JrVillabona/resonance-data

Resonance Macro and Market Data

Daily-normalized macro, economic, market, crypto, on-chain, DeFi, derivatives, and ETF time series for research and model training.

This Hugging Face dataset package is generated from the Resonance data collector. The default table is long-format partitioned Parquet:

data/compatibility/**/*.parquet

Canonical data is published as independent configurations:

binance_bulk_bars
compatibility
cross_venue_bars
crypto_derivatives
derived_feature_lineage
documents
events
fiscal_observations
fundamentals
macro_point_in_time
market_bars_1d
official_macro_observations
positioning_observations
scalar_observations
sec_entities
sec_filings

Columns

• symbol: canonical series name from config/symbols.json
• date: normalized observation date
• open, high, low, close, volume: canonical OHLCV-compatible values
• source, ticker, category, data_type, frequency, description: series metadata

The compatibility configuration preserves the legacy compatibility schema. New consumers should use market_bars or scalar_observations; scalar observations expose a single value instead of synthetic OHLCV columns.

macro_point_in_time preserves ALFRED release and revision dates for point-in-time training.

fiscal_observations contains dimension-preserving US Treasury Fiscal Data, including debt, operating cash, receipts/outlays, and average interest rates.

documents contains source documents prepared for language-model training. Phase 6 document sources include bounded SEC filing primary-document text and official Federal Reserve, ECB, Bank of England, Bank of Japan, and Bank of Canada, Reserve Bank of Australia, and Swiss National Bank publication text. Raw payload checksums are retained in quality flags and exact raw responses are stored separately in the local lake.

events contains derived event/model features, currently including bounded GDELT event-export features. These are intentionally separate from raw source documents.

official_macro_observations contains direct BLS, BEA, ECB, Eurostat, BIS, BOJ, and EIA observations with source-native dimensions and conservative availability metadata. BOJ coverage includes monetary base, reserve balances, money stock, total assets, JGB holdings, ETF/J-REIT holdings, loans, and current deposits for Japan liquidity analysis. BIS coverage includes bounded keyed debt-securities and cross-border banking series for Japan, the United States, the United Kingdom, Switzerland, China where available, Canada, Australia, and Sweden in addition to the broader prices, credit, property, policy-rate, and exchange-rate feeds. IMF IFS coverage includes bounded keyed reserves, reserve-money, broad-money, private-credit, money-market-rate, policy-rate, and FX-rate series for major liquidity countries where available.

scalar_observations also includes calculated Japan liquidity factors: BOJ_OFFICIAL_ASSETS_USD_B, BOJ_MONETARY_BASE_USD_B, BOJ_RESERVE_BALANCES_USD_B, and BOJ_OFFICIAL_ASSETS_MOM_USD_B. It also includes native-unit BOJ total-assets, monetary-base, and reserve-balance month-over-month and year-over-year growth factors. It includes calculated IMF IFS reserve factors for the United States, Japan, euro area, China, the United Kingdom, Switzerland, Canada, Australia, and Sweden, plus IMF_SELECTED_RESERVES_USD_B and IMF_SELECTED_RESERVES_MOM_USD_B. It also includes date-aligned IMF IFS reserve-money and broad-money month-over-month and year-over-year percentage-change features for selected liquidity countries where those IFS series are available. Current monthly private-sector credit growth factors are included for Japan, Australia, and Sweden; stale country series are intentionally excluded. Policy-rate change factors are included over one-month, three-month, and twelve-month horizons for the United States, Japan, euro area, China, the United Kingdom, Switzerland, Canada, and Australia. It also includes DB.NOMICS BIS global liquidity series for USD, EUR, and JPY total credit plus private-sector credit.

derived_feature_lineage maps each materialized calculated feature to its source series, deterministic transformation, and date-alignment policy. Join it to scalar observations by symbol; referenced official inputs remain in official_macro_observations or the corresponding source table.

positioning_observations contains weekly CFTC Commitments of Traders positions, trader counts, open interest, and concentration measures.

Quick Start

from datasets import load_dataset

ds = load_dataset("JrVillabona/resonance-data", split="train")
btc = ds.filter(lambda row: row["symbol"] == "BTC")

macro = load_dataset(
    "JrVillabona/resonance-data",
    "macro_point_in_time",
    split="train",
)

import pandas as pd

url = "hf://datasets/JrVillabona/resonance-data/data/compatibility/**/*.parquet"
df = pd.read_parquet(url)

Package Manifest

The full machine-readable package metadata is kept in manifest.json; file checksums are kept in publication_manifest.json. Summary:

• generated at: 2026-07-25T06:34:11+00:00
• configured symbols: 1,566
• materialized symbols: 392
• compatibility rows: 1,119,587
• date range: 1854-12-01 to 2026-07-27
• target shard size: 384 MB

Dataset configs:

• binance_bulk_bars: 65,985,580 rows, 31 shards
• compatibility: 1,119,587 rows, 21 shards
• cross_venue_bars: 66,280 rows, 45 shards
• crypto_derivatives: 29,353 rows, 5 shards
• derived_feature_lineage: 76 rows, 1 shards
• documents: 1,073 rows, 5 shards
• events: 1,010 rows, 2 shards
• fiscal_observations: 1,506,408 rows, 25 shards
• fundamentals: 111,155,374 rows, 220 shards
• macro_point_in_time: 121,267 rows, 15 shards
• market_bars_1d: 315,606 rows, 11 shards
• official_macro_observations: 268,831,194 rows, 250 shards
• positioning_observations: 30,766,759 rows, 57 shards
• scalar_observations: 774,628 rows, 21 shards
• sec_entities: 972,970 rows, 20 shards
• sec_filings: 26,915,487 rows, 218 shards

Documentation

• DATA_DICTIONARY.md: researcher-friendly series descriptions by category.
• DATA_CATALOG.md: source, endpoint, unit, materialization status, and quality notes.
• USAGE_GUIDE.md: loading and alignment guidance.
• quality/report.json: release-level quality summary.
• quality/series_quality.parquet: coverage, freshness, gaps, and availability.
• quality/source_quality.parquet: health, license, and redistribution status.
• quality/revision_quality.parquet: ALFRED revision coverage.
• quality/sec_quality.parquet: SEC entity/table coverage and primary-key checks.
• quality/treasury_quality.parquet: Treasury partition coverage and primary-key checks.
• quality/official_macro_quality.parquet: official macro coverage, primary-key checks, BOJ unit/cadence/staleness checks, and availability coverage.
• quality/cftc_quality.parquet: CFTC report-family coverage and primary-key checks.
• run_status.json: latest local scheduler result, step durations, and publish target.
• publication_manifest.json: SHA-256 checksum for every managed file.

Refresh Policy

The dataset is currently refreshed by local scheduled jobs to avoid exhausting GitHub Actions free-tier minutes. Daily market updates publish bounded changes; monthly macro jobs refresh revised official sources such as BOJ, ECB, Eurostat, BIS, BLS, BEA, ALFRED, and Treasury Fiscal Data. Historical immutable archives are fetched once and then updated through bounded overlap windows.

Important Notes

This dataset aggregates third-party public data sources. Check each upstream source's terms before commercial redistribution or regulated use.