1 files changed, 221 insertions, 0 deletions
diff --git a/backend/app/services/news_service.py b/backend/app/services/news_service.py
new file mode 100644
index 0000000..0f95096
--- /dev/null
+++ b/backend/app/services/news_service.py
@@ -0,0 +1,221 @@
+"""News feed service for Prism v2 — Finnhub company news + sentiment, FMP fallback."""
+from __future__ import annotations
+
+import hashlib
+import logging
+import os
+from datetime import datetime, timedelta, timezone
+from typing import Any
+from urllib.parse import urlparse
+
+import httpx
+from cachetools import TTLCache, cached
+
+FINNHUB_BASE = "https://finnhub.io/api/v1"
+FMP_BASE = "https://financialmodelingprep.com/api/v3"
+NEWS_CACHE: TTLCache = TTLCache(maxsize=128, ttl=600)
+
+logger = logging.getLogger(__name__)
+
+POSITIVE_WORDS = [
+    "beats", "surges", "rises", "gains", "profit", "record", "upgrade", "buy", "outperform",
+    "growth", "strong", "higher", "rally", "raises", "expands", "upside",
+]
+NEGATIVE_WORDS = [
+    "misses", "falls", "drops", "loss", "cut", "downgrade", "sell", "underperform", "weak",
+    "lower", "decline", "warning", "layoff", "lawsuit", "probe", "cuts",
+]
+
+
+def _finnhub_api_key() -> str:
+    return os.getenv("FINNHUB_API_KEY", "")
+
+
+def _fmp_api_key() -> str:
+    return os.getenv("FMP_API_KEY", "")
+
+
+def _today_window(days: int = 30) -> tuple[str, str]:
+    end = datetime.now(timezone.utc).date()
+    start = end - timedelta(days=days)
+    return start.isoformat(), end.isoformat()
+
+
+def _http_get_json(url: str, params: dict[str, Any] | None = None, timeout: float = 10.0) -> Any:
+    try:
+        with httpx.Client(timeout=timeout) as client:
+            resp = client.get(url, params=params)
+            resp.raise_for_status()
+            return resp.json()
+    except Exception as exc:  # noqa: BLE001
+        logger.warning("news upstream request failed: %s - %s", url, exc)
+        return None
+
+
+def _finnhub_company_news(symbol: str) -> list[dict]:
+    start, end = _today_window(30)
+    data = _http_get_json(
+        f"{FINNHUB_BASE}/company-news",
+        params={"symbol": symbol.upper(), "from": start, "to": end, "token": _finnhub_api_key()},
+    )
+    if not isinstance(data, list):
+        return []
+    return data[:100]
+
+
+def _finnhub_news_sentiment(symbol: str) -> dict | None:
+    data = _http_get_json(
+        f"{FINNHUB_BASE}/news-sentiment",
+        params={"symbol": symbol.upper(), "token": _finnhub_api_key()},
+    )
+    if isinstance(data, dict):
+        return data
+    return None
+
+
+def _fmp_stock_news(symbol: str) -> list[dict]:
+    data = _http_get_json(
+        f"{FMP_BASE}/stock_news",
+        params={"tickers": symbol.upper(), "limit": 100, "apikey": _fmp_api_key()},
+    )
+    if not isinstance(data, list):
+        return []
+    return data[:100]
+
+
+def _classify_sentiment(article: dict) -> str:
+    headline = str(article.get("headline") or article.get("title") or "").lower()
+    summary = str(article.get("summary") or article.get("text") or "").lower()
+    text = headline + " " + summary
+    pos = sum(1 for w in POSITIVE_WORDS if w in text)
+    neg = sum(1 for w in NEGATIVE_WORDS if w in text)
+    if pos > neg:
+        return "bullish"
+    if neg > pos:
+        return "bearish"
+    return "neutral"
+
+
+def _normalize_source(raw: Any) -> str:
+    text = str(raw or "").strip().lower()
+    if text.startswith("www."):
+        text = text[4:]
+    # If it looks like a domain, drop TLD.
