copper-mind / app /data_manager.py
ifieryarrows's picture
Sync from GitHub (tests passed)
31a340a verified
"""
Data Manager: News and price data ingestion.
Handles:
- NewsAPI fetching (if API key provided)
- RSS feed fallback (Google News)
- Fuzzy deduplication for RSS noise
- Multi-symbol yfinance price ingestion
- Language filtering for FinBERT compatibility
Usage:
python -m app.data_manager --fetch
python -m app.data_manager --fetch --news-only
python -m app.data_manager --fetch --prices-only
"""
import argparse
import logging
from datetime import datetime, timedelta, timezone
from typing import Optional
import requests
import yfinance as yf
from rapidfuzz import fuzz
from langdetect import detect, LangDetectException
from sqlalchemy.dialects.sqlite import insert as sqlite_insert
from sqlalchemy.dialects.postgresql import insert as pg_insert
from sqlalchemy.orm import Session
from app.db import SessionLocal, init_db, get_db_type
from app.models import NewsArticle, PriceBar
from app.settings import get_settings
from app.rss_ingest import fetch_google_news
from app.utils import (
clean_text,
canonical_title,
normalize_url,
generate_dedup_key,
truncate_text,
)
from app.lock import pipeline_lock
logging.basicConfig(
level=logging.INFO,
format="%(asctime)s - %(levelname)s - %(message)s"
)
logger = logging.getLogger(__name__)
def get_upsert_stmt(table, values: dict, index_elements: list, update_set: dict = None):
"""Create database-agnostic upsert statement."""
db_type = get_db_type()
if db_type == "postgresql":
stmt = pg_insert(table).values(**values)
if update_set:
stmt = stmt.on_conflict_do_update(index_elements=index_elements, set_=update_set)
else:
stmt = stmt.on_conflict_do_nothing(index_elements=index_elements)
else:
# SQLite
stmt = sqlite_insert(table).values(**values)
if update_set:
stmt = stmt.on_conflict_do_update(index_elements=index_elements, set_=update_set)
else:
stmt = stmt.on_conflict_do_nothing(index_elements=index_elements)
return stmt
# =============================================================================
# NewsAPI Fetching
# =============================================================================
def fetch_newsapi_articles(
api_key: str,
query: str,
language: str = "en",
lookback_days: int = 30,
page_size: int = 100
) -> list[dict]:
"""
Fetch articles from NewsAPI.
Note: Free plan limits to ~1 month of history.
"""
logger.info(f"Fetching from NewsAPI: query='{query}', language={language}")
# Calculate date range
to_date = datetime.now(timezone.utc)
from_date = to_date - timedelta(days=min(lookback_days, 30)) # API limit
url = "https://newsapi.org/v2/everything"
params = {
"apiKey": api_key,
"q": query,
"language": language,
"from": from_date.strftime("%Y-%m-%d"),
"to": to_date.strftime("%Y-%m-%d"),
"sortBy": "publishedAt",
"pageSize": page_size,
}
try:
response = requests.get(url, params=params, timeout=30)
response.raise_for_status()
data = response.json()
if data.get("status") != "ok":
logger.error(f"NewsAPI error: {data.get('message', 'Unknown error')}")
return []
articles = []
for item in data.get("articles", []):
try:
published_str = item.get("publishedAt", "")
published_at = datetime.fromisoformat(published_str.replace("Z", "+00:00")) if published_str else datetime.now(timezone.utc)
articles.append({
"title": item.get("title", ""),
"description": item.get("description", ""),
"content": item.get("content", ""),
"url": item.get("url", ""),
"source": item.get("source", {}).get("name", ""),
"author": item.get("author", ""),
"published_at": published_at,
})
except Exception as e:
logger.debug(f"Error parsing NewsAPI article: {e}")
continue
logger.info(f"Fetched {len(articles)} articles from NewsAPI")
return articles
except requests.RequestException as e:
logger.error(f"NewsAPI request failed: {e}")
return []
# =============================================================================
# Language Detection
# =============================================================================
def detect_language(text: str) -> Optional[str]:
"""Detect language of text. Returns None if detection fails."""
if not text or len(text) < 20:
return None
try:
return detect(text)
except LangDetectException:
return None
def filter_by_language(
articles: list[dict],
target_language: str = "en"
) -> tuple[list[dict], int]:
"""
Filter articles by language.
