Push the rest

This commit is contained in:
2026-05-11 10:58:46 +02:00
parent adb5c1a439
commit 0031caf16c
94 changed files with 11777 additions and 3474 deletions

View File

View File

@@ -0,0 +1,258 @@
from __future__ import annotations
import logging
from opentelemetry import metrics, trace
from sqlalchemy.orm import sessionmaker, Session
from app.core.audit import append_audit
from app.domain.aw.models import AWSalesForecast, AWRepScore, AWProductDemand, AWAnomalyRun
LOGGER = logging.getLogger(__name__)
tracer = trace.get_tracer("otel-bi.domain.aw")
meter = metrics.get_meter("otel-bi.domain.aw")
_persist_counter = meter.create_counter(
"aw_persist_writes_total",
description="Number of AW PostgreSQL write operations",
)
def _current_span_context() -> tuple[str | None, str | None]:
ctx = trace.get_current_span().get_span_context()
if not ctx.is_valid:
return None, None
return f"{ctx.trace_id:032x}", f"{ctx.span_id:016x}"
def _actor_type(trigger_source: str) -> str:
return "scheduler" if trigger_source.startswith("scheduler") else "api"
# ---------------------------------------------------------------------------
# Persist functions — called after Go service returns data
# ---------------------------------------------------------------------------
def persist_forecast(
factory: sessionmaker[Session],
data: list[dict],
horizon_days: int,
trigger_source: str,
) -> None:
trace_id, span_id = _current_span_context()
try:
with factory() as session:
session.add(AWSalesForecast(
horizon_days=horizon_days,
point_count=len(data),
trigger_source=trigger_source,
trace_id=trace_id,
span_id=span_id,
payload=data,
))
session.commit()
_persist_counter.add(1, {"entity": "sales_forecast"})
except Exception as exc: # noqa: BLE001
LOGGER.warning("Failed to persist AW forecast: %s", exc)
append_audit(
factory,
action="forecast.generated",
actor_type=_actor_type(trigger_source),
actor_id=trigger_source,
domain="aw",
service="otel-bi-backend",
entity_type="sales_forecast",
payload={"horizon_days": horizon_days, "point_count": len(data)},
)
def persist_rep_scores(
factory: sessionmaker[Session],
data: list[dict],
top_n: int,
trigger_source: str,
) -> None:
trace_id, span_id = _current_span_context()
try:
with factory() as session:
session.add(AWRepScore(
rep_count=len(data),
trigger_source=trigger_source,
trace_id=trace_id,
span_id=span_id,
payload=data,
))
session.commit()
_persist_counter.add(1, {"entity": "rep_scores"})
except Exception as exc: # noqa: BLE001
LOGGER.warning("Failed to persist AW rep scores: %s", exc)
append_audit(
factory,
action="scores.generated",
actor_type=_actor_type(trigger_source),
actor_id=trigger_source,
domain="aw",
service="otel-bi-backend",
entity_type="rep_scores",
payload={"rep_count": len(data), "top_n": top_n},
)
def persist_product_demand(
factory: sessionmaker[Session],
data: list[dict],
top_n: int,
trigger_source: str,
) -> None:
trace_id, span_id = _current_span_context()
try:
with factory() as session:
session.add(AWProductDemand(
product_count=len(data),
top_n=top_n,
trigger_source=trigger_source,
trace_id=trace_id,
span_id=span_id,
payload=data,
))
session.commit()
_persist_counter.add(1, {"entity": "product_demand"})
