Returns structured narrative data: persistent story threads (global narratives),
daily article clusters, individual articles, and sentiment scores. Use this to
understand what stories are driving a stock and how sentiment is evolving.

Args:
    tickers: Comma-separated ticker symbols. Examples: "NVDA" or "NVDA,TSLA,AAPL".
             Maximum 10 tickers per request. Free tier: TSLA, NVDA, AAPL, MSFT,
             AMZN, GOOGL, META, AMD, NFLX, JPM. Plus/Scale: all 109 tickers.
    from_date: Start date in YYYY-MM-DD format. Defaults to 7 days ago.
    to_date: End date in YYYY-MM-DD format. Defaults to today.

Returns:
    Dict with narrative data per ticker, including global_narratives (persistent story
    threads with title, sentiment, and dominance), daily_topics (day-level article
    clusters), and articles (individual news items with relevance and sentiment scores).

Groups narratives across all holdings by macro theme (e.g. "AI Regulation",
"Interest Rate Sensitivity") and identifies which stories create concentrated
narrative exposure across multiple positions simultaneously.

Args:
    holdings: Portfolio positions as TICKER:WEIGHT pairs, comma-separated.
              Weights represent portfolio allocation (should sum to ~1.0).
              Example: "NVDA:0.30,AAPL:0.25,TSLA:0.20,MSFT:0.25"
              Maximum 50 holdings.

Returns:
    Dict with macro risk themes, each showing affected tickers, dominant narrative
    titles, sentiment trajectory, and weighted exposure score. Also includes an
    overall portfolio narrative concentration score.