Returns:
Tuple of (filtered_articles, num_filtered_out)
"""
filtered = []
filtered_out = 0
for article in articles:
# Try to detect from title + description
text = f"{article.get('title', '')} {article.get('description', '')}"
lang = detect_language(text)
if lang is None or lang == target_language:
filtered.append(article)
else:
filtered_out += 1
logger.debug(f"Filtered out ({lang}): {article.get('title', '')[:50]}")
if filtered_out > 0:
logger.info(f"Language filter: kept {len(filtered)}, filtered out {filtered_out}")
return filtered, filtered_out
# =============================================================================
# Fuzzy Deduplication
# =============================================================================
def get_recent_titles(
session: Session,
window_hours: int = 48
) -> list[str]:
"""Get canonical titles from recent articles for fuzzy dedup."""
cutoff = datetime.now(timezone.utc) - timedelta(hours=window_hours)
articles = session.query(NewsArticle.canonical_title).filter(
NewsArticle.published_at >= cutoff,
NewsArticle.canonical_title.isnot(None)
).all()
return [a[0] for a in articles if a[0]]
def is_fuzzy_duplicate(
title: str,
existing_titles: list[str],
threshold: int = 85
) -> bool:
"""
Check if title is too similar to existing titles.
Uses token_set_ratio for robust matching.
"""
if not title or not existing_titles:
return False
canon = canonical_title(title)
for existing in existing_titles:
similarity = fuzz.token_set_ratio(canon, existing)
if similarity >= threshold:
logger.debug(f"Fuzzy duplicate ({similarity}%): '{title[:50]}...'")
return True
return False
# =============================================================================
# News Ingestion
# =============================================================================
def ingest_news(session: Session) -> dict:
"""
Ingest news from all configured sources.
Returns:
Dict with stats: imported, duplicates, language_filtered, fuzzy_filtered
"""
settings = get_settings()
# Strategic queries based on S&P Global 2026 copper market report
# Each pipeline run focuses on different strategic topics for diversity
STRATEGIC_QUERIES = [
# Supply Crisis / Deficit Focus
"copper supply deficit 2026",
"copper shortage AI data center",
"copper inventory LME warehouse",
# Key Players (Majors & Producers)
"Freeport-McMoRan copper outlook",
"BHP copper production news",
"Rio Tinto copper investment",
"Southern Copper SCCO forecast",
# China & Emerging Markets
"Zijin Mining copper investment",
"China copper demand stimulus",
"copper demand EV battery",
# M&A & Strategic Moves
"copper mining acquisition merger",
"Ivanhoe Mines copper grade",
"Lundin Mining copper deal",
# Price & Macro Analysis
"copper price forecast Goldman Sachs",
"copper futures CME analysis",
"grade decline copper mining",
]
logger.info(f"🕵️ Strategic News Agent: Investigating {len(STRATEGIC_QUERIES)} topics...")
stats = {
"imported": 0,
"duplicates": 0,
"language_filtered": 0,
"fuzzy_filtered": 0,
"source": "unknown",
"queries_used": len(STRATEGIC_QUERIES),
}
# Collect articles from ALL strategic queries
all_articles = []
seen_urls = set() # Track URLs to avoid duplicates across queries
for i, strategic_query in enumerate(STRATEGIC_QUERIES, 1):
logger.info(f" [{i}/{len(STRATEGIC_QUERIES)}] Searching: '{strategic_query}'")
query_articles = []
# Try NewsAPI first if key is available
if settings.newsapi_key:
articles = fetch_newsapi_articles(
api_key=settings.newsapi_key,
query=strategic_query,
language=settings.news_language,
lookback_days=settings.lookback_days,
)
if articles:
query_articles.extend(articles)
# RSS fallback/supplement
if not query_articles or not settings.newsapi_key:
rss_articles = fetch_google_news(
query=strategic_query,
language=settings.news_language,
)
query_articles.extend(rss_articles)
# Deduplicate within this batch (by URL)
new_articles = 0
for article in query_articles:
url = article.get('url', '')
if url and url not in seen_urls:
seen_urls.add(url)
all_articles.append(article)
new_articles += 1
if new_articles > 0:
logger.info(f" → Found {new_articles} new articles ({len(query_articles) - new_articles} duplicates skipped)")
stats["source"] = "newsapi+rss" if settings.newsapi_key else "rss"
if not all_articles:
logger.warning("No articles fetched from any source")
return stats
logger.info(f"Total unique articles fetched: {len(all_articles)}")
# Language filter
all_articles, lang_filtered = filter_by_language(
all_articles,
target_language=settings.news_language
)
stats["language_filtered"] = lang_filtered
# Get recent titles for fuzzy dedup
recent_titles = get_recent_titles(
