wash-alpha 0.55.1

Cross-account wash sale detection for stocks, options, and ETFs — local-first, IRS Pub 550 rules, with a tax-harvest planner and pre-trade simulator.

Cross-account wash sale detection for stocks, options, and ETFs — local-first, IRS Pub 550 rules, with a tax-harvest planner and pre-trade simulator.

Overview · Quickstart · Service management · Features · Usage · How the rules work · Architecture

---

Overview

When you trade across multiple brokerages, each platform tracks wash sales only within its own ecosystem. A loss sale on Schwab can be silently neutralized by a repurchase on Fidelity, and you won't find out until tax season — long after the window to act has closed.

The problem compounds when you trade options alongside the underlying, or ETFs that track the same index.

net-alpha is a local-first Python tool that gives you a single, unified view of your wash sale exposure across every account, asset class, and tax year — before it's too late to act.

[!NOTE] The package is published to PyPI as wash-alpha but the CLI command is net-alpha. Both names refer to the same project.

Quickstart

net-alpha requires Python 3.11+ and uv.

# Install with the local web UI
uv tool install 'wash-alpha[ui]'

# One-time: install the always-on background service
net-alpha service install

# Open the dashboard (browser opens automatically)
net-alpha ui

The service runs locally at http://127.0.0.1:8765. It refreshes prices every 4 hours and re-runs forward-looking wash-sale + §1091 detection daily. Stop it any time with net-alpha service stop (data at ~/.net_alpha/ is left untouched).

Prefer a one-shot launch without the background service? net-alpha ui still works in ephemeral mode (Ctrl-C to stop). The service is opt-in.

Prefer the terminal? The CLI works without UI extras:

uv tool install wash-alpha
net-alpha schwab.csv --account personal --detail

Service management

Command	Effect
net-alpha service install	Provision the runtime venv at ~/.net_alpha/venv/, write the launchd plist + sandbox profile, and (re)load the agent. Idempotent — safe to re-run.
net-alpha service start	Start (or reload) the service
net-alpha service stop	Stop the service. Survives reboots until start
net-alpha service restart	Restart
net-alpha service pause	Freeze background jobs but keep the dashboard reachable
net-alpha service resume	Unfreeze jobs
net-alpha service status	Health report (also --json for scripting)
net-alpha service logs -f	Tail the service log
net-alpha service uninstall	Remove plist, sandbox profile, and runtime venv. Your data at ~/.net_alpha/net_alpha.db is preserved.

The dashboard at /settings/service exposes the same controls in a UI surface, plus recent-runs history and a status pill in the site header.

What install does

Everything the running service needs lives inside ~/.net_alpha/, the only path the sandbox profile allows the service to write to:

• ~/.net_alpha/venv/ — a dedicated runtime venv (homebrew Python 3.11 via uv) with wash-alpha installed from your local project source.
• ~/.net_alpha/bin/net-alpha-wrap — the launchd wrapper script.
• ~/.net_alpha/run/sandbox.sb — the sandbox-exec profile applied to the running service.
• ~/Library/LaunchAgents/com.netalpha.service.plist — the LaunchAgent registration.

The runtime venv intentionally lives outside ~/Documents/. Under launchd, the service runs with a TCC identity that can't read user-data folders, so the project's own .venv/ is unreadable; ~/.net_alpha/ is TCC-clear.

Re-running install is safe — it bootouts a stale plist before re-bootstrapping and reinstalls wash-alpha with uv pip install --reinstall-package wash-alpha (deps stay cached, so the refresh is fast).

Re-installing after code changes

The runtime venv is a non-editable snapshot of your project source at install time. To make the running service reflect new code or pyproject.toml changes, re-run net-alpha service install.

For active development, run net-alpha … directly from your project's editable .venv — the LaunchAgent is for the always-on background service, not for the dev edit-run loop.

Features

• Cross-account intelligence — match a loss sale on one broker against a repurchase on another in a single pass. The whole point of the tool.
• Bundled broker parsers — Schwab transactions and Schwab Realized G/L are supported out of the box. Other brokers can be added by contributing a parser.
• Local web UI — Portfolio · Positions · Tax · Sim · Imports · per-ticker drill-down, all rendered server-side via HTMX. No Node, no npm, no CDN at runtime, dependencies vendored.
• Pre-trade simulation — net-alpha sim TSLA 10 --price 180 shows FIFO lot consumption, realized P&L, and a cross-account wash-sale verdict per account before you execute. The Sim page surfaces one-click suggestion chips (largest unrealized loss, wash-sale risk, largest unrealized gain).
• Tax-harvest planner — a forward-looking, plan-builder assistant turns the harvest queue into a ranked, editable plan (greedy by tax saved, capped by §1211's $3,000 ordinary-loss limit). Honors user-declared position targets so it never closes something you want to keep.
• §1256 awareness — index options (SPX, NDX, RUT, VIX, etc.) are recognized as §1256 contracts: wash-sale-exempt with statutory 60/40 LT/ST classification.
• Auditable explanations — every wash-sale flag includes rule citation, source trades, match reason, math, and confidence reasoning — inline in the UI or via --detail on the CLI. Per-symbol reconciliation cross-checks computed P&L against your broker's Realized G/L file.
• After-tax performance — the Tax → Performance view shows realized P&L after estimated taxes, with a tax-drag breakdown and an ST/LT/§1256 mix bar.
• Manual trade CRUD — add, edit, transfer, or delete trades by hand from the web UI; wash sales recompute automatically over the affected window.
• 100% local & zero-knowledge — your trade data, accounts, and P&L never leave the box. Symbols are sent to Yahoo Finance for live quotes only; disable with prices.enable_remote: false.

