ce-optimize

Automate metric-driven code optimization by orchestrating repeated experiments against a measurable goal, using baselines, gates, and judge scoring to converge on the best solution. The system runs batches and records results for audit and rollback.

Iterative build time optimization is managed by defining a repeatable harness, seeding a backlog of hypotheses, and running batch experiments serially or in parallel. Strict checkpoints and safety gates ensure tests stay green and enforce reproducibility.

LLM-based judging for semantic quality optimization is supported with strict spend controls. The system orchestrates experiments, applies judge-based scoring to variants, and merges the best-in-batch results while maintaining cost control.

Crash recovery for iterative experiment harnesses relies on durable on-disk logs and explicit checkpoints. These mechanisms enforce data integrity, allowing the optimization loop to resume from the last safe state without losing previous measurement results.

Python3 is required to run metric-driven optimization batches. The environment uses Python3 scripts to orchestrate the measurement harness, execute variant tests, and record results for durable, resumable improvement cycles.

Rollback after a failed optimization experiment is enabled by durable on-disk logs and explicit checkpoints. The system records all results for audit, allowing you to safely revert variants that fail safety gates or regress metrics.