session,
window_hours=settings.fuzzy_dedup_window_hours
)
# Process articles
for article in all_articles:
try:
title = clean_text(article.get("title", ""))
if not title:
continue
# Fuzzy dedup check
if is_fuzzy_duplicate(
title,
recent_titles,
threshold=settings.fuzzy_dedup_threshold
):
stats["fuzzy_filtered"] += 1
continue
# Prepare fields
description = clean_text(article.get("description", ""))
content = clean_text(article.get("content", ""))
url = normalize_url(article.get("url", ""))
source = article.get("source", "Unknown")
author = article.get("author", "")
published_at = article.get("published_at", datetime.now(timezone.utc))
# Generate keys
dedup_key = generate_dedup_key(
url=url,
title=title,
published_at=published_at,
source=source
)
canon_title = canonical_title(title)
# Upsert
stmt = get_upsert_stmt(
NewsArticle,
values={
"dedup_key": dedup_key,
"title": truncate_text(title, 500),
"canonical_title": truncate_text(canon_title, 500),
"description": truncate_text(description, 2000) if description else None,
"content": truncate_text(content, 10000) if content else None,
"url": url or None,
"source": source,
"author": author or None,
"language": settings.news_language,
"published_at": published_at,
"fetched_at": datetime.now(timezone.utc),
},
index_elements=["dedup_key"]
)
result = session.execute(stmt)
if result.rowcount > 0:
stats["imported"] += 1
# Add to recent titles for this batch
recent_titles.append(canon_title)
else:
stats["duplicates"] += 1
except Exception as e:
logger.warning(f"Error processing article: {e}")
continue
session.commit()
logger.info(
f"News ingestion complete: "
f"{stats['imported']} imported, "
f"{stats['duplicates']} duplicates, "
f"{stats['fuzzy_filtered']} fuzzy filtered, "
f"{stats['language_filtered']} language filtered"
)
return stats
# =============================================================================
# Price Ingestion
# =============================================================================
def fetch_symbol_with_retry(symbol: str, start_date, end_date, max_retries: int = 3, retry_delay: int = 30):
"""
Fetch price data for a symbol with retry on rate limit.
Args:
symbol: Ticker symbol
start_date: Start date
end_date: End date
max_retries: Maximum retry attempts
retry_delay: Seconds to wait between retries
Returns:
DataFrame or None if all retries failed
"""
import time
for attempt in range(max_retries):
try:
ticker = yf.Ticker(symbol)
df = ticker.history(
start=start_date.strftime("%Y-%m-%d"),
end=end_date.strftime("%Y-%m-%d"),
interval="1d"
)
return df
except Exception as e:
error_msg = str(e).lower()
if "rate limit" in error_msg or "too many requests" in error_msg:
if attempt < max_retries - 1:
logger.warning(f"{symbol}: Rate limited, waiting {retry_delay}s before retry {attempt + 2}/{max_retries}")
time.sleep(retry_delay)
else:
logger.error(f"{symbol}: Rate limit exceeded after {max_retries} retries")
raise
else:
raise
return None
def ingest_prices(session: Session) -> dict:
"""
Ingest price data for all configured symbols.
Uses INCREMENTAL fetching: checks latest bar date per symbol in DB
and only fetches from that point forward (plus 3-day overlap for corrections).
Falls back to full lookback if no existing data found for a symbol.
Returns:
Dict with stats per symbol
"""
import time
settings = get_settings()
# Fetch union of dashboard and training symbols (training may have different symbols)
dashboard_symbols = set(settings.symbols_list)
training_symbols = set(settings.training_symbols)
symbols = list(dashboard_symbols | training_symbols)
logger.info(f"Ingesting prices for {len(symbols)} symbols (dashboard={len(dashboard_symbols)}, training={len(training_symbols)})")
stats = {}
# Full lookback range (used only for first-time fetches)
end_date = datetime.now(timezone.utc)
full_start_date = end_date - timedelta(days=settings.lookback_days)
# Overlap buffer: re-fetch last 3 days to catch any corrections/adjustments
OVERLAP_DAYS = 3
for i, symbol in enumerate(symbols):
try:
# Check latest bar in DB for incremental fetch
latest_bar = session.query(PriceBar.date).filter(
PriceBar.symbol == symbol
).order_by(PriceBar.date.desc()).first()
if latest_bar and latest_bar.date:
# Incremental: fetch from (latest - overlap) to now
latest_date = latest_bar.date
if latest_date.tzinfo is None:
latest_date = latest_date.replace(tzinfo=timezone.utc)
start_date = latest_date - timedelta(days=OVERLAP_DAYS)
mode = "incremental"
else:
# First time: full lookback
start_date = full_start_date
mode = "full"
logger.info(f"Fetching prices for {symbol} ({mode})...")
