Getting Started

ThetaDataDx connects directly to ThetaData's servers — no Java terminal process to install or babysit. Pick a language, install, save your credentials, and make a request.

1. Install

# Cargo.toml
[dependencies]
thetadatadx = "12"
tokio = { version = "1", features = ["rt-multi-thread", "macros"] }

The historical client is async; tokio provides the runtime.

pip install thetadatadx

# Optional DataFrame adapters:
pip install thetadatadx[pandas]    # pandas
pip install thetadatadx[polars]    # polars
pip install thetadatadx[arrow]     # pyarrow only

Requires Python 3.12+. Pre-built abi3 wheels for Linux x86_64, macOS, and Windows — no Rust toolchain needed. On other platforms, build from source with maturin (maturin develop --release in sdks/python).

npm install thetadatadx

Requires Node.js 18+. Pre-built native binaries install automatically per platform.

# Prerequisites: C++17 compiler, CMake 3.16+, Rust toolchain
git clone https://github.com/userFRM/ThetaDataDx.git
cd ThetaDataDx
cargo build --release -p thetadatadx-ffi

cd sdks/cpp
cmake -B build && cmake --build build

Include sdks/cpp/include/thetadx.hpp (header-only over the C ABI) and link the built thetadatadx_ffi library. Any other language with C interop can target the same C ABI directly.

cargo install thetadatadx-server --git https://github.com/userFRM/ThetaDataDx

The server binary exposes the full surface as local HTTP REST and WebSocket endpoints — no SDK required. Existing scripts written against the v3 route surface point at it unchanged.

2. Save credentials

Create a creds.txt in your working directory: your ThetaData account email on line 1, password on line 2.

you@example.com
your-password

No subscription yet? Create an account at thetadata.net — several endpoints work on the free tier (look for the Free badge on reference pages).

3. First request

use thetadatadx::{Credentials, DirectConfig, ThetaDataDxClient};

#[tokio::main]
async fn main() -> Result<(), thetadatadx::Error> {
    let creds = Credentials::from_file("creds.txt")?;
    let tdx = ThetaDataDxClient::connect(&creds, DirectConfig::production()).await?;

    let rows = tdx.stock_history_eod("AAPL", "20250303", "20250306").await?;
    for t in &rows {
        println!("{}: open={} close={} volume={}", t.date, t.open, t.close, t.volume);
    }
    Ok(())
}

from thetadatadx import Config, Credentials, ThetaDataDxClient

creds = Credentials.from_file("creds.txt")
tdx = ThetaDataDxClient(creds, Config.production())

rows = tdx.stock_history_eod("AAPL", "20250303", "20250306")
for t in rows:
    print(t.date, t.open, t.close, t.volume)

import { ThetaDataDxClient } from 'thetadatadx';

const tdx = ThetaDataDxClient.connectFromFile('creds.txt');

const rows = await tdx.stockHistoryEOD('AAPL', '20250303', '20250306');
for (const t of rows) {
  console.log(t.date, t.open, t.close, t.volume);
}

#include "thetadx.hpp"
#include <iostream>

int main() {
    auto creds = tdx::Credentials::from_file("creds.txt");
    auto client = tdx::MddsClient::connect(creds, tdx::Config::production());

    auto rows = client.stock_history_eod("AAPL", "20250303", "20250306");
    for (const auto& t : rows) {
        std::cout << t.date << ": open=" << t.open
                  << " close=" << t.close << " volume=" << t.volume << "\n";
    }
}

thetadatadx-server --creds creds.txt &

curl 'http://127.0.0.1:25503/v3/stock/history/eod?symbol=AAPL&start_date=2025-03-03&end_date=2025-03-06'

Every endpoint follows this shape. Browse the API Reference — each page carries the signature and a runnable sample in all five surfaces.

Good to knows

• Dates are YYYYMMDD strings in the SDKs ("20250303"); the HTTP server also accepts ISO YYYY-MM-DD. Timestamps come back as milliseconds since midnight Eastern Time — see Symbology & Contract Identity.
• Connect once, reuse the client. One client multiplexes any number of historical requests and an optional streaming session; per-request connections waste the authentication round trip.
• Markets closed? Connect with Config.dev() / DirectConfig::dev() to stream a replayed historical session, and prefer historical endpoints over snapshots on weekends.