:
raise HTTPException(status_code=404, detail="Import run not found")
require_household_access(run.household_id, local_session)
summary = run.summary_json or {}
observation_ids = list(
dict.fromkeys(
[
*summary.get("imported_observation_ids", []),
*summary.get("duplicate_observation_ids", []),
]
)
)
event_ids = summary.get("created_event_ids", [])
observations = (
list(
session.scalars(
select(models.RawObservation).where(models.RawObservation.id.in_(observation_ids))
).all()
)
if observation_ids
else []
)
events = (
list(
session.scalars(
select(models.FinancialEvent)
.where(models.FinancialEvent.id.in_(event_ids))
.options(selectinload(models.FinancialEvent.split_lines))
).all()
)
if event_ids
else []
)
_attach_event_tags(session, events)
return ImportRunDetailRead(
run=ImportRunRead.model_validate(run),
observations=[RawObservationRead.model_validate(observation) for observation in observations],
events=[EventRead.model_validate(event) for event in events],
)