GitHub

DatabentoAPI.jl Documentation

Julia client for the Databento market-data APIs:

  • Historical — HTTP API for backfills, metadata, batch jobs, and symbology.
  • Live — TCP API for low-latency real-time feeds with reconnect, multiple schemas, and crash-safe capture to disk.

Built on top of DatabentoBinaryEncoding.jl, which handles all binary record decoding. DatabentoAPI.jl is a thin transport

  • session layer; the heavy lifting on the wire format lives in the sister

package.

What this package wraps

  • Authenticated HTTP requests to the Historical gateway.
  • The Live TCP protocol: CRAM handshake, subscription frames, multi-schema routing, full-jitter exponential backoff on reconnect.
  • Capture to .dbn[.zst] files with optional time-based file rotation and in-file zstd frame rotation for crash safety.
  • Conversion helpers wrapping DBN.jl's CSV / JSON / Parquet / DataFrame converters.

Quick taste

using DatabentoAPI, Dates

# Historical: pull a minute of trades and convert to a DataFrame.
client = Historical()
store  = get_range(client;
    dataset  = "XNAS.ITCH",
    schema   = Schema.TRADES,
    symbols  = ["AAPL"],
    start    = DateTime(2024, 1, 2, 14, 30),
    end_     = DateTime(2024, 1, 2, 14, 31),
    stype_in = SType.RAW_SYMBOL)
df = to_dataframe(store)

# Live: subscribe and iterate inside a do-block; clean shutdown on Ctrl-C.
Live(dataset = "GLBX.MDP3") do live
    connect!(live)
    subscribe!(live; schema = Schema.TRADES, symbols = ["ES.FUT"], stype_in = SType.PARENT)
    start!(live)
    for rec in live
        @info "tick" rec
    end
end

Where to go next

Status

The package is in active development. Core functionality is tested (1500+ offline tests against mocked TCP gateways and HTTP servers) and used in production captures, but the public API may still evolve. Pin to the patch version when stability matters.