Skip to content

alpha-visualizer v0.2.0 — Try the full web visualizer in one command

Released: May 11, 2026 / Version: v0.2.0 / Distribution: PyPI · GitHub Release

alpha-visualizer is a standalone OSS package that renders backtest results produced by forge in a web browser. v0.2.0 focuses on making the very first run feel like a finished product — you can see every screen end-to-end without producing any data of your own, alongside large security and quality improvements.

Highlights

1. A fully synthetic sample dataset is now bundled (--use-bundled-samples)

Until v0.1.x, alpha-visualizer needed actual forge output to show anything meaningful — the screens were empty on a fresh install. From v0.2.0 onward, a legally redistribution-free synthetic dataset ships inside the wheel, so you can explore every feature with a single command:

pip install alpha-visualizer==0.2.0
vis serve --use-bundled-samples --no-open
# Open http://127.0.0.1:8000

The bundle contains 5 years (1250 business days) of synthetic OHLCV across 8 textbook strategies, producing 40 backtests + 2 WFO runs + 2 Grid search runs + 5 sample idea memos.

Symbol Character Strategy fit
EQUITY_SYNTH 2008-style crash → recovery → mild correction Trend-following strategies shine
INDEX_SYNTH Long calm → 2020-style V-shaped recovery Composite filter strategies work well
COMMODITY_SYNTH Sideways → spike → blow-off → slow bleed Breakout strategies dominate
FX_SYNTH Mean-reverting throughout (AR(1)) Reversion strategies prevail
CRYPTO_SYNTH Bubble → crash × 2 + flash crash → recovery Trend-following + breakout

The strategy × symbol compatibility matrix is intentionally distributed so that the Browse table sort/filter, the Compare heatmap, and the WFO IS/OOS stability chart all look meaningful out of the box.

Prices are synthesized from Geometric Brownian Motion + Poisson jumps + AR(1). All symbol names carry the _SYNTH suffix so they cannot be confused with real tickers. The regeneration script samples/build_samples.py is deterministic, and CI runs git diff --exit-code on every PR to keep the bundle byte-identical.

The bundled strategies are textbook indicator combinations only — SMA / RSI / MACD / Bollinger / ADX / Donchian. Proprietary differentiators in forge itself, such as HMM regime detection and MTF-optimized parameters, are intentionally excluded.

2. Live-performance view (Detail screen)

A new Live tab on the Detail screen reads data/live/ summaries and trades and renders a period-aligned diff against the corresponding backtest (#57). If your backtest looked great but the live deployment is bleeding edge, the divergence shows up immediately. Period misalignment is normalized automatically.

3. HMM / regime overlays on the equity chart

The Equity Chart can now render HMM states (high-vol / mid-vol / low-vol, etc.) as color bands behind the equity curve (#56). The Risk tab gained a per-regime summary card too, breaking down trade counts, win rate, and average return for each regime in a single view. This pairs especially well with forge-side strategies that infer HMM regimes.

4. Security and quality lift

  • 14 CodeQL alerts resolved (#175): path-injection, log-injection, and SSRF detectors all clear.
  • Distinguished "resource not found (404)" from "DB failure (500)" — responses on optimization_runs table absence or SQLAlchemy OperationalError now correctly return 500 instead of masquerading as 404 (#106).
  • Fixed Browse-screen crash on strategies with latest_* fields undefined (#23).
  • Caught up with forge's default DB filename change to backtest_results.db (#177).
  • Non-finite best_metric values are now returned as null and rendered as — in the UI (#172).

5. CI and developer experience

  • Lighthouse CI continuously measures Performance / Accessibility / Best-Practices on every PR (#136, #162).
  • E2E fixture drift check and OSS sample-forge drift check enforce that test fixtures and the bundled sample regenerate byte-identically — drift fails the PR (#161, #178).
  • Major frontend dependencies bumped to React 19, react-router-dom 7, Storybook 10, Vite 8, etc.

Upgrading

The package is on PyPI now:

# pip
pip install -U alpha-visualizer

# uv
uv add alpha-visualizer@latest         # add to a project
uv tool install alpha-visualizer        # install as a CLI tool

The easiest sanity check is --use-bundled-samples. To point at your own forge project, use the existing flag:

vis serve --forge-dir /path/to/your/alpha-strategies

No forge.yaml schema changes — existing forge projects continue to work unchanged.

Please report bugs and feature requests through GitHub Issues.