Decision SystemIBKR · READ ONLY简体中文
Read-only by design · Open source

Decisions
before orders.

A self-hosted IBKR decision-support system that turns account facts, portfolio risk and market context into auditable actions—without placing a trade.

Open synthetic demo View on GitHub

No broker connection. No real account data. No persistence.

DECISION / SIM-AREVIEW
Suggested actionReduce 3%

Concentration exceeds the personal risk budget while the position remains above its stop.

Risk score72/ 100
Position27.4%of equity
Evidence4sources
R1 concentrationTechnical intactManual execution
IBKR ACCOUNT TRUTH•RULES R1–R7•MANUAL EXECUTION•POINT-IN-TIME REVIEW
01 / SYSTEM

One evidence chain.
Three deliberate stages.

Built for the moment between seeing a number and acting on it.

01

Account truth

Read balances, positions, executions, PnL and margin from IBKR without exposing an order route.

02

Decision context

Combine portfolio risk, technical state, market regime and existing plans into explainable actions.

03

Review loop

Trace every recommendation, threshold, execution and outcome without rewriting the historical record.

02 / SAFETY BOUNDARY

The broker is a source.
Never an actuator.

Read-only mode is enforced at the API client and Gateway. The application contains no order, cancellation, exercise or transfer route. Adopting a recommendation creates a local draft only.

Read the safety model
01Dedicated broker userLeast-privilege account access
02Gateway read-only modeIB_READONLY=yes
03API read-only modereadonly=True
04No execution routesDecisions remain drafts
03 / RISK ENGINE

Rules that leave
an audit trail.

Each evaluation stores status, evidence, source batch and notification state.

R1Single-name concentrationPERSONAL LIMIT
R2Sector concentrationPERSONAL LIMIT
R3Margin usageZERO MARGIN
R4Repeated risk increasesPERSONAL LIMIT
R5Monthly activityPERSONAL LIMIT
R6Decision-level lossPERSONAL LIMIT
R7Suggested stop reachedSTOP QUALITY
04 / PUBLIC DEMO

Change the scenario.
Watch the decision change.

Explore three synthetic portfolio states and see how concentration, drawdown and liquidity alter the risk score and recommended action.

Launch demo
05 / OPEN SOURCE

Inspect the logic.
Own the data.

Python, FastAPI, SQLite and Docker Compose. MIT licensed, self-hosted, documented in English and Chinese, and designed around synthetic tests.

GitHub repository →中文 README →MIT License →