AlgoTune Agent
Command-driven algorithm optimizer with evaluation tools, profiling, multi-file edits, and best-snapshot restore.
name: algotune_agent
display_name: AlgoTune Agent
organization: AlgoTune
type: test_time_search
family: autonomous_agent
summary: "Command-driven algorithm optimizer with evaluation tools, profiling, multi-file edits, and best-snapshot restore."
source: "https://github.com/oripress/AlgoTune"
tags: [algotune, algorithm-engineering, command-agent, profiling, performance-optimization]
license: MIT
controller: galapagos.scaffolds.algotune_agent.scaffold.AlgoTuneScaffold
components:
population: {kind: best_snapshot_archive}
selection_policy: {kind: mutable_workspace_with_revert}
prompt_builder: {kind: algotune_command_transcript}
proposer: {kind: command_driven_optimizer}
evaluator: {kind: task}
memory: {kind: bounded_role_history}
model:
default: openai/o4-mini
host: openrouter
examples:
- "galapagos run --scaffold algotune_agent --task algotune_svm --set general.max_iterations=9999"