Trading Research Workbench

A lightweight local trading research environment — think simplified open-source TradingView.

Streamlit UI, Plotly candlestick charts, Backtrader behind an adapter, SQLite cache for OHLCV. Designed to be understandable, hackable, and AI-coding friendly.

What it does

• Pick a ticker (any symbol yfinance / Alpha Vantage supports — stocks, ETFs, indices, crypto).
• Download historical OHLCV bars, cached locally in SQLite.
• Toggle technical indicators (SMA, EMA, RSI, MACD, Bollinger, Momentum, Stochastic, ATR) — overlays on the price pane, oscillators in their own sub-panels.
• Run rule-based strategies (SMA Cross, RSI Reversion, MACD Cross) via Backtrader.
• See entry / exit triangles directly on the candles, plus return / Sharpe / drawdown / win-rate metrics and a full trade log.

Quick start

conda create -n equity-research python=3.11 -y
conda activate equity-research
conda install -c conda-forge ta-lib -y
pip install streamlit plotly backtrader yfinance scikit-learn pandas numpy requests

Then either:

• Double-click start.bat (Windows) — handles env activation and launches the app on http://localhost:8520
• Or manually: streamlit run app.py

Architecture

Small files, single-purpose, no deep abstractions.

File	Role
app.py	Streamlit entrypoint
ui.py	Sidebar + main panel layout
chart.py	Plotly candlestick figure builder
data_loader.py	Provider adapters (yfinance, Alpha Vantage) with caching
indicators.py	TA-Lib-backed indicator math
strategy.py	Backtrader strategy classes (rule-based)
backtest.py	Backtrader adapter — only file that imports backtrader
database.py	SQLite OHLCV cache
config.py	Config load/save
config.json	All defaults (tickers, periods, cash, commission, …)
smoke_test.py	Backend smoke check (no UI)

Backtrader sits behind backtest.run() returning a plain BacktestResult, so the engine can be swapped without touching anything else.

Configuration

Edit config.json. Indicator periods, strategy thresholds, default ticker/date range, backtest cash and commission all live there.

For Alpha Vantage, set the environment variable named in data.alpha_vantage_api_key_env (default ALPHAVANTAGE_API_KEY).

Project direction

See PROJECT.md for the full project philosophy, MVP milestones, and explicit non-goals.

The TL;DR:

• Goal: simplified, visual, hackable TradingView clone for local research.
• Not a goal: live trading, broker integration, microservices, anything you'd build for an HFT shop.

Status

First MVP milestone (visual candles + SMA cross backtest with markers on chart) is reachable from the UI.

Roadmap from PROJECT.md: parameter search, multi-symbol batches, vectorbt for sweeps, eventual ML extensions.