except Exception as exc: # noqa: BLE001
LOGGER.warning("Failed to persist AW product demand: %s", exc)
append_audit(
factory,
action="scores.generated",
actor_type=_actor_type(trigger_source),
actor_id=trigger_source,
domain="aw",
service="otel-bi-backend",
entity_type="product_demand",
payload={"product_count": len(data), "top_n": top_n},
)
def persist_anomaly_run(
factory: sessionmaker[Session],
data: list[dict],
trigger_source: str,
) -> None:
anomaly_count = sum(1 for p in data if p.get("is_anomaly"))
trace_id, span_id = _current_span_context()
try:
with factory() as session:
session.add(AWAnomalyRun(
anomaly_count=anomaly_count,
series_days=365,
window_days=30,
threshold_sigma=2.0,
trigger_source=trigger_source,
trace_id=trace_id,
span_id=span_id,
payload=data,
))
session.commit()
_persist_counter.add(1, {"entity": "anomaly_run"})
except Exception as exc: # noqa: BLE001
LOGGER.warning("Failed to persist AW anomaly run: %s", exc)
append_audit(
factory,
action="anomaly_detection.ran",
actor_type=_actor_type(trigger_source),
actor_id=trigger_source,
domain="aw",
service="otel-bi-backend",
entity_type="anomaly_detection",
payload={"series_days": 365, "window_days": 30, "anomaly_count": anomaly_count},
)
# ---------------------------------------------------------------------------
# Read functions — query PostgreSQL for stored results
# ---------------------------------------------------------------------------
def list_forecasts(factory: sessionmaker[Session], limit: int = 50) -> list[dict]:
with factory() as session:
rows = (
session.query(AWSalesForecast)
.order_by(AWSalesForecast.created_at.desc())
.limit(limit)
.all()
)
return [
{
"id": r.id,
"created_at": r.created_at.isoformat(),
"horizon_days": r.horizon_days,
"point_count": r.point_count,
"trigger_source": r.trigger_source,
"trace_id": r.trace_id,
}
for r in rows
]
def list_rep_scores(factory: sessionmaker[Session], limit: int = 50) -> list[dict]:
with factory() as session:
rows = (
session.query(AWRepScore)
.order_by(AWRepScore.computed_at.desc())
.limit(limit)
.all()
)
return [
{
"id": r.id,
"computed_at": r.computed_at.isoformat(),
"rep_count": r.rep_count,
"trigger_source": r.trigger_source,
"trace_id": r.trace_id,
"payload": r.payload,
}
for r in rows
]
def list_product_demand(factory: sessionmaker[Session], limit: int = 50) -> list[dict]:
with factory() as session:
rows = (
session.query(AWProductDemand)
.order_by(AWProductDemand.computed_at.desc())
.limit(limit)
.all()
)
return [
{
"id": r.id,
"computed_at": r.computed_at.isoformat(),
"product_count": r.product_count,
"top_n": r.top_n,
"trigger_source": r.trigger_source,
"trace_id": r.trace_id,
"payload": r.payload,
}
for r in rows
]
def list_anomaly_runs(factory: sessionmaker[Session], limit: int = 20) -> list[dict]:
with factory() as session:
rows = (
session.query(AWAnomalyRun)
.order_by(AWAnomalyRun.detected_at.desc())
.limit(limit)
.all()
)
return [
{
"id": r.id,
"detected_at": r.detected_at.isoformat(),
"anomaly_count": r.anomaly_count,
"series_days": r.series_days,
"window_days": r.window_days,
"threshold_sigma": r.threshold_sigma,
"trigger_source": r.trigger_source,
"trace_id": r.trace_id,
}
for r in rows
]

