mirror of
https://github.com/NousResearch/hermes-agent.git
synced 2026-07-31 19:16:29 +00:00
refactor(skills): tighten comfyui MCP routing + trim skill below pre-PR size
Rework the salvaged Route First section and cut fat so the skill SHRINKS despite the new MCP routing guidance (6,310 tokens vs 6,342 before the routing PR; contributor version was 6,748): - Route First: remote/cloud -> comfy-cloud MCP from the catalog (hermes mcp install comfy-cloud); this skill owns local runs and headless/CI API-key use. Neutral on local-vs-cloud choice (dropped the 24GB-VRAM cloud-default gate — user choice wins when local is viable). Notes the MCP's ~10k-token always-on schema cost. - description: 60 chars (was 217) per skill authoring standard - Step 0: collapsed quoted onboarding script to a plain choice list - Paths B/C: collapsed to two-liners; model downloads: 4 examples -> 2 - docs page + skills-catalog line regenerated/updated in lockstep
This commit is contained in:
Teknium
parent
07e4c45c9f
commit
883a465efd
3 changed files with 81 additions and 151 deletions
|
|
@ -1,6 +1,6 @@
|
|||
---
|
||||
name: comfyui
|
||||
description: "Generate images, video, and audio with ComfyUI — install, launch, manage nodes/models, run workflows with parameter injection. Uses the official comfy-cli for lifecycle and direct REST/WebSocket API for execution."
|
||||
description: "ComfyUI image/video/audio generation — local or Comfy Cloud."
|
||||
version: 5.2.0
|
||||
author: [kshitijk4poor, alt-glitch, purzbeats]
|
||||
license: MIT
|
||||
|
|
@ -78,30 +78,27 @@ SDXL inpaint, ESRGAN upscale, AnimateDiff video, Wan T2V. See
|
|||
|
||||
## Route First: Comfy Cloud MCP vs. This Skill
|
||||
|
||||
Hermes can also drive Comfy Cloud through the `comfy-cloud` MCP server from
|
||||
the Nous-approved catalog (`hermes mcp install comfy-cloud`) — a hosted
|
||||
remote server with browser OAuth sign-in, no API keys to copy around, and no
|
||||
local compute. Decide which surface to use **before** running anything from
|
||||
this skill:
|
||||
For **remote/cloud** generation, prefer the `comfy-cloud` MCP server from
|
||||
the Nous-approved catalog over this skill's API-key path:
|
||||
|
||||
1. **Default to the Comfy Cloud MCP.** It is the easiest path — one-command
|
||||
install, OAuth on first connect, nothing to run locally. Recommend it
|
||||
especially when the machine is a Mac or the local GPU has less than
|
||||
24 GB VRAM.
|
||||
2. **Offer a choice only when local is genuinely viable**: the local GPU has
|
||||
24 GB+ VRAM AND a local ComfyUI setup already exists. Then ask the user
|
||||
whether they want to run Comfy locally or on cloud.
|
||||
3. **If the user chooses local**, remind them it takes a bit longer, and use
|
||||
this skill to handle the rest — it will find or set up the ComfyUI
|
||||
instance (`comfyui_setup.sh`) and install any missing models/nodes
|
||||
(`auto_fix_deps.py`) if they don't have them already.
|
||||
4. **The MCP and this skill are alternatives, not layers.** Pick one for the
|
||||
task at hand; don't mix MCP tool calls with this skill's CLI/REST scripts
|
||||
against the same job.
|
||||
```bash
|
||||
hermes mcp install comfy-cloud # browser OAuth on first connect, no API keys
|
||||
```
|
||||
|
||||
The API-key Cloud path in this skill (**Path A** below) remains the right
|
||||
choice for headless / CI / scripted use where a browser OAuth flow isn't
|
||||
available.
|
||||
Hosted, zero local compute, a curated tool subset enabled by default
|
||||
(`hermes mcp configure comfy-cloud` to adjust). Its tool schemas add ~10k
|
||||
tokens to every request while enabled — right for regular Comfy work, heavy
|
||||
for a one-off generation.
