Skip to main content

Build Engine, a self-improving adaptive prediction and paper-trading engine

Build Engine

Self-improving adaptive prediction and paper-trading engine that bridges build-oracle forecasting models to build-finance execution through an accuracy feedback loop.

Project Telos | gather | crucible | index | forum | telos | emet | buildlang

CI version: 1.0.1 python: 3.10+ core deps: numpy/scipy license: fair-source

Not financial advice. Paper trading is the default; live broker execution is explicit opt-in and gated; the engine custodies no funds. Read SECURITY.md before enabling live mode.

Build Engine is a paper-first adaptive prediction engine for market research, strategy simulation, and model feedback loops.

It connects build-oracle forecasting models with build-finance paper trading and backtesting primitives so predictions can be evaluated against outcomes instead of left as static charts.

This repository is published as fair-source software. The code is available for review, use, and building on, while commercial use that competes with the project is reserved under the license in this repository.

What It Does

Market data -> models train -> predictions -> paper trades
       ^                                      |
       |                                      v
       +--------- results and weights <-------+
  • Trains forecasting models through build-oracle.
  • Converts model forecasts into strategy signals.
  • Runs those signals through build-finance paper execution primitives.
  • Tracks directional accuracy and adjusts model weights over time.
  • Keeps live broker execution behind explicit private-risk gates.

Install

pip install -e ../build-oracle
pip install -e ../build-finance
pip install -e .

The GUI code remains private/license-gated until the Qt binding posture is resolved. The public-ready package path is CLI and library first.

Paper Mode

Paper trading is the default.

build-engine run --symbols AAPL,BTC-USD --paper --cycles 10

Backtest the prediction strategy:

build-engine backtest --symbols AAPL --days 252 --monte-carlo

Check status:

build-engine status

Live Broker Mode

Live broker mode is not the default public path. It requires all of the following:

  • --live
  • BUILD_ENGINE_LIVE_ACK=I_UNDERSTAND_LIVE_RISK or matching --live-ack
  • APCA_API_KEY_ID
  • APCA_API_SECRET_KEY

Example with placeholders:

$env:BUILD_ENGINE_LIVE_ACK="I_UNDERSTAND_LIVE_RISK"
$env:APCA_API_KEY_ID="<paper-or-live-key-id>"
$env:APCA_API_SECRET_KEY="<paper-or-live-secret>"
build-engine run --symbols AAPL --live --live-ack I_UNDERSTAND_LIVE_RISK --cycles 1

The engine does not persist broker API keys or the live-mode acknowledgement in its saved state.

Python API

from build_engine.adaptive_engine import AdaptiveEngine
from build_engine.config import EngineConfig

config = EngineConfig(
    symbols=["AAPL", "BTC-USD"],
    models=["arima", "prophet"],
    paper_trading=True,
)

engine = AdaptiveEngine(config)
result = engine.run_cycle()
print(f"Equity: ${result['equity']:,.2f}")
print(f"Accuracy: {result['accuracy']:.0%}")

Risk Posture

This software is research and engineering tooling. It is not investment, financial, legal, tax, or trading advice. Model output can be wrong, delayed, overfit, incomplete, or inappropriate for real capital allocation. Paper mode is the supported default. Live broker mode is a private-risk capability and should be used only by an operator who understands the consequences.

Architecture

Module Role
config.py Central configuration dataclass and live-mode acknowledgement constant.
model_trainer.py Wraps build-oracle model fitting and prediction.
prediction_strategy.py Bridges oracle forecasts to build-finance signals.
performance_tracker.py Tracks accuracy and computes dynamic model weights.
adaptive_engine.py Wires model training, prediction, broker setup, and feedback.
persistence.py Saves non-secret engine state and trade history.
cli.py Command-line interface.

Verification Snapshot

Current release-gate evidence is recorded in the workspace roadmap contract:

project-docs/roadmaps/contracts/product-use-case-build-engine-2026-06-12.json

The public claim is intentionally narrow: adaptive forecasting and paper-first execution are verified; live broker mode remains explicit and private-risk gated.

About The Author

Build Engine is part of Zain Dana Harper's Build workspace: a set of language, state, forecasting, color, and verification tools built around practical experiments rather than conventional product categories. The work is deliberately cross-disciplinary: compiler thinking, market simulation, visual systems, and local-first safety boundaries.

License

Copyright (c) 2022-2026 Zain Dana Harper. All rights reserved.

Build Engine is released under the FSL-1.1-MIT. The source is available: you may read, run, copy, modify, and redistribute it for any purpose other than a Competing Use. Commercial use that competes with the project is reserved to the Licensor and requires a separate commercial license. See LICENSE for the full terms.

Release files for build-engine 1.0.1

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for build-engine 1.0.1
File Size Uploaded
build_engine-1.0.1.tar.gz 68.4 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for build-engine 1.0.1
File Interpreter ABI Platform
build_engine-1.0.1-py3-none-any.whl Python 3 none any Details

Total release size: 133.6 kB

Release files / build_engine-1.0.1.tar.gz

Download URL build_engine-1.0.1.tar.gz
Size 68.4 kB
Tags Source
SHA-256 checksum
How to use checksums
c2c40e90da6e6d785d69c580754788ac372d1aadbecd60eb1c93f3c0cee9598c
BLAKE2b-256 checksum
How to use checksums
efbbd7bce700443b543771336b629940e2586b8df7ffa82a9ab323d5ef8c1958
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.12.10

Release files / build_engine-1.0.1-py3-none-any.whl

Download URL build_engine-1.0.1-py3-none-any.whl
Size 65.2 kB
Tags Python 3
SHA-256 checksum
How to use checksums
0fe572f3ff409e8872cd333b0841ef66d49a57815c094825bf4f31554494072c
BLAKE2b-256 checksum
How to use checksums
030d0afe1426f242b5a5ff31233ab3026540ab69dfd09ee01059ea67ec849b30
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.12.10

Release history Release notifications | RSS feed

This release

1.0.1 This release

2 release files

1.0.0

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page