View File

@@ -0,0 +1,77 @@
from __future__ import annotations
from datetime import datetime, timezone
from uuid import uuid4
from sqlalchemy import JSON, DateTime, Integer, String
from sqlalchemy.orm import DeclarativeBase, Mapped, mapped_column
def _utcnow() -> datetime:
return datetime.now(timezone.utc)
class AWBase(DeclarativeBase):
pass
class AWSalesForecast(AWBase):
"""Persisted AW sales forecast runs."""
__tablename__ = "aw_sales_forecasts"
id: Mapped[str] = mapped_column(String(36), primary_key=True, default=lambda: str(uuid4()))
created_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), default=_utcnow, index=True)
horizon_days: Mapped[int] = mapped_column(Integer)
point_count: Mapped[int] = mapped_column(Integer)
trigger_source: Mapped[str] = mapped_column(String(64), index=True)
trace_id: Mapped[str | None] = mapped_column(String(32), nullable=True, index=True)
span_id: Mapped[str | None] = mapped_column(String(16), nullable=True)
payload: Mapped[list[dict]] = mapped_column(JSON, default=list)
class AWRepScore(AWBase):
"""Persisted AW sales rep performance scoring runs."""
__tablename__ = "aw_rep_scores"
id: Mapped[str] = mapped_column(String(36), primary_key=True, default=lambda: str(uuid4()))
computed_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), default=_utcnow, index=True)
rep_count: Mapped[int] = mapped_column(Integer)
trigger_source: Mapped[str] = mapped_column(String(64), index=True)
trace_id: Mapped[str | None] = mapped_column(String(32), nullable=True, index=True)
span_id: Mapped[str | None] = mapped_column(String(16), nullable=True)
payload: Mapped[list[dict]] = mapped_column(JSON, default=list)
class AWProductDemand(AWBase):
"""Persisted AW product demand scoring runs."""
__tablename__ = "aw_product_demand"
id: Mapped[str] = mapped_column(String(36), primary_key=True, default=lambda: str(uuid4()))
computed_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), default=_utcnow, index=True)
product_count: Mapped[int] = mapped_column(Integer)
top_n: Mapped[int] = mapped_column(Integer)
trigger_source: Mapped[str] = mapped_column(String(64), index=True)
trace_id: Mapped[str | None] = mapped_column(String(32), nullable=True, index=True)
span_id: Mapped[str | None] = mapped_column(String(16), nullable=True)
payload: Mapped[list[dict]] = mapped_column(JSON, default=list)
class AWAnomalyRun(AWBase):
"""Persisted AW revenue anomaly detection runs."""
__tablename__ = "aw_anomaly_runs"
id: Mapped[str] = mapped_column(String(36), primary_key=True, default=lambda: str(uuid4()))
detected_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), default=_utcnow, index=True)
anomaly_count: Mapped[int] = mapped_column(Integer)
series_days: Mapped[int] = mapped_column(Integer)
window_days: Mapped[int] = mapped_column(Integer)
threshold_sigma: Mapped[float] = mapped_column(default=2.0)
trigger_source: Mapped[str] = mapped_column(String(64), index=True)
trace_id: Mapped[str | None] = mapped_column(String(32), nullable=True, index=True)
span_id: Mapped[str | None] = mapped_column(String(16), nullable=True)
# Full annotated series (date, revenue, rolling_mean, lower_band, upper_band, is_anomaly, z_score)
payload: Mapped[list[dict]] = mapped_column(JSON, default=list)

