AI Prompts

Download high-frequency stock data from the HF Data Library and give me a summary.

Configuration:
  ticker     = AAPL
  version    = clean        # "clean" = filtered, "raw" = as-received
  api_key    = YOUR_API_KEY

Step 1 — Download the parquet file (entire history, ~24 years, single request):

  curl -H "X-API-Key: YOUR_API_KEY" \
       "https://api.hfdatalibrary.com/v1/bars/AAPL?version=clean" \
       -o AAPL.parquet

Step 2 — Load with pandas:

  import pandas as pd
  df = pd.read_parquet("AAPL.parquet")

Columns: datetime (Eastern Time, tz-naive), open, high, low, close, volume, source.
"source" is "pitrading" for pre-2022-03-04 and "iex" for post-2022-03-07.

Step 3 — Report:
  - total bars
  - date range (first and last datetime)
  - mean / min / max close price
  - number of trading days with at least one bar
  - first 5 rows

Download one-minute OHLCV data from the HF Data Library for several tickers and combine them into a single DataFrame.

Configuration:
  tickers    = AAPL, MSFT, GOOG, NVDA, AMZN, META, TSLA
  version    = clean
  api_key    = YOUR_API_KEY

For each ticker, fetch the parquet:

  curl -H "X-API-Key: YOUR_API_KEY" \
       "https://api.hfdatalibrary.com/v1/bars/${TICKER}?version=clean" \
       -o "${TICKER}.parquet"

Combine in Python:

  import pandas as pd, glob, os
  dfs = []
  for path in sorted(glob.glob("*.parquet")):
      d = pd.read_parquet(path)
      d["ticker"] = os.path.basename(path).replace(".parquet", "")
      dfs.append(d)
  combined = pd.concat(dfs, ignore_index=True)
  combined = combined[["ticker", "datetime", "open", "high", "low", "close", "volume", "source"]]

Report:
  - bars per ticker
  - shared (overlapping) date range across all tickers
  - mean daily volume per ticker
  - total combined row count and memory footprint

Download HF Data Library data for one ticker, slice to a specific date range, and compute daily realized volatility.

Configuration:
  ticker     = NVDA
  version    = clean
  start_date = 2024-10-01
  end_date   = 2024-12-31
  api_key    = YOUR_API_KEY

Step 1 — Download (the API ships entire-ticker files; slice client-side):

  curl -H "X-API-Key: YOUR_API_KEY" \
       "https://api.hfdatalibrary.com/v1/bars/NVDA?version=clean" \
       -o NVDA.parquet

Step 2 — Slice and compute realized variance (sum of squared 1-min log returns per day):

  import numpy as np, pandas as pd
  df = pd.read_parquet("NVDA.parquet")
  df["datetime"] = pd.to_datetime(df["datetime"])
  window = df[(df.datetime >= "2024-10-01") & (df.datetime < "2025-01-01")].copy()
  window["date"] = window.datetime.dt.date
  window["log_ret"] = np.log(window.close).diff()
  daily_rv = window.groupby("date")["log_ret"].apply(lambda r: (r**2).sum())

Report:
  - highest-RV trading day and its value
  - the 5 trading days with the largest absolute close-to-close moves
  - mean RV across the window
  - a simple chart of daily RV over the window

I have an HF Data Library API key (https://hfdatalibrary.com) and need a shell command to download 1-minute OHLCV bars for a ticker.

Configuration:
  ticker  = AAPL
  version = clean
  api_key = YOUR_API_KEY

The endpoint is:
  GET https://api.hfdatalibrary.com/v1/bars/{TICKER}?version={VERSION}
  Header: X-API-Key: YOUR_API_KEY

Give me a single curl command that:
  1. authenticates with X-API-Key
  2. downloads the parquet file to ./AAPL_clean.parquet
  3. fails with a non-zero exit code on HTTP error (so the shell knows it broke)
  4. prints a short success message with the file size on completion

Also give me the equivalent commands in:
  - Python (using requests, save to disk)
  - R (using httr2 or curl, save to disk)

No explanation, just the four commands.

Build a one-minute OHLCV dataset for a sector / index from the HF Data Library.

Configuration:
  bundle     = Magnificent 7        # or: S&P 500, Dow 30, sector-tech, sector-financials
  tickers    = AAPL, MSFT, GOOG, NVDA, AMZN, META, TSLA
  version    = clean
  api_key    = YOUR_API_KEY
  out_dir    = mag7_data/

If the bundle name is recognized but no tickers are supplied, look up the constituents (you can use https://hfdatalibrary.com/pages/tickers as a reference or a public source for index membership).

Download each ticker, throttled to ~1 request/second to stay under the 100/min rate limit:

  mkdir -p mag7_data
  for T in AAPL MSFT GOOG NVDA AMZN META TSLA; do
    curl -H "X-API-Key: YOUR_API_KEY" \
         "https://api.hfdatalibrary.com/v1/bars/${T}?version=clean" \
         -o "mag7_data/${T}.parquet"
    sleep 1
  done

Then in Python:
  import pandas as pd, glob, os
  files = sorted(glob.glob("mag7_data/*.parquet"))
  combined = pd.concat([
      pd.read_parquet(f).assign(ticker=os.path.basename(f).replace(".parquet",""))
      for f in files
  ], ignore_index=True)
  combined.to_parquet("mag7_combined.parquet")

Report total bars, ticker count, date range, and combined-file size.

Download HF Data Library data at a non-1-minute timeframe. The /v1/bars/{TICKER} endpoint only serves 1min — for other intervals, use the signed-URL flow.

Configuration:
  ticker     = SPY
  version    = clean
  timeframe  = 5min          # allowed: 1min, 5min, 15min, 30min, hourly, daily, weekly, monthly
  format     = parquet       # or csv
  api_key    = YOUR_API_KEY

Step 1 — Request a signed download URL (the token is valid 10 minutes):

  curl -H "Authorization: Bearer YOUR_API_KEY" \
       "https://api.hfdatalibrary.com/v1/download-token/SPY?version=clean&timeframe=5min&format=parquet"

Response (JSON):
  {
    "url": "https://api.hfdatalibrary.com/v1/download/SPY?token=...",
    "expires_at": "...",
    "version": "clean", "timeframe": "5min", "format": "parquet"
  }

Step 2 — Fetch the `url` value (no auth header needed; the token authorizes it):

  curl "" -o SPY_clean_5min.parquet

Then load with pandas as in prompt 1. Show me the first 5 rows and confirm the inter-bar spacing is 5 minutes.