Frequently Asked QuestionsCommon questions and answers about using WecoGetting Started How do I log in to Weco from the CLI?How do I log out of the CLI?Which Python versions are supported?Does Weco work on Windows, macOS and Linux?How do I track my runs with the weco agent/where can I view my run history? Basic Usage & Setup What exactly does weco run modify in my code?How can I keep my original file safe (version control / backup)?How do I revert changes after a run completes?What does the --eval-command flag do?Where is the evaluation script run? Evaluation Scripts & Metrics Does my evaluation script need to be in Python?Does the evaluation script have to be in the same language as the code I'm optimizing?How does Weco read my metric?How do I choose a metric and goal?Does the metric printed need to exactly match the string passed to the `--metric` flag?How do I pass environment variables to my evaluation script?Are there any templates/examples of common evaluation scripts? Optimization Capabilities Which kinds of tasks can Weco optimize?Which programming languages can Weco modify?What metrics can I optimize for?Can I point Weco at a directory instead of a single file?Can Weco optimize multiple files at once?How do I constrain the search space? Performance & Best Practices How long can my evaluation take?Can I optimize for multiple objectives/metrics?Is nondeterminism in my evaluation a deal-breaker?How can I provide richer feedback than just one number?How do I pass extra context like markdown or text files? Advanced Features & Troubleshooting What does the solution tree represent?What does the value under each node represent?How do I inspect or download a specific solution from the tree?How do I share a run or solution with teammates?How do I limit the maximum runtime or cost of a run?Can I run multiple Weco runs in parallel?Does Weco support private/self-hosted LLM endpoints?Can Weco automatically stop if the metric plateaus? Privacy & Security What data does Weco send to the cloud?Prompt EngineeringIteratively improve a prompt for solving AIME problems