# Fetch with retry mechanism
df = fetch_symbol_with_retry(symbol, start_date, end_date)
if df is None or df.empty:
logger.warning(f"No data returned for {symbol}")
stats[symbol] = {"imported": 0, "updated": 0, "error": "no_data"}
continue
imported = 0
updated = 0
for date_idx, row in df.iterrows():
try:
# Convert index to datetime
if hasattr(date_idx, 'to_pydatetime'):
bar_date = date_idx.to_pydatetime()
else:
bar_date = date_idx
# Ensure timezone
if bar_date.tzinfo is None:
bar_date = bar_date.replace(tzinfo=timezone.utc)
# Upsert
stmt = get_upsert_stmt(
PriceBar,
values={
"symbol": symbol,
"date": bar_date,
"open": float(row.get("Open", 0)) if row.get("Open") else None,
"high": float(row.get("High", 0)) if row.get("High") else None,
"low": float(row.get("Low", 0)) if row.get("Low") else None,
"close": float(row["Close"]),
"volume": float(row.get("Volume", 0)) if row.get("Volume") else None,
"adj_close": float(row.get("Adj Close", row["Close"])),
"fetched_at": datetime.now(timezone.utc),
},
index_elements=["symbol", "date"],
update_set={
"close": float(row["Close"]),
"adj_close": float(row.get("Adj Close", row["Close"])),
"fetched_at": datetime.now(timezone.utc),
}
)
result = session.execute(stmt)
if result.rowcount > 0:
imported += 1
else:
updated += 1
except Exception as e:
logger.debug(f"Error processing price bar: {e}")
continue
session.commit()
stats[symbol] = {"imported": imported, "updated": updated, "mode": mode}
logger.info(f"{symbol}: {imported} bars imported, {updated} unchanged ({mode}, {len(df)} fetched)")
# Add delay between symbols to avoid rate limiting
if i < len(symbols) - 1:
time.sleep(2) # 2 second delay between symbols
except Exception as e:
logger.error(f"Failed to fetch {symbol}: {e}")
stats[symbol] = {"imported": 0, "updated": 0, "error": str(e)}
return stats
# =============================================================================
# Main Entry Point
# =============================================================================
def fetch_all(
news: bool = True,
prices: bool = True
) -> dict:
"""
Run full data ingestion pipeline.
Args:
news: Whether to fetch news
prices: Whether to fetch prices
Returns:
Combined stats dict
"""
logger.info("Starting data ingestion pipeline...")
results = {
"news": None,
"prices": None,
"timestamp": datetime.now(timezone.utc).isoformat(),
}
with SessionLocal() as session:
if news:
results["news"] = ingest_news(session)
if prices:
results["prices"] = ingest_prices(session)
logger.info("Data ingestion complete")
return results
def main():
parser = argparse.ArgumentParser(
description="Fetch news and price data"
)
parser.add_argument(
"--fetch",
action="store_true",
help="Run data fetch"
)
parser.add_argument(
"--news-only",
action="store_true",
help="Fetch only news"
)
parser.add_argument(
"--prices-only",
action="store_true",
help="Fetch only prices"
)
parser.add_argument(
"--no-lock",
action="store_true",
help="Skip pipeline lock (for testing)"
)
parser.add_argument(
"--verbose", "-v",
action="store_true",
help="Verbose logging"
)
args = parser.parse_args()
if args.verbose:
logging.getLogger().setLevel(logging.DEBUG)
if not args.fetch:
parser.print_help()
return
# Initialize database
logger.info("Initializing database...")
init_db()
# Determine what to fetch
fetch_news = not args.prices_only
fetch_prices = not args.news_only
# Run with or without lock
if args.no_lock:
results = fetch_all(news=fetch_news, prices=fetch_prices)
else:
try:
with pipeline_lock():
results = fetch_all(news=fetch_news, prices=fetch_prices)
except RuntimeError as e:
logger.error(f"Could not acquire lock: {e}")
logger.info("Another pipeline process may be running. Use --no-lock to bypass.")
return
# Print summary
print("\n" + "=" * 50)
print("DATA INGESTION SUMMARY")
print("=" * 50)
if results.get("news"):
news = results["news"]
print(f"\nNews ({news.get('source', 'unknown')}):")
print(f" - Imported: {news.get('imported', 0)}")
print(f" - Duplicates: {news.get('duplicates', 0)}")
print(f" - Fuzzy filtered: {news.get('fuzzy_filtered', 0)}")
print(f" - Language filtered: {news.get('language_filtered', 0)}")
if results.get("prices"):
print("\nPrices:")
for symbol, stats in results["prices"].items():
status = f"{stats.get('imported', 0)} imported"
if stats.get("error"):
status = f"ERROR: {stats['error']}"
print(f" - {symbol}: {status}")
print(f"\nTimestamp: {results.get('timestamp', 'N/A')}")
if __name__ == "__main__":
main()