autoresearch

Automate hyperparameter tuning by defining a program.md file with your objective, eval command, and metric direction, then run an autonomous optimize, measure, and keep or discard iteration loop to automatically adjust targets based on performance metrics.

An autonomous iteration loop is an automated process that repeatedly executes an optimize, measure, and keep or discard cycle on any target file with a measurable metric, eliminating the need for manual experimentation and analysis.

Set up automated optimization by creating a program.md file specifying your objective, target files, eval command, metric extraction, direction, constraints, and max iterations, then execute the autoresearch run command with the program file.

Yes, you can optimize prompts and templates by defining a program.md that specifies the target file, evaluation command, and metric extraction, then running the autonomous iteration loop to automatically refine and keep or discard variations.

Yes, defining constraints and max iterations in your program.md is required to control the autonomous optimization loop, ensuring the measurable targets are optimized within safe boundaries and a bounded number of cycles.