Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Unreleased

Tournament-based LLM evaluation engine with multi-round elimination
Knowledge Bank for preserving insights from eliminated models
Rubric-based scoring with configurable weights
YAML workflow system for custom AI pipelines
CLI interface with certamen command (tournament and workflow modes)
Python API via Certamen.from_settings()
Support for 19 LLM providers via LiteLLM (OpenAI, Anthropic, Google, Ollama, etc.)
Provenance tracking and cost accounting
Auto-discovery of Ollama local models
SQLite response cache for deduplication
Clean Architecture with import-linter enforcement
Comprehensive pre-commit hooks (14 layers of checks)
Cross-platform CI (Linux, macOS, Windows) with Python 3.10-3.13
Mutation testing on domain layer
PEP 561 type stubs (py.typed marker)