Brue

Brue is the trade-entry language. A Brue script writes buy/sell/close decisions over bars, and the same file runs identically in a historical run and live against a broker. Brue is for two things and nothing else: writing trade entries, and connecting them to brokers.

Two products, one install:

• Brue, the language (this package, import brue): the script engine and the Python entry points. .brue scripts are pure trading logic: no header, no credentials, no configuration block. What a script trades, at what quantity, from which data source, at which timeframe is written where it acts.
• Brue Connect (brueconnect, bundled inside this package): the broker connector. One wire protocol, every broker an adapter, the built-in paper account first. It ships with the language, so there is nothing else to install; the language calls it.

pip install brue-language

(brue-language is the distribution; the import is brue. The bare brue name on PyPI belongs to someone else.)

A script

fast = ema(EURUSD.h1.close, 9)
slow = ema(EURUSD.h1.close, 21)
if crossover(fast, slow):
    buy("EURUSD", 0.5, sl=EURUSD.h1.close * 0.995, tp=EURUSD.h1.close * 1.01)
if crossunder(fast, slow):
    close("EURUSD")

Over history, from the command line:

python -m brue run trend.brue eurusd_h1.csv --capital 100000 --commission 0.001

The same, from Python, and then live:

import brue

bars = brue.parse_bars_csv(open("eurusd_h1.csv").read())
out = brue.run_script(open("trend.brue").read(), bars)      # historical run
out.commands                                                # entries, exits, pnl

brue.login(api_key="lse_...")          # the only credential that ever appears
brue.buy("EURUSD", 0.5, sl=1.05, tp=1.12)                   # live, via Brue Connect
brue.close("EURUSD")

Historical runs simulate the fills (next-bar-open, brackets, commission, slippage) and report what was executed: the trade records, the log() lines, the canonical JSON. There is no stats block, no equity frame and no pandas indicator surface in this package; research on the results happens in plain Python on the host. Indicator scripts (plot() calls, no orders) are exposed through brue.Indicator for chart hosts. The language reference is SYNTAX.md; the boundary of what Brue owns is ARCHITECTURE.md.

Why the decisions can be trusted

The engine is conformance-locked: conformance/run_all_py.sh runs the corpus (every construct of the language plus the loud-error cases) and byte-compares the execution JSON against locked goldens. A fill or a formula cannot silently drift, and a live run re-derives the same order intents from the same script text.

Status

Version 0.0.1. Pure Python, stdlib-only engine (pandas is optional and only used by the convenience frames in brue.live and brue.Indicator.compute).

Account login

Anything account-bound (LSE data, placing orders) starts with your LSE API key, the same key you use on the data platform:

import brue
brue.login(api_key="lse_...")   # or BRUE_API_KEY / LSE_API_KEY env,
brue.account()                  # or ~/.config/brue/config.json

Historical runs on your own bars never need a key: the engine runs fully offline.