chore: release v0.2.0

Bundles up the multi-chain runner, discrete distributions, HMC gradient
checker, first-class warmup workflow, serde-serializable diagnostics, the
gamma log-probability accuracy fix, and the input-validation hardening
landed since 0.1.0. All changes since 0.1.0 are additive — trait additions
ship with default impls — so existing 0.1.x users upgrade without source
changes.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: kholdrex

CHANGELOG.md

@@ -5,6 +5,38 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.2.0] - 2026-06-06
+
+### Added
+- Discrete distributions module with discrete sampling primitives.
+- Hamiltonian Monte Carlo gradient checker with `finite_difference_gradient` and
+  `gradient_check` helpers for validating user-supplied gradients.
+- Multi-chain runner (`multi_chain::run_multiple_chains` and `MultiChainOutput`)
+  that drives several seeded samplers through a shared warmup/sample schedule
+  and returns combined diagnostics.
+- First-class warmup/burn-in workflow on the `Sampler` trait: `run_with_warmup`,
+  `sample_with_warmup`, `WarmupMetadata`, and `WarmupRun`. Sampler statistics
+  are reset between the warmup and retained phases.
+- Optional `serde` support for `McmcDiagnostics`, `McmcDiagnosticSummary`,
+  `ParameterDiagnosticSummary`, `TracePlot`, `WarmupMetadata`, `WarmupRun`, and
+  `MultiChainOutput`, gated behind the `serde` feature.
+- Normal-normal and additional conjugate-model examples
+  (`examples/conjugate_models.rs`, `examples/serde_diagnostics.rs`,
+  `examples/discrete_distributions.rs`) plus a conjugate-models test suite.
+- MCSE summaries and richer per-parameter diagnostic output.
+
+### Changed
+- Hardened sampler input validation across Metropolis-Hastings, Gibbs, and HMC
+  constructors to reject non-finite or mis-sized inputs earlier.
+- Tightened gradient-check input validation (empty or non-finite points are
+  rejected before evaluating the analytic gradient).
+
+### Fixed
+- Improved numerical accuracy of the Gamma log-probability computation; locked
+  with a regression test.
+
+[0.2.0]: https://github.com/SyntaxSpirits/bayes-rs/releases/tag/v0.2.0
+
 ## [0.1.0] - 2025-07-03
 
 ### Added

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "bayes-rs"
-version = "0.1.0"
+version = "0.2.0"
 edition = "2021"
 authors = ["Alex Kholodniak <alexandrkholodniak@gmail.com>"]
 description = "A Rust library for Bayesian inference with MCMC samplers"

README.md

@@ -21,14 +21,14 @@ Add this to your `Cargo.toml`:
 
 ```toml
 [dependencies]
-bayes-rs = "0.1.0"
+bayes-rs = "0.2.0"
 ```
 
 Enable the optional `serde` feature when you want to serialize user-facing MCMC output:
 
 ```toml
 [dependencies]
-bayes-rs = { version = "0.1.0", features = ["serde"] }
+bayes-rs = { version = "0.2.0", features = ["serde"] }
 ```
 
 ### Simple Example