Home
finasys
From raw market data to ML-ready features in five lines of code.
finasys is a toolkit for financial data processing — not manual wrangling — for ML pipelines and AI agents. It lets you go from raw market data to production-ready features in a few lines of code, whether you're building trading models, running portfolio analysis, or powering financial AI agents.
finasys is Polars-first — every indicator and feature runs as a native Polars expression, making it 10-100x faster than pandas-based alternatives with zero C dependencies (no ta-lib build headaches). It supports 37+ international markets, crypto, forex, commodities, and macro indicators out of the box.
Quick Start¶
import finasys as fs
df = fs.load("AAPL", start="2024-01-01")
df = fs.features.add_all(df)
print(fs.agents.summarize(df))
Install¶
Key Modules¶
| Module | What it does |
|---|---|
fs.load() |
Load from Yahoo Finance, CSV, Parquet with auto-caching |
fs.features |
15+ indicators, returns, rolling stats, lags, calendar, cross-sectional |
fs.features (targets) |
Forward returns, classification labels, triple-barrier labeling |
fs.features (distributions) |
Rolling skewness, kurtosis, tail ratio, Jarque-Bera |
fs.stats |
Sharpe, Sortino, Calmar, VaR, CVaR, alpha/beta, drawdown duration |
fs.profiler |
One-call data profiling: quality checks, distribution analysis |
fs.agents |
LLM summaries, OpenAI tools, RAG context, LangChain integration |
Why finasys?¶
- Polars-first -- 10-100x faster than pandas-ta, zero C dependencies
- Financial-native -- every function understands OHLCV, ticks, fundamentals
- ML-ready -- target engineering, risk metrics, and data profiling built in
- Agent-ready -- structured outputs designed for LLM consumption
- Symbol-aware -- multi-symbol DataFrames work correctly out of the box