repository·master·Indexed 11 days ago
https://github.com/goldmansachs/gs-quantA Python-based quantitative finance toolkit developed by Goldman Sachs for building trading strategies, managing risk, and performing derivative analysis. It provides comprehensive modules for pricing and risk across Rates, FX, Equities, and Credit, as well as tools for portfolio management, backtesting engines, factor models, and index analysis via Marquee.
GS Quant is a Python toolkit for quantitative finance designed to accelerate the development of quantitative trading strategies and risk management solutions. It is used for:
The gs_quant repository contains a collection of research, tactical case studies, and analytical notebooks. You can use these to learn how to implement specific financial workflows using the library.
Key areas of research include:
The GS Quant documentation is organized into functional modules, each containing Jupyter notebooks that serve as practical examples for specific tasks. You can find examples for:
00_data): Querying datasets, managing data contexts, retrieving volatility surfaces, and using screeners.01_markets): Working with securities and managing relative dates/business day calendars.02_pricing_and_risk):The gs_quant documentation is structured into two primary types of resources to help you learn the library:
The Portfolio class groups instruments for pricing, resolution, and analysis as a single unit. Portfolios can be created from lists or dictionaries (where keys are instrument names) and can be nested.
{name: instrument}.Portfolio can contain other Portfolio objects.portfolio[0]) or by name (portfolio['name']).portfolio.all_instruments returns all instruments across all nested portfolios.portfolio.resolve() resolves all instruments within the portfolio in place.portfolio.calc(measure) calculates risk measures (e.g., DollarPrice, IRDelta) for the entire portfolio.from gs_quant.instrument import IRSwap, IRSwaption
from gs_quant.markets.portfolio import Portfolio
from gs_quant.risk import DollarPrice, IRDelta
# Creating from a list
swap = IRSwap('Pay', '10y', 'USD', name='USD 10y Payer')
swaption = IRSwaption('Receive', '10y', 'EUR', expiration_date='1y', name='EUR 1y10y Receiver')
portfolio = Portfolio([swap, swaption], name='My Portfolio')
# Creating from a dictionary
portfolio = Portfolio(
{
'USD 10y Payer': IRSwap('Pay', '10y', 'USD'),
'EUR 5y Receiver': IRSwap('Receive', '5y', 'EUR'),
}
)
# Nesting portfolios
book_a = Portfolio([IRSwap('Pay', '5y', 'USD')], name='Book A')
book_b = Portfolio([IRSwap('Pay', '5y', 'EUR')], name='Book B')
master = Portfolio([book_a, book_b], name='Master Book')
# Operations
master.resolve() # resolves all instruments in place
prices = master.calc(DollarPrice) # single risk measure
results = master.calc([DollarPrice, IRDelta]) # multiple risk measuresWhen you run the installer, the files are placed in the target .claude/skills/ directory as follows:
.claude/skills/<name> to the source directory inside the installed gs_quant package. Updating the gs_quant package automatically updates the skills.install command after upgrading gs_quant to refresh the skills.Warning: If a target directory already exists at the destination, it will be overwritten.
The backtesting framework is built on three core abstractions that compose to form a strategy:
Trigger objects.Action objects.A backtest Engine executes the strategy over a specified date range, resolving instruments, computing risks, and generating P&L time series.
from gs_quant.backtests.strategy import Strategy
from gs_quant.backtests.triggers import PeriodicTrigger
from gs_quant.backtests.actions import AddTradeAction
# Example composition
strategy = Strategy(None, PeriodicTrigger(trig_req, action))All new measure code must comply with strict formatting and naming standards. Enforcement is done via ruff.
snake_case (e.g., swap_rate)CamelCase (e.g., EventType)UPPER_SNAKE_CASE (e.g., CENTRAL_BANK_WATCH_START_DATE)_extract_series_from_df)import pandas as pd (never from pandas import ...)import numpy as npimport datetime as dtgs_quant.target.common (use gs_quant.common instead).pytz (use zoneinfo or datetime.timezone).stdlib $\rightarrow$ third-party $\rightarrow$ local. Alphabetize within groups."""). Must include :param, :return:, **Usage**, and **Examples** sections.Optional[X] and Union[X, Y] where appropriate.Run these before committing:
ruff check --fix
ruff formatThe server supports two authentication modes via the --auth flag:
local (Default)Best for single-user / desktop use. The server creates and authenticates one GsSession at startup using the provided --client-id and --client-secret (or environment variables). This single shared session is reused for every request made to the server.
python -m gs_quant.mcp server --auth local \
--client-id "$CLIENT_ID" --client-secret "$CLIENT_SECRET"passthroughBest for multi-user / remote deployments. No session is created at startup. Instead, the server inspects each incoming HTTP request to construct a per-user GsSession. It detects authentication from:
GSSSO cookie $\rightarrow$ GSSSO SSO tokenMarqueeLogin cookie $\rightarrow$ MARQUEE_LOGINAuthorization: Bearer <3-part JWT> $\rightarrow$ JWTAuthorization: Bearer <otherwise> $\rightarrow$ OAUTH access tokenpython -m gs_quant.mcp server --auth passthroughpython -m gs_quant.mcp server --auth local --client-id "$CLIENT_ID" --client-secret "$CLIENT_SECRET"Triggers determine when an action (like a trade) is executed during a backtest. Available trigger types include:
AND/OR logic.The 01_markets/02_rdates module provides tools for handling financial dates. Key capabilities include:
The backtesting framework allows for complex logic by combining multiple triggers:
AND or OR operations.NOT).