AutoBott v2

AutoBott v2 is being built as an automated options trading system for long calls and long puts.

The product is the trading bot itself: signal generation, risk controls, broker execution, position management, operator oversight, and auditability. Research, replay, backtests, and historical analysis are crucial support systems, but they are not the product.

Product Identity

AutoBott has five major layers:

• signal: regime, direction, volatility, and contract selection
• risk: trade eligibility, sizing, exposure limits, and kill switches
• execution: broker routing, order lifecycle, retries, and cancel/replace behavior
• positions: open-position monitoring and exits
• operator: dashboard, health, alerts, and administrative controls

Supporting systems exist to improve and audit the bot:

• capture: real market snapshot collection
• replay: deterministic campaign evaluation
• history: historical corpus generation and backtest-style analysis
• scorecards: edge review, drift review, and gate reporting

Today the repository is still weighted toward the support layers. The execution product is not complete yet.

Current State

The committed runtime currently provides:

• a decision-engine foundation
• paper-market capture and status plumbing
• replay, scorecard, and campaign analysis
• a Render-hosted operator console
• no committed live order-routing path yet

That means the repo contains important building blocks, but it is not yet the complete automated trading system described above.

Validation

Local paper dashboard, no PowerShell setup:

Double-click start_paper_dashboard.cmd

That launcher auto-loads C:\Users\flavo\Downloads\AutoBott.env, applies the local paper defaults, starts the dashboard on http://127.0.0.1:8000, and uses dashboard token autobott-local unless you override it in the env file.

Run the test suite:

pytest

Validate a captured market/options snapshot and optionally append the decision card to a JSONL ledger:

.\.venv\Scripts\python.exe -m autobott_v2.phase1_validate --snapshot .\path\to\real_snapshot.json --ledger .\data\learning_ledger.jsonl

The validator expects real captured inputs: market bars, option-chain quotes, SPY/QQQ/VIX context, event blackout flags, and IV history. It does not synthesize market data.

Architecture Docs

Read:

• docs/DOCTRINE.md
• docs/AUTOBOTT_V2_PURPOSE_LOCK.md
• docs/BUILD_PLAN.md
• docs/REPO_LANES.md
• docs/TRADESCOUT_DEFAULT_PROFILE.md