+    if "." in text:
+        text = text.split(".")[0]
+    return text or "unknown"
+
+
+def _parse_published_at(raw: Any) -> str | None:
+    if raw is None:
+        return None
+    if isinstance(raw, (int, float)):
+        try:
+            dt = datetime.fromtimestamp(float(raw), tz=timezone.utc)
+            return dt.strftime("%Y-%m-%dT%H:%M:%SZ")
+        except Exception:  # noqa: BLE001
+            return None
+    text = str(raw).strip()
+    if not text:
+        return None
+    # ISO strings
+    norm = text.replace("Z", "+00:00")
+    try:
+        dt = datetime.fromisoformat(norm)
+        if dt.tzinfo is None:
+            dt = dt.replace(tzinfo=timezone.utc)
+        else:
+            dt = dt.astimezone(timezone.utc)
+        return dt.strftime("%Y-%m-%dT%H:%M:%SZ")
+    except Exception:  # noqa: BLE001
+        pass
+    # FMP-style "YYYY-MM-DD HH:MM:SS"
+    for fmt in ("%Y-%m-%d %H:%M:%S", "%Y-%m-%d"):
+        try:
+            dt = datetime.strptime(text[: len(fmt)], fmt).replace(tzinfo=timezone.utc)
+            return dt.strftime("%Y-%m-%dT%H:%M:%SZ")
+        except Exception:  # noqa: BLE001
+            pass
+    return None
+
+
+def _is_safe_url(url: Any) -> bool:
+    if not url:
+        return False
+    parsed = urlparse(str(url))
+    return parsed.scheme in ("http", "https")
+
+
+def _build_item(article: dict) -> dict | None:
+    title = str(article.get("headline") or article.get("title") or "").strip()
+    if not title:
+        title = "Untitled item"
+    summary = str(article.get("summary") or article.get("text") or "").strip()
+    source = _normalize_source(article.get("source") or article.get("site"))
+    url = article.get("url") or article.get("newsURL")
+    if not _is_safe_url(url):
+        return None
+    published_at = _parse_published_at(article.get("datetime") or article.get("publishedDate"))
+    if published_at is None:
+        return None
+    item_id = hashlib.sha1(str(url).encode("utf-8")).hexdigest()[:16]
+    return {
+        "id": item_id,
+        "title": title,
+        "summary": summary,
+        "source": source,
+        "url": str(url),
+        "sentiment": _classify_sentiment(article),
+        "published_at": published_at,
+    }
+
+
+def _build_aggregate(raw: dict | None) -> dict | None:
+    if not raw:
+        return None
+    buzz = raw.get("buzz") or {}
+    score = raw.get("sentiment") or {}
+    weekly = buzz.get("articlesInLastWeek")
+    bull = score.get("bullishPercent")
+    bear = score.get("bearishPercent")
+    neutral = score.get("neutralPercent")
+    if weekly is None and bull is None and bear is None and neutral is None:
+        return None
+    return {
+        "buzz_articles_last_week": int(weekly) if weekly is not None else 0,
+        "bullish_pct": float(bull) if bull is not None else 0.0,
+        "bearish_pct": float(bear) if bear is not None else 0.0,
+        "neutral_pct": float(neutral) if neutral is not None else 0.0,
+    }
+
+
+@cached(NEWS_CACHE, key=lambda symbol: symbol.upper())
+def get_news(symbol: str) -> dict:
+    """Return recent news articles and aggregate sentiment for a ticker."""
+    symbol = symbol.upper()
+    provider = "finnhub"
+
+    raw_articles = _finnhub_company_news(symbol)
+    if not raw_articles:
+        raw_articles = _fmp_stock_news(symbol)
+        provider = "fmp" if raw_articles else None
+
+    items = []
+    for article in raw_articles:
+        if not article:
+            continue
+        item = _build_item(article)
+        if item:
+            items.append(item)
+
+    items.sort(key=lambda i: i["published_at"], reverse=True)
+    items = items[:100]
+
+    aggregate = _build_aggregate(_finnhub_news_sentiment(symbol))
+
+    return {
+        "items": items,
+        "aggregate": aggregate,
+        "provider": provider,
+        "fetched_at": datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ"),
+    }