Talyn is a first-principles simulation kernel for probabilistic modeling and inference, designed for clarity, extensibility, and mathematical rigor.

Talyn is a simulation-native probabilistic programming system. It empowers you to define, simulate, and analyze complex probabilistic models using a trace-first, execution-based approach—no abstraction layers or symbolic graphs. Talyn is designed for researchers, educators, and engineers who want full transparency and control over every probabilistic computation.

• Simulation-based inference (no static graphs, pure execution)
• Transparent execution traces for every run
• Supports: Monte Carlo, importance sampling, belief propagation, particle filtering, and more
• Composable, dynamic model definitions (YAML or Python DSL)
• CLI and Python API for flexible workflows
• Extensible: add new distributions or inference methods easily
• Academic-grade testing and validation suite

pip install talyn # or for local development pip install -e .

Minimal CLI example:

talyn simulate coin.yaml

• talyn/ — Core library (sampler, trace, inference, etc.)
• docs/ — Documentation and Jupyter notebooks
• tests/ — Unit and regression tests
• examples/ — Example models and scripts

See NOTEBOOKS.md for a categorized list of Jupyter notebooks, including tutorials, validation, and benchmarks.

This project is licensed under the MIT License.

For academic use, please cite using the BibTeX entry in CITATION.md.