Installation¶

Tradetropy requires Python 3.12+.

Base install¶

pip install tradetropy

The base package already includes the full backtesting engine, the indicator library, the plotting stack (Bokeh) and data IO for CSV and the NumPy .npz binary format. .npz is the base binary format: it needs no compiler and works everywhere NumPy does (including Termux/Android), so you do not need any extra to save, load or record data.

Optional extras¶

Extras add broker/venue integrations and optional file formats. Install only what you need:

pip install tradetropy[mt5]       # MetaTrader 5 connector
pip install tradetropy[ccxt]      # CCXT crypto exchanges (live + streaming)
pip install tradetropy[bybit]     # Bybit connector (pybit)
pip install tradetropy[parquet]   # Parquet read/save support (pyarrow)
pip install tradetropy[hdf5]      # HDF5 read/save support (PyTables)
pip install tradetropy[all]       # all runtime extras above at once

Extra	Adds	Use it for
`mt5`	`MetaTrader5`	Trading and data through MetaTrader 5
`ccxt`	`ccxt`	Crypto live trading and WebSocket streaming
`bybit`	`pybit`	The Bybit connector
`parquet`	`pyarrow`	Reading/saving `.parquet` files
`hdf5`	`tables` (PyTables)	Reading/saving `.h5` / `.hdf5` files
`all`	all of the above	Everything at once

Note

The connectors import their third-party library lazily, so a missing extra only raises when you actually construct that connector - the rest of the library keeps working. Likewise, the HDF5 path raises a clear error only when you pass format='hdf5' (or a .h5/.hdf5 path) without the hdf5 extra; use the base .npz format instead.

With uv¶

If you develop with uv:

uv add tradetropy
uv add "tradetropy[ccxt]"

Termux (Android)¶

Most users install on Windows, macOS or Linux, where every dependency (numpy, pandas, bokeh) ships as a prebuilt wheel and pip install tradetropy just works with no compilation. Termux is the exception: it is Android on aarch64, and PyPI publishes no wheels for it. The scientific stack, though, already ships as prebuilt Termux packages (pkg install), so no compiler and no --no-build-isolation build are needed - install numpy / pandas / contourpy from pkg first, then pip install tradetropy picks them up.

Crucially, the base install does not require PyTables: it cannot build on Termux (it links the HDF5 C library through Blosc2, which Termux does not package), so tradetropy uses the NumPy .npz binary format instead. .npz is as fast as HDF5 for whole-file save/load and needs nothing beyond NumPy, so you get the full data-IO stack on Termux with no HDF5 at all. Do not try to install the hdf5 extra there.

# 1. Update Termux and install Python.
pkg update && pkg upgrade -y
pkg install -y python

# 2. If you ever ran `pip install numpy` (or pandas/contourpy) by mistake
#    BEFORE this step, pip's own wheel is now shadowing the pkg version and
#    the pkg install below will fail or silently not take effect. Undo it
#    first: uninstall the pip-installed copy, then purge both caches so a
#    stale wheel is never reused, and only then install via pkg.
#      pip uninstall -y numpy pandas contourpy
#      pkg clean
#      pip cache purge

# 3. Install the prebuilt scientific stack via pkg (NOT pip). contourpy is
#    pulled in by Bokeh (the plotting stack). One command is equivalent to
#    installing them separately.
pkg install -y python-numpy python-pandas python-contourpy

# Optional: only if you want to read/save .parquet files (pyarrow has a
# prebuilt Termux package). HDF5 does not work on Termux - do not install
# the hdf5 extra there (see note above).
pkg install -y python-pyarrow

# 4. Restart Termux (close and reopen the app) so the new libraries are
#    picked up by a fresh shell, then install tradetropy normally.
pip install tradetropy

Verify the versions tradetropy will use:

python -c "import numpy, pandas; print('numpy', numpy.__version__); print('pandas', pandas.__version__)"

Verified on-device

Confirmed working on Termux with numpy==2.4.4 and pandas==3.0.4 (both satisfy tradetropy's numpy>=2.0,<2.5 / pandas>=3.0,<4 pins), installing tradetropy==0.1.4 with no compilation step.

Warning

Do not run pip install --upgrade pip on Termux. It ships a patched pip that fixes Termux-specific build issues; the upstream version breaks native builds. Also use the official Termux app (from F-Droid or the GitHub releases), not the Google Play build, which is unmaintained.

If a future Termux update ships an older numpy/pandas that falls outside tradetropy's pins, fall back to building them from source against Termux's toolchain (pkg install build-essential cmake ninja libopenblas, then pip install --no-build-isolation numpy pandas before pip install tradetropy).

From source¶

git clone https://github.com/michiTrader/tradetropy.git
cd tradetropy
uv sync            # installs the project + dev tooling (pytest)
uv run pytest      # run the test suite

Documentation toolchain¶

To build this documentation site locally, install the docs extra:

pip install tradetropy[docs]
mkdocs serve