Layered, broker-agnostic architecture (Alpaca adapter included, others drop in)
guest@make-directory:~$ open ./apps/tradeflow
make-directory:~/apps/tradeflow
TradeFlow
TradeFlow is a small but complete, layered trading engine built around clean abstractions: a universe scanner picks symbols, a strategy turns market data into signals, and an engine either backtests on history or trades live (paper by default). It is broker-agnostic — everything above the broker layer depends on interfaces, with Alpaca as the bundled adapter — and ships with pure pandas/numpy indicators (RSI, EMA, ATR, volume spikes, beta) instead of TA-Lib, plus optional parameter optimization and constraint-solver (OR-Tools) portfolio allocation. It is educational tooling and a reference architecture, not financial advice or a managed trading product.
guest@make-directory:~$ cat ./highlights.md
Backtest and live (paper) trading driven by the same strategy
Universe scanner, parameter optimization (grid/random/Bayesian), and OR-Tools portfolio allocation
Pure pandas/numpy indicators (no TA-Lib); uv- or Docker-based setup
Offline test suite + CI, with explicit educational-use and risk disclaimers
guest@make-directory:~$ cat ./technical-notes.md
Public repository: makedirectory/tradeflow
Python 3.10+ managed with uv (or run via Docker); requires user-provided Alpaca API credentials (paper by default)
Optional extras: scikit-learn for Bayesian tuning, Google OR-Tools for portfolio allocation
Useful lower-hanging-fruit app/tool page because the project is public and easy to explain
guest@make-directory:~$ cat ./privacy-and-policy.md
privacy and policy