steliosot/comfy-agent

configure-optimizations

Return explicit optimization mode config for cache, memoization, ModelManager, and attention reuse.

optimization-attention-reuse-cpu-offload

Enable attention reuse plus ModelManager CPU offload.

optimization-attention-reuse-ltxv-cpu-offload

Enable LTXV attention reuse preset with CPU offload.

optimization-attention-reuse-ltxv

Enable attention reuse preset tuned for LTXV temporal/transformer pipelines.

optimization-attention-reuse-sd-cpu-offload

Enable SD attention reuse preset with CPU offload.

optimization-attention-reuse-sd

Enable attention reuse preset tuned for SD image pipelines.

optimization-attention-reuse

Enable attention map reuse plugin mode.

optimization-attention-reuse-wan21-cpu-offload

Enable WAN2.1 attention reuse preset with CPU offload.

optimization-attention-reuse-wan21

Enable attention reuse preset tuned for WAN2.1 video pipelines.

optimization-attention-reuse-wan22-cpu-offload

Enable WAN2.2 attention reuse preset with CPU offload.

optimization-attention-reuse-wan22

Enable attention reuse preset tuned for WAN2.2 pipelines.

optimization-cache

Enable deterministic node cache only.

optimization-cpu-offload

Enable ModelManager with CPU offload.

optimization-full-stack

Enable full optimization stack (cache + memoization + model manager).

optimization-lazy-loading

Enable lazy-loading focused ModelManager profile.

optimization-memoization

Enable deterministic cache plus graph memoization.

optimization-model-lru

Enable ModelManager LRU without CPU offload.

optimization-none

Disable cache, memoization, and model manager optimizations.

curated-ep47-acestep-v1-3-5b-text-to-music-lyrics

Curated ComfyUI workflow imported from `EP47 AceStep V1 3.5b text to music - Lyrics.json`.

select-comfy-server

Resolve a named Comfy server from YAML registry and return run-ready connection params.

agentic-execute

Execute a precomputed agentic plan payload.

agentic-plan

Build a structured plan for agentic generation before execution.

assess-server-resources

Assess queue load and hardware/resource compatibility (VRAM/storage) before installs or workflow runs.

delete-image-job

Delete a ComfyUI history job by prompt_id (image/video specific checks).

delete-video-job

Delete a ComfyUI history job by prompt_id (image/video specific checks).

download-image

Download ComfyUI image outputs to local files using deterministic names.

download-model

Install/download models via ComfyUI-Manager API, with source-aware token support for HuggingFace and Civitai.

download-video

Download generated video outputs from ComfyUI history to local storage.

get-progress

Get progress percentage for a prompt; uses queue/history fallback if /progress is unavailable.

get-queue-status

Query ComfyUI queue status for running and pending jobs.

get-server-status

Query ComfyUI server health, queue counters, and system stats.

get-workflow-download-links

Extracts model/dependency links from notes embedded in workflow JSON files and groups them by provider.

install-custom-node

Install a ComfyUI custom node package from a git URL through ComfyUI-Manager API and verify node availability.

list-comfy-assets

List available ComfyUI assets (checkpoints, vae, clip, lora, unet) and local input/output image files.

list-curated-workflows

List curated workflow skills and filter them by capability family or model family hints.

match-curated-workflow

Match user intent to the most appropriate curated workflow skills.

model-folder-guide

Return model type placement guidance for ComfyUI (for example checkpoints vs diffusion_models).

predict-job-success-likelihood

Predicts the likelihood of successful workflow execution using installed dependencies, resource status, and similar historical jobs.

prepare-workflow-dependencies

Detect workflow dependency gaps (models/custom nodes), assess server resources, and auto-fix through install skills.

remove-model

Remove a model from ComfyUI through Manager API and verify that loaders no longer list it.