View File

@@ -0,0 +1,131 @@
from __future__ import annotations
# ---------------------------------------------------------------------------
# AdventureWorksDW2022 — read-only MSSQL queries
# Each list contains fallback variants tried in order.
# ---------------------------------------------------------------------------
# Daily sales combining FactInternetSales + FactResellerSales
AW_DAILY_SALES: list[str] = [
"""
SELECT
CAST(d.FullDateAlternateKey AS date) AS sale_date,
SUM(f.SalesAmount) AS revenue,
SUM(f.TotalProductCost) AS cost,
SUM(f.OrderQuantity) AS quantity,
COUNT_BIG(*) AS orders
FROM dbo.FactInternetSales AS f
INNER JOIN dbo.DimDate AS d ON d.DateKey = f.OrderDateKey
GROUP BY CAST(d.FullDateAlternateKey AS date)
UNION ALL
SELECT
CAST(d.FullDateAlternateKey AS date) AS sale_date,
SUM(r.SalesAmount) AS revenue,
SUM(r.TotalProductCost) AS cost,
SUM(r.OrderQuantity) AS quantity,
COUNT_BIG(*) AS orders
FROM dbo.FactResellerSales AS r
INNER JOIN dbo.DimDate AS d ON d.DateKey = r.OrderDateKey
GROUP BY CAST(d.FullDateAlternateKey AS date)
ORDER BY sale_date;
""",
# Fallback: internet sales only using OrderDate column directly
"""
SELECT
CAST(OrderDate AS date) AS sale_date,
SUM(SalesAmount) AS revenue,
SUM(TotalProductCost) AS cost,
SUM(OrderQuantity) AS quantity,
COUNT_BIG(*) AS orders
FROM dbo.FactInternetSales
GROUP BY CAST(OrderDate AS date)
ORDER BY sale_date;
""",
]
# Sales rep performance — reseller sales attributed to employees
AW_REP_PERFORMANCE: list[str] = [
"""
SELECT
e.EmployeeKey AS employee_key,
e.FirstName + ' ' + e.LastName AS rep_name,
COALESCE(e.Title, 'Sales Rep') AS rep_title,
COALESCE(st.SalesTerritoryRegion, 'Unknown') AS territory,
SUM(r.SalesAmount) AS revenue,
SUM(r.TotalProductCost) AS cost,
COUNT_BIG(*) AS orders,
AVG(r.SalesAmount) AS avg_deal_size
FROM dbo.FactResellerSales AS r
INNER JOIN dbo.DimEmployee AS e
ON e.EmployeeKey = r.EmployeeKey
INNER JOIN dbo.DimSalesTerritory AS st
ON st.SalesTerritoryKey = r.SalesTerritoryKey
WHERE e.SalesPersonFlag = 1
GROUP BY
e.EmployeeKey,
e.FirstName, e.LastName,
e.Title,
st.SalesTerritoryRegion
ORDER BY revenue DESC;
""",
# Fallback without SalesPersonFlag filter
"""
SELECT
e.EmployeeKey AS employee_key,
e.FirstName + ' ' + e.LastName AS rep_name,
COALESCE(e.Title, 'Employee') AS rep_title,
'Unknown' AS territory,
SUM(r.SalesAmount) AS revenue,
SUM(r.TotalProductCost) AS cost,
COUNT_BIG(*) AS orders,
AVG(r.SalesAmount) AS avg_deal_size
FROM dbo.FactResellerSales AS r
INNER JOIN dbo.DimEmployee AS e ON e.EmployeeKey = r.EmployeeKey
GROUP BY e.EmployeeKey, e.FirstName, e.LastName, e.Title
ORDER BY revenue DESC;
""",
]
# Product demand — internet sales with full category hierarchy
AW_PRODUCT_DEMAND: list[str] = [
"""
SELECT
p.ProductAlternateKey AS product_id,
p.EnglishProductName AS product_name,
COALESCE(pc.EnglishProductCategoryName, 'Unknown') AS category,
SUM(f.SalesAmount) AS revenue,
SUM(f.TotalProductCost) AS cost,
SUM(f.OrderQuantity) AS quantity,
COUNT_BIG(*) AS orders
FROM dbo.FactInternetSales AS f
INNER JOIN dbo.DimProduct AS p
ON p.ProductKey = f.ProductKey
LEFT JOIN dbo.DimProductSubcategory AS sc
ON sc.ProductSubcategoryKey = p.ProductSubcategoryKey
LEFT JOIN dbo.DimProductCategory AS pc
ON pc.ProductCategoryKey = sc.ProductCategoryKey
GROUP BY
p.ProductAlternateKey,
p.EnglishProductName,
pc.EnglishProductCategoryName
ORDER BY revenue DESC;
""",
# Fallback: no category join
"""
SELECT
CAST(f.ProductKey AS nvarchar(50)) AS product_id,
COALESCE(p.EnglishProductName, CAST(f.ProductKey AS nvarchar(50))) AS product_name,
'Unknown' AS category,
SUM(f.SalesAmount) AS revenue,
SUM(f.TotalProductCost) AS cost,
SUM(f.OrderQuantity) AS quantity,
COUNT_BIG(*) AS orders
FROM dbo.FactInternetSales AS f
LEFT JOIN dbo.DimProduct AS p ON p.ProductKey = f.ProductKey
GROUP BY f.ProductKey, p.EnglishProductName
ORDER BY revenue DESC;
""",
]