2.0.1

• Removed pre-v2 dead code (model.py, pyaugmecon.py, process_handler.py, queue_handler.py, solver_adapter.py, solver_process.py) superseded by the solver/ subpackage.
• Added conda-forge distribution. Install via conda install -c conda-forge pyaugmecon. Updated README with conda/mamba instructions and solver package mapping.

2.0.0

Breaking changes

• Moved the package to src/pyaugmecon and removed the options module. Use PyAugmeconConfig directly.
• Replaced the broad options object with one strict Pydantic config model. Unsupported setting names are rejected instead of being silently accepted.
• Kept the public result surface small: removed undocumented PyAugmecon attributes such as coverage_*, exact_guarantee, objective_order_*, nadir_source, grid_points_effective, and summary.
• Exposes advanced parallel controls as work_distribution and flag_policy.

Core behavior

• Added explicit exact and sampled run modes.
• Built epsilon levels deterministically and added objective ordering with objective_order="auto_range" or "given".
• Automatically deactivates all user objectives before solving.
• Checks for reserved internal component names before modifying the Pyomo model.
• Computes auto-safe nadir bounds when explicit nadir_points are not provided.
• Keeps decision-variable payloads optional with store_decision_variables.
• Normalizes objective keys deterministically with round_decimals. Workers also snap near-integer noise on objective values via objective_tolerance before sending results.

Parallel solve path

• Uses spawn-based multiprocessing with shared-memory model handoff.
• Splits work with work_distribution="dynamic", "fixed", or "outer_grid". The default "auto" picks "outer_grid" for exact multi-worker runs and "dynamic" otherwise.
• Uses flag_policy="local" or "shared". The default "auto" picks "shared" for exact multi-worker runs and "local" otherwise.
• Adds clearer worker failure, timeout, early-stop, job queue, result queue, and error queue handling.
• Workers emit structured Solution records instead of nested result maps.
• The worker hot loop caches Pyomo component handles and objective expressions once per worker to reduce repeated Pyomo lookups.
• Shared outer-grid skip state avoids repeat infeasible solves between workers; missed concurrent updates can only cost an extra solve, not a wrong answer.

Solver support

• Sets HiGHS as the default solver family.
• Adds fallback chains for HiGHS, Gurobi, CPLEX, XPRESS, CBC, and SCIP families.
• Adds solver_options, solver_io, and solve_warmstart controls.
• Handles solver statuses through a shared adapter so infeasible, infeasible-or-unbounded, time-limit, and unsupported terminations are reported consistently.

Logging and artifacts

• Replaced print output with tqdm progress and Loguru-backed run logging.
• Added log_sink= to forward run messages to objects with .info(message).
• Added artifact_folder, artifact_name, write_csv, progress_bar, log_to_console, and process_logging.
• Writes payoff, grid, and solution tables as CSV files through a small tabular helper.

Docs, examples, and packaging

• Solver backends are opt-in through extras: pyaugmecon[highs], pyaugmecon[gurobi], pyaugmecon[xpress], pyaugmecon[cbc]. cbc now installs cbcbox to match Pyomo's executable backend path. scip remains supported through Pyomo when a scip executable is installed. The base install does not pull in any solver.
• Documented solver fallback order, backend selection rationale, and common free/community license notes.
• Packaged the knapsack CSV data used by pyaugmecon.example_models, so the bundled examples work from installed wheels.
• Rewrote benchmarks as a modular CLI (benchmarks/cli.py, cases.py, engines.py) with engine pluggability supporting both this project and the augmecon-py implementation. Bundled knapsack benchmark cases are maintained and documented in benchmarks/README.md.
• Restructured tests into tests/unit, tests/integration, and tests/support, with knapsack regression reference data under tests/input/.
• Switched project tooling to uv, Ruff, ty, pytest, and prek.
• Replaced the previous CI setup with GitHub Actions for Ruff, ty, and pytest on Python 3.12, 3.13, and 3.14.
• Added release workflow to verify builds and publish to PyPI.
• Removed obsolete files: dev/make_distribution.sh, gurobi.env, and pytest.ini.

Fixed Issues and PRs

• Queue and scheduling work: Issue #14.
• Auto publish pipeline: Issue #15.
• CI coverage updates: Issue #16.
• Solution storage updates: Issue #17.
• Solver support including CBC: Issue #20.
• Objective auto-deactivation: Issue #22.
• Reserved-name conflict checks: Issue #23.
• Thanks @yvanoers for PR #6 and PR #7.
• Thanks @kschulze26 for PR #25.

1.0.8

• Add method to check if user-provided model component names mask pyaugmecon-added component names
• Update LICENSE

1.0.7

• Update readme to include other solvers
• Fix Pandas warning

1.0.6

Fix dependencies

1.0.5

• Update dependencies

1.0.4

• Correctly enqueue items such that the work can be redistributed when a process finishes its work

1.0.3

• Use new Pyomo 6.6.1 method to close() Gurobi and prevent duplicate instances, thanks @torressa

1.0.2

• Update dependencies

1.0.1

• Make QueueHandler work on MacOS #5, thanks @yvanoers