fundcloud 0.3.0

Portfolio research, end-to-end, with a Rust core.

Fundcloud

Portfolio research, end-to-end, with a Rust core.

Fundcloud is a beginner-friendly, headless-for-advanced portfolio research framework. One install replaces the usual stack of quantstats + pyportfolioopt + skfolio + vectorbt + xlsxwriter, with a coherent .fc pandas surface for beginners and a full sklearn-compatible estimator API for advanced users. Matrix-heavy math lives in a Rust core via PyO3 and ships as a single abi3 wheel per platform.

Install

uv add fundcloud                  # core
uv add "fundcloud[data]"          # + all network data providers (yf, fmp, av, binance)
uv add "fundcloud[pf,ta,data]"    # + skfolio + TA-Lib + data sources
uv add "fundcloud[all]"           # everything

Extra	Adds
pf	skfolio — portfolio optimisation
ta	TA-Lib — 170+ technical indicators
data-yf / data-fmp / data-av / data-bn	individual data providers
data	bundle of every data provider above
viz	matplotlib + kaleido (static plot export)
reports	WeasyPrint (PDF) + XlsxWriter (Excel)
all	everything above

Exploratory data analysis (fundcloud.explore.{profile, compare, quickview}) ships in core — no extra needed.

Quickstart (60 seconds)

import pandas as pd
import fundcloud  # registers the .fc accessor on pandas

# Any returns Series gets instant analytics
returns = pd.Series([0.012, -0.005, 0.008, -0.010, 0.015], name="strategy")
returns.fc.sharpe(periods=252)           # annualised Sharpe
returns.fc.max_drawdown()
returns.fc.drawdown_series()

# Purged CV that plugs into sklearn out of the box
from fundcloud.validate import PurgedKFold
from sklearn.model_selection import cross_val_score

cv = PurgedKFold(n_splits=5, purge=3, embargo=1)
# cross_val_score(estimator, X, y, cv=cv)   # drop-in

The library ships DCA/Hold strategies, a simulator, skfolio-backed optimisers, native EDA, and HTML/PDF/Excel tear sheets out of the box. Prefer one composed figure over a full report? fundcloud.plots.summary(returns) returns a multi-panel plotly.graph_objects.Figure (cumulative, drawdown, rolling Sharpe, distribution, monthly heatmap) with Plotly theme support via fc.set_theme("dark") (re-exported at the top level for import fundcloud as fc); every builder also accepts multi-asset DataFrames so comparisons stay one line.

sklearn & skfolio interop

Every Fundcloud estimator, transformer, and CV splitter passes sklearn.utils.estimator_checks.check_estimator and round-trips through skfolio. Example:

from sklearn.pipeline import Pipeline
from sklearn.model_selection import GridSearchCV
from fundcloud.features import FeaturePipeline
from fundcloud.features.indicators import RSI, SMA
from fundcloud.optimize import MeanRisk, RiskMeasure
from fundcloud.validate import PurgedKFold

pipe = Pipeline([
    ("features", FeaturePipeline([("rsi", RSI(timeperiod=14)), ("sma", SMA(timeperiod=20))])),
    ("optim",    MeanRisk(risk_measure=RiskMeasure.CVAR)),
])
search = GridSearchCV(pipe, param_grid={"optim__min_weights": [0.0, 0.02, 0.05]},
                     cv=PurgedKFold(n_splits=5, purge=3))
search.fit(returns_panel)

Architecture

 ┌────────────────────────────────────────────────────────────────────┐
 │  End-user surfaces:  fluent accessor .fc   |   estimator API       │
 ├───────────────┬──────────────┬───────────────┬──────────────────────┤
 │   reports     │   explore    │    plots      │    datasets          │
 ├───────────────┴──────────────┴───────────────┴──────────────────────┤
 │                       metrics │ validate │ optimize                 │
 ├──────────────────────────┬──────────────────────────────────────────┤
 │      portfolio           │                    sim                   │
 ├──────────────────────────┼──────────────────────────────────────────┤
 │       strategies         │                features                  │
 ├──────────────────────────┴──────────────────────────────────────────┤
 │                              data                                   │
 │  Backends (YF, FMP, …, Parquet, DuckDB, Memory, CSV) ─ Catalog      │
 ├─────────────────────────────────────────────────────────────────────┤
 │                   kernels  (Rust, PyO3, abi3)                       │
 └─────────────────────────────────────────────────────────────────────┘

Python compatibility

Supported on Python 3.10, 3.11, 3.12, 3.13, 3.14. Wheels are built with PyO3's abi3-py310 feature, so one wheel per platform works across every supported version.

Acknowledgments

Fundcloud stands on the shoulders of excellent open-source work:

• scikit-learn (BSD-3-Clause) — estimator, transformer, and CV-splitter contracts used throughout.
• skfolio (BSD-3-Clause) — portfolio optimisation algorithms; Portfolio/Population objects. Install with uv add 'fundcloud[pf]'.
• quantstats (Apache-2.0) — inspiration for our tear-sheet and pandas-accessor design.
• vectorbt (Apache-2.0) and vectorbt.pro — inspiration for the vectorised simulation model.
• TA-Lib / ta-lib-python (BSD-2-Clause) — all 170+ technical indicators in fundcloud.features.indicators.
• PyO3, rust-numpy, maturin, uv — the build-and-ship story.

See NOTICE for the full attribution.

Contributing

Read CONTRIBUTING.md. TL;DR: uv sync, uv run pytest, cargo test --workspace, add a test, open a PR.

License

MIT.