---

name: configure_optimizations description: > Return explicit optimization mode config for cache, memoization, ModelManager, and attention reuse. homepage: https://github.com/steliosot/comfy-agent metadata.clawdbot.os: ["darwin", "linux"] metadata.clawdbot.requires.bins: ["python3"] metadata.clawdbot.requires.env: [] metadata.clawdbot.files: ["skill.py", "skill.yaml", "scripts/*"] metadata.clawdbot.tags: ["comfyui", "skill", "ops", "optimization", "cache", "memoization", "model-manager", "attention-reuse"] metadata.clawdbot.category: "ops" metadata.clawdbot.input_type: "application/json" metadata.clawdbot.output_type: "application/json" metadata.clawdbot.output_can_feed_into: [] metadata.clawdbot.accepts_input_from: [] metadata.clawdbot.priority: 74

configure_optimizations

Build an optimization profile for:

• no optimization
• deterministic node cache
• graph memoization
• model LRU
• CPU offload
• lazy loading
• attention reuse
• full stack

Returns:

• workflow flags (cache_enabled, memoization_enabled, policy/size)
• model_manager flags (max_models_in_vram, enable_cpu_offload, lazy_loading)
• attention_reuse flags (threshold, cache_device, store_frequency, reuse_layers)
• python_snippet you can paste in scripts.