dsgekit

A Python toolkit for Dynamic Stochastic General Equilibrium (DSGE) modeling.

Status

• Version: 0.1.0rc1 (release candidate pre-release)
• API stability: evolving
• Minimum Python version: 3.13 (from pyproject.toml)

What Is Included

• Model loading from .mod, YAML, and Python dict/API sources
• Linear workflow: linearization, QZ/BK solving, IRFs, simulation, moments, FEVD
• State-space pipeline: Kalman filter, RTS smoother, forecasting, historical decomposition
• Estimation: MLE, MAP, MCMC (with diagnostics), marginal data density
• Nonlinear solvers: first-, second-, and third-order perturbation (with pruning)
• Deterministic workflows: perfect foresight and one-constraint occbin-lite
• Policy and welfare tooling: OSR sweeps, Ramsey/discretionary LQ, welfare comparisons
• Baseline regression harness and CI workflow for compatibility checks

Installation

From source (recommended at this stage)

git clone https://github.com/gustavovargas/dsgekit.git
cd dsgekit
python3.13 -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

Optional extras

pip install -e ".[plot]"   # plotting
pip install -e ".[sym]"    # symbolic derivatives
pip install -e ".[speed]"  # numba acceleration
pip install -e ".[all]"    # all optional extras

Quick Start

from dsgekit import load_model
from dsgekit.simulate import irf
from dsgekit.solvers import solve_linear
from dsgekit.transforms import linearize

model, cal, ss = load_model("tests/fixtures/models/ar1.yaml")
solution = solve_linear(linearize(model, ss, cal))
responses = irf(solution, "e", periods=20)

print(responses.data.head())

CLI Overview

dsgekit info model.mod
dsgekit solve model.mod
dsgekit irf model.mod -s e -p 20
dsgekit simulate model.mod -n 200 --seed 42
dsgekit estimation model.mod --method mle --params rho
dsgekit forecast model.mod -p 12
dsgekit decompose model.mod
dsgekit osr model.mod --grid phi_pi=1.0:3.0:9 --loss pi=1.0 --loss y=0.5
dsgekit baseline_regression --baselines tests/fixtures/baselines

Documentation Map

• docs/quickstart.md
• docs/formats.md
• docs/first_order_perturbation.md
• docs/second_order_perturbation.md
• docs/third_order_perturbation.md
• docs/perfect_foresight.md
• docs/occbin_lite.md
• docs/forecast_decomposition.md
• docs/osr.md
• docs/ramsey_lq.md
• docs/discretion_lq.md
• docs/welfare.md
• docs/performance.md
• docs/mod_workflow_migration.md

Full index: docs/index.md

Examples

python examples/ar1_pipeline.py
python examples/nk_forward_looking.py
python examples/nonlinear_girf.py
python examples/occbin_zlb_toy.py
python examples/reference_model_zoo.py

Contributing and Security

• Contribution guide: CONTRIBUTING.md
• Code of conduct: CODE_OF_CONDUCT.md
• Security policy: SECURITY.md

Legal

• dsgekit is an independent project and is not affiliated with third-party tool maintainers.
• Implementation provenance and attribution policy: docs/legal_and_attribution.md.
• Naming policy for third-party references: docs/brand_naming_policy.md.
• Dependency notices: THIRD_PARTY_NOTICES.md.

License

Apache-2.0