quantlib-irs-toolkit

Minimal, opinionated QuantLib helpers to bootstrap interest-rate swap curves from market par rates, with sensible conventions for the EM and frontier currencies most often missing from generic tutorials.

Currently supported out of the box:

• PLN (WIBOR 3M)
• HUF (BUBOR 3M)
• CZK (PRIBOR 3M)
• ZAR (JIBAR 3M, quarterly fixed)

Adding a new currency is a single dict entry; see Adding a currency.

Install

QuantLib's Python bindings are the only dependency.

pip install quantlib-irs-toolkit

For local development:

git clone https://github.com/fawraw/quantlib-irs-toolkit.git
cd quantlib-irs-toolkit
pip install -e ".[test]"
pytest

QuantLib wheels are available on PyPI for the major platforms. On macOS Apple Silicon you may need to install QuantLib via brew install quantlib first.

Quick start

from quantlib_irs_toolkit import build_irs_curve

result = build_irs_curve(
    ccy="PLN",
    rates={
        "1Y": 4.235,
        "2Y": 4.150,
        "3Y": 4.100,
        "5Y": 4.050,
        "7Y": 4.100,
        "10Y": 4.200,
        "15Y": 4.350,
    },
)

print(result["zero_rates"])
# {'ON': 4.187..., '1W': 4.187..., '1M': 4.190..., ..., '15Y': 4.391...}

print(result["forward_rates"])
# {'1y1y': 4.067..., '5y5y': 4.231..., ...}

rates are quoted in percent: 4.235 means 4.235 %, not 0.04235.

What you get back

{
    "currency":         "PLN",
    "curve_date":       "2026-05-13",
    "settlement_date":  "20260515",
    "conventions": {
        "day_count":       "ACT/365",
        "fixed_frequency": "Annual",
        "index":           "WIBOR3M",
        "settlement_days": 2,
        "calendar":        "Poland",
    },
    "input_tenors": ["1Y", "2Y", "3Y", "5Y", "7Y", "10Y", "15Y"],
    "input_rates":  { ... },
    "discount_factors": { "ON": 0.9999..., "1M": 0.9965..., ..., "15Y": 0.5235... },
    "zero_rates":       { "ON": 4.187, "1M": 4.190, ..., "15Y": 4.391 },
    "forward_rates":    { "1y1y": 4.067, "2y1y": 4.055, ..., "5y5y": 4.231 },
    "error":            None,
}

If something goes wrong (bad currency, missing rates, unbootstrappable inputs), error is a human-readable message and the other fields may be absent.

Conventions per currency

The conventions table (CCY_CONVENTIONS in curves.py) is the source of truth. If you disagree with a value (eg. your desk uses a different settlement period for ZAR), patch it at runtime:

from quantlib_irs_toolkit import CCY_CONVENTIONS
CCY_CONVENTIONS["ZAR"]["settlement_days"] = 2   # T+2 instead of T+0

Adding a currency

from quantlib_irs_toolkit import CCY_CONVENTIONS, build_irs_curve

CCY_CONVENTIONS["TRY"] = {
    "day_count":        "ACT/360",
    "fixed_frequency":  "Quarterly",
    "index_name":       "TRLIBOR3M",
    "index_tenor":      3,
    "settlement_days":  2,
    "calendar":         "TARGET",   # use TARGET or WeekendsOnly if QL has no native calendar
}

result = build_irs_curve("TRY", { "1Y": 45.0, "2Y": 40.0, "5Y": 32.0 })

Supported calendar names (from _ql_calendar): Poland, Hungary, CzechRepublic, SouthAfrica, TARGET, WeekendsOnly. Extend the mapping in curves.py if you need more.

Output tenors

By default, the toolkit returns:

• Discount factors and zero rates at: ON, 1W, 1M, 3M, 6M, 1Y, 2Y, 3Y, 4Y, 5Y, 6Y, 7Y, 8Y, 9Y, 10Y, 12Y, 15Y
• Forward rates at: 1y1y, 2y1y, 3y1y, 5y1y, 1y4y, 5y5y, 2y3y, 3y2y

Both sets are exposed as tuples (OUTPUT_TENORS, FORWARD_PAIRS) and can be overridden by copying and editing curves.py.

Method

The bootstrapping uses QuantLib's PiecewiseFlatForward with SwapRateHelper instances for each input tenor. Extrapolation is enabled by default so queries slightly past the longest swap don't raise; pass extrapolate=False to build_irs_curve if you'd rather fail loudly.

The IBOR index is built generically (ql.IborIndex) rather than relying on QuantLib's built-in Euribor-style classes, because the EM indices (WIBOR, BUBOR, PRIBOR, JIBAR) are not all natively shipped in QuantLib.

Not in scope

• Multi-curve pricing (separate discount and projection curves)
• OIS-discounted swap pricing
• Cross-currency basis
• Sensitivity / bucketed risk

These are deliberately out of scope for v0. If you need them, this code is a reasonable starting point but you'll want to extend the helpers and the curve construction.

License

MIT. See LICENSE.