|
||||
|
||||
Use this skill instead when:
|
||||
|
||||
- The user wants to run ComfyUI **locally** — this skill handles setup
|
||||
(`comfyui_setup.sh`), models/nodes (`auto_fix_deps.py`), and execution.
|
||||
- **Headless / CI / scripted** cloud use where browser OAuth isn't
|
||||
available — **Path A** below (API key).
|
||||
|
||||
Pick one surface per task; don't mix MCP tool calls with this skill's
|
||||
CLI/REST scripts against the same job.
|
||||
|
||||
## Architecture: Two Layers
|
||||
|
||||
|
|
@ -257,15 +254,10 @@ The scripts emit JSON to stdout describing every output file:
|
|||
|
||||
## Setup & Onboarding
|
||||
|
||||
When a user asks to set up ComfyUI, **the FIRST thing to do is ask whether
|
||||
they want Comfy Cloud (hosted, zero install, API key) or Local (install
|
||||
ComfyUI on their machine)**. Don't start running install commands or hardware
|
||||
checks until they've answered.
|
||||
|
||||
(If you haven't already routed between the Comfy Cloud MCP and this skill,
|
||||
do that first — see **Route First** above. Interactive cloud users are
|
||||
usually better served by `hermes mcp install comfy-cloud` than by this
|
||||
skill's API-key path.)
|
||||
Route first (see **Route First** above) — interactive cloud users are usually
|
||||
better served by `hermes mcp install comfy-cloud` than by this skill. If the
|
||||
user is setting up *this skill*, **ask whether they want Comfy Cloud
|
||||
(API key) or Local before running any install commands or hardware checks**.
|
||||
|
||||
**Official docs:** https://docs.comfy.org/installation
|
||||
**CLI docs:** https://docs.comfy.org/comfy-cli/getting-started
|
||||
|
|
@ -274,20 +266,14 @@ skill's API-key path.)
|
|||
|
||||
### Step 0: Ask Local vs Cloud (ALWAYS FIRST)
|
||||
|
||||
Suggested script:
|
||||
Offer the choice plainly:
|
||||
|
||||
> "Do you want to run ComfyUI locally on your machine, or use Comfy Cloud?
|
||||
>
|
||||
> - **Comfy Cloud** — hosted on RTX 6000 Pro GPUs, all common models pre-installed,
|
||||
> zero setup. Requires an API key (paid subscription required to actually run
|
||||
> workflows; free tier is read-only). Best if you don't have a capable GPU.
|
||||
> - **Local** — free, but your machine MUST meet the hardware requirements:
|
||||
> - NVIDIA GPU with **≥6 GB VRAM** (≥8 GB for SDXL, ≥12 GB for Flux/video), OR
|
||||
> - AMD GPU with ROCm support (Linux), OR
|
||||
> - Apple Silicon Mac (M1+) with **≥16 GB unified memory** (≥32 GB recommended).
|
||||
> - Intel Macs and machines with no GPU will NOT work — use Cloud instead.
|
||||
>
|
||||
> Which would you like?"
|
||||
- **Comfy Cloud** — hosted (RTX 6000 Pro), common models pre-installed, zero
|
||||
setup. Needs an API key; paid subscription required to run workflows
|
||||
(free tier is read-only). Best without a capable GPU.
|
||||
- **Local** — free, but requires: NVIDIA ≥6 GB VRAM (≥8 GB SDXL, ≥12 GB
|
||||
Flux/video), OR AMD ROCm (Linux), OR Apple Silicon ≥16 GB unified
|
||||
(≥32 GB recommended). Intel Macs / no-GPU machines will NOT work.
|
||||
|
||||
Routing:
|
||||
|
||||
|
|
@ -376,24 +362,14 @@ required for `/api/prompt`, `/api/upload/*`, `/api/view`, etc.
|
|||
|
||||
### Path B: ComfyUI Desktop