Usage

Local web UI

net-alpha ui [--port 8765] [--no-browser] [--reload]

Drop CSVs onto the dashboard, drill into any ticker, build a harvest plan, hand-enter trades, run pre-trade sims with suggested chips. Server-side HTMX, dies on Ctrl-C.

Import + check from the CLI

net-alpha schwab.csv --account personal

Imports the CSV, recomputes wash sales over the affected window, and prints a watch list plus a YTD impact summary. Add --detail for a per-violation breakdown.

Pre-trade simulation

net-alpha sim TSLA 10 --price 180

Lists every account that holds the ticker, with FIFO lot consumption, realized P&L, and a cross-account wash-sale verdict per account.

Manage past imports

net-alpha imports                # list
net-alpha imports rm 3 --yes     # remove (recomputes wash sales)

Migrate from v0.x

net-alpha migrate-from-v1 --yes

Reads ~/.net_alpha/net_alpha.db (v1 schema) and writes a fresh v2 DB at ~/.net_alpha/net_alpha.db.v2. Helper exists only on the v2.0.x line.

How the rules work

net-alpha strictly follows IRS Publication 550. A wash sale is triggered when you sell a security at a loss and buy a substantially identical security within 30 days before or after the sale.

Asset type	Scenario	Confidence
Equities	Sold ticker X at a loss, bought ticker X within 30 days	🟢 Confirmed
Options	Sold option at a loss, bought same option (exact strike + expiry)	🟢 Confirmed
Options	Sold option at a loss, bought option on the same underlying	🟡 Probable
ETFs	Sold ETF at a loss, bought the exact same ETF ticker	🟢 Confirmed
ETFs	Sold ETF at a loss, bought ETF tracking the same index (e.g., SPY → VOO)	🔵 Unclear

Bundled "substantially identical" pairs cover the major index-tracking ETFs (S&P 500: SPY/VOO/IVV/SPLG, Nasdaq-100: QQQ/QQQM, etc.). Extend with your own pairs by editing ~/.net_alpha/etf_pairs.yaml — your file adds to the bundled defaults, never replaces them.

Adding a broker

Contribute a parser at src/net_alpha/brokers/<name>.py — implement the BrokerParser Protocol and register it in brokers/registry.py. Realized G/L parsers go in src/net_alpha/audit/brokers/.

Architecture

CSV → BrokerParser → Trade (Pydantic) ──► SQLite (~/.net_alpha/net_alpha.db)
                                              │
                                              ├─► Wash-sale engine  (incremental ±30-day window)
                                              ├─► §1256 classifier  (60/40 LT/ST split)
                                              ├─► Reconciliation    (against broker Realized G/L)
                                              └─► Portfolio / planner / pricing  (pure functions)
                                                                │
                                                  ┌─────────────┴─────────────┐
                                                  ▼                           ▼
                                            Typer CLI                 FastAPI + HTMX UI

• Stack — Python 3.11+, pydantic v2, sqlmodel over SQLite, typer CLI, optional fastapi + Jinja + HTMX UI, ruff, pytest. Managed with uv.
• Storage — single SQLite DB at ~/.net_alpha/net_alpha.db; cross-year window detection works because all trades live in one place. Schema versioned via a meta table; hand-written ALTER TABLE migrations.
• Engine — wash-sale recompute is incremental, only the ±30-day window around affected trade dates is recalculated on import or removal. Every emitted violation is reproducible from source trades.
• Web UI — optional subpackage at src/net_alpha/web/. Calls only existing public seams (Repository, engine functions, audit/portfolio/planner pure functions); no business logic in web/. Static assets vendored — no CDN at runtime, no Node toolchain.

Development

uv sync --extra ui --extra dev      # install everything
uv run pytest                       # tests (or: make test)
uv run ruff check .                 # lint
uv run ruff format .                # format
make build-css                      # rebuild the Tailwind bundle

Disclaimer

[!IMPORTANT] This tool is for informational purposes only and does not constitute tax or legal advice. Wash sale rules — especially around options and ETFs — involve unsettled areas of tax law. Anything labeled Probable or Unclear should be reviewed with a qualified CPA before making filing decisions.

_{If net-alpha saves you a CPA call this tax season, consider starring the repo — it's the best signal for other traders looking for a tool they can trust with their tax data.}