207 lines
8.3 KiB
Python
207 lines
8.3 KiB
Python
#!/usr/bin/env python3
|
||
# -*- coding: utf-8 -*-
|
||
"""Host data adapter for Webcrumbs shared-ui plugins.
|
||
|
||
This module is intentionally read-only. It converts the canonical
|
||
MOMO/PChome competitor intelligence payload into the small
|
||
``window.StockPlatformSharedUI`` shape consumed by the shared plugin
|
||
runtime.
|
||
"""
|
||
|
||
from __future__ import annotations
|
||
|
||
from typing import Any
|
||
|
||
from services.competitor_intel_repository import (
|
||
fetch_competitor_coverage,
|
||
fetch_competitor_review_queue,
|
||
fetch_top_competitor_risks,
|
||
summarize_review_decision_envelopes,
|
||
)
|
||
|
||
|
||
def _num(value: Any, default: float = 0.0) -> float:
|
||
try:
|
||
return float(value)
|
||
except (TypeError, ValueError):
|
||
return default
|
||
|
||
|
||
def _money(value: Any) -> str:
|
||
return f"NT${_num(value):,.0f}"
|
||
|
||
|
||
def _short_text(value: Any, limit: int = 42) -> str:
|
||
text = str(value or "").strip()
|
||
if len(text) <= limit:
|
||
return text
|
||
return f"{text[:limit - 1]}…"
|
||
|
||
|
||
def _risk_level(gap_pct: float) -> str:
|
||
if gap_pct >= 15:
|
||
return "high"
|
||
if gap_pct >= 8:
|
||
return "medium"
|
||
return "watch"
|
||
|
||
|
||
def _is_direct_price_alert(item: dict) -> bool:
|
||
return (
|
||
str(item.get("match_type") or "") == "exact"
|
||
and str(item.get("price_basis") or "") == "total_price"
|
||
and str(item.get("alert_tier") or "") == "price_alert_exact"
|
||
)
|
||
|
||
|
||
def _empty_payload(reason: str = "no_price_alert_exact") -> dict:
|
||
return {
|
||
"marketSnapshot": [
|
||
{
|
||
"name": "MOMO/PChome exact price alert",
|
||
"price": "not_available",
|
||
"change_pct": "not_available",
|
||
"freshness_status": reason,
|
||
}
|
||
],
|
||
"aiCandidate": {
|
||
"ticker": "-",
|
||
"name": "目前沒有可直接告警的 exact 同款價差",
|
||
"thesis": "資料源已接入，但目前沒有符合 exact / total_price / price_alert_exact 的高風險候選；非同款、單位價或變體候選會進入 AI 例外決策隊列。",
|
||
"confidence_score": "not_available",
|
||
"risk_level": "none",
|
||
"release_status": "blocked",
|
||
"evidence_refs": ["competitor_prices", "price_records", reason],
|
||
"updated_at": "read_only",
|
||
},
|
||
}
|
||
|
||
|
||
def _coverage_metadata(coverage: dict, row_count: int) -> dict:
|
||
return {
|
||
"source": "competitor_intel_repository",
|
||
"matched_count": int(coverage.get("valid_matches") or coverage.get("matched_count") or 0),
|
||
"coverage_rate": _num(coverage.get("match_rate")),
|
||
"identity_coverage_rate": _num(coverage.get("match_rate")),
|
||
"decision_ready_count": int(
|
||
coverage.get("decision_ready_matches")
|
||
or coverage.get("fresh_matches")
|
||
or coverage.get("fresh_match_count")
|
||
or 0
|
||
),
|
||
"decision_ready_rate": _num(coverage.get("decision_ready_rate")),
|
||
"decision_support_count": int(coverage.get("decision_support_count") or 0),
|
||
"decision_support_rate": _num(coverage.get("decision_support_rate")),
|
||
"decision_support_non_exact_count": int(
|
||
coverage.get("decision_support_non_exact_count") or 0
|
||
),
|
||
"catalog_comparable_count": int(coverage.get("catalog_comparable_count") or 0),
|
||
"catalog_comparable_rate": _num(coverage.get("catalog_comparable_rate")),
|
||
"catalog_variant_review_count": int(coverage.get("catalog_variant_review_count") or 0),
|
||
"catalog_unit_review_count": int(coverage.get("catalog_unit_review_count") or 0),
|
||
"catalog_identity_review_count": int(coverage.get("catalog_identity_review_count") or 0),
|
||
"catalog_review_plan": coverage.get("catalog_review_plan") or {},
|
||
"unit_comparable_count": int(coverage.get("unit_comparable_count") or 0),
|
||
"fresh_match_count": int(coverage.get("fresh_matches") or coverage.get("fresh_match_count") or 0),
|
||
"fresh_match_rate": _num(coverage.get("fresh_match_rate")),
|
||
"stale_match_count": int(coverage.get("stale_matches") or coverage.get("stale_match_count") or 0),
|
||
"pending_match_count": int(coverage.get("pending") or coverage.get("pending_match_count") or 0),
|
||
"row_count": row_count,
|
||
"writes_database": False,
|
||
"calls_llm": False,
|
||
"fetches_external": False,
|
||
}
|
||
|
||
|
||
def _review_decision_brief(engine, limit: int) -> dict:
|
||
"""Read the shared review decision brief without blocking host-data preview."""
|
||
try:
|
||
review_queue = fetch_competitor_review_queue(engine, limit=min(max(limit, 1), 8))
|
||
return summarize_review_decision_envelopes(review_queue, limit=min(max(limit, 1), 5))
|
||
except Exception:
|
||
return summarize_review_decision_envelopes([], limit=min(max(limit, 1), 5))
|
||
|
||
|
||
def _attach_review_brief(payload: dict, review_brief: dict) -> dict:
|
||
payload["reviewDecisionBrief"] = review_brief
|
||
metadata = payload.setdefault("metadata", {})
|
||
metadata["review_queue_count"] = len(review_brief.get("items") or [])
|
||
metadata["hitl_count"] = 0
|
||
metadata["primary_human_gate_count"] = 0
|
||
metadata["ai_exception_count"] = int(review_brief.get("auto_execute_blocked_count") or 0)
|
||
metadata["auto_execute_blocked_count"] = int(
|
||
review_brief.get("auto_execute_blocked_count") or 0
|
||
)
|
||
metadata["decision_envelope_source"] = "competitor_intel_repository"
|
||
return payload
|
||
|
||
|
||
def build_webcrumbs_marketplace_host_data(engine=None, limit: int = 5) -> dict:
|
||
"""Build read-only host data for the Webcrumbs diagnostic page.
|
||
|
||
Args:
|
||
engine: Optional SQLAlchemy engine. When omitted, the default
|
||
momo-pro database manager is used.
|
||
limit: Maximum number of price-alert rows exposed to the shared-ui
|
||
preview. Kept small because this powers a diagnostic surface.
|
||
"""
|
||
if engine is None:
|
||
from database.manager import DatabaseManager
|
||
|
||
engine = DatabaseManager().engine
|
||
|
||
limit = max(1, min(int(limit or 5), 8))
|
||
raw_risks = fetch_top_competitor_risks(engine, limit=max(limit * 4, 20)) or []
|
||
risks = [item for item in raw_risks if _is_direct_price_alert(item)][:limit]
|
||
coverage = fetch_competitor_coverage(engine) or {}
|
||
review_brief = _review_decision_brief(engine, limit=limit)
|
||
|
||
if not risks:
|
||
payload = _empty_payload("no_current_exact_risk")
|
||
payload["metadata"] = _coverage_metadata(coverage, row_count=0)
|
||
return _attach_review_brief(payload, review_brief)
|
||
|
||
rows = []
|
||
for item in risks[:limit]:
|
||
gap_pct = _num(item.get("gap_pct"))
|
||
momo_price = _num(item.get("momo_price"))
|
||
pchome_price = _num(item.get("pchome_price"))
|
||
rows.append({
|
||
"name": f"{item.get('sku') or '-'} {_short_text(item.get('name'))}",
|
||
"price": pchome_price,
|
||
"change_pct": gap_pct,
|
||
"freshness_status": item.get("alert_tier") or "price_alert_exact",
|
||
"momo_price": momo_price,
|
||
"pchome_price": pchome_price,
|
||
"match_score": _num(item.get("match_score")),
|
||
"competitor_product_id": item.get("pchome_id") or "",
|
||
"competitor_product_name": _short_text(item.get("pchome_name"), 48),
|
||
})
|
||
|
||
top = risks[0]
|
||
top_gap = _num(top.get("gap_pct"))
|
||
top_score = _num(top.get("match_score"))
|
||
sku = str(top.get("sku") or "-")
|
||
evidence_refs = ["competitor_prices", "price_records", "exact", "total_price", "price_alert_exact"]
|
||
|
||
payload = {
|
||
"marketSnapshot": rows,
|
||
"aiCandidate": {
|
||
"ticker": sku,
|
||
"name": _short_text(top.get("name"), 58),
|
||
"thesis": (
|
||
f"MOMO {_money(top.get('momo_price'))} vs PChome {_money(top.get('pchome_price'))}，"
|
||
f"價差 {top_gap:+.1f}%（MOMO - PChome）。"
|
||
f"此候選已通過 exact / total_price / price_alert_exact 只讀過濾，"
|
||
f"match_score={top_score:.2f}；由 AI 決策信封檢查促銷、庫存與商品頁條件後再採取價格或曝光調整。"
|
||
),
|
||
"confidence_score": round(top_score, 2),
|
||
"risk_level": _risk_level(top_gap),
|
||
"release_status": "review_required",
|
||
"evidence_refs": evidence_refs,
|
||
"updated_at": top.get("crawled_at") or "latest_competitor_prices",
|
||
},
|
||
"metadata": _coverage_metadata(coverage, row_count=len(rows)),
|
||
}
|
||
return _attach_review_brief(payload, review_brief)
|