Candles & Funding

Historical OHLCV and funding rates from immutable dataset releases.

Both endpoints read from validated, checksummed dataset releases — the same bytes the backtest engine uses. Coverage per venue/pair is in /context/datasets.

GET/context/candles

Query parameters

ParamTypeNotes
venuestringRequired
symbolstringRequired — the canonical symbol exactly as returned by /context/markets (never a framework alias)
timeframestring1m, 5m, 15m, 1h, 4h, 1d (aggregated server-side from 1m)
from / toISO 8601Required window; max 10,000 candles per request
cursorstringContinue a windowed read

Response — 200

json
{
  "venue": "lighter",
  "symbol": "SOL-PERP.LIGHTER",
  "timeframe": "1h",
  "release": "rel_2026-07-27T02",
  "candles": [
    { "t": "2026-07-01T00:00:00Z", "o": 141.22, "h": 143.05, "l": 140.9, "c": 142.71, "v": 183422.5 }
  ],
  "gaps": [],
  "next_cursor": null
}
  • release pins the immutable dataset release the rows came from — quote it and any re-read returns identical bytes. On venues whose release pipeline is still pending (Hyperliquid, Binance — see datasets), release is null and reproducibility is not claimed.
  • gaps lists intervals the venue itself never served (venue-side holes are recorded, never interpolated). An empty array is a completeness statement, not a default.
  • Bulk export: send Accept: text/csv for the same window as CSV — one row per candle, same fields.

GET/context/funding

Hourly funding rates for perp venues.

Query parameters: venue, symbol, from, to (same semantics as candles).

Response — 200

json
{
  "venue": "lighter",
  "symbol": "SOL-PERP.LIGHTER",
  "release": "rel_2026-07-27T02",
  "fundings": [
    {
      "t": "2026-07-01T00:00:00Z",
      "rate": 0.0000125,
      "rate_annualized": 0.1095
    }
  ],
  "gaps": [
    { "from": "2025-01-21T14:00:00Z", "to": "2025-01-21T15:00:00Z", "reason": "venue_history_hole" }
  ]
}

rate is the decimal fraction per funding interval; rate_annualized is derived for convenience. Funding history is the input for carry analysis and cost-of-hold modeling — the things a directional backtest silently ignores.

Freshness

Releases publish on backfill cadence (hourly for active venues), so the most recent ~1–2 hours may not be in a release yet. to values beyond the latest release are clamped, and the response's release timestamp tells you the effective horizon. Context is a research surface; execution-time data belongs to your deployed framework's live venue connection.

Previous
Markets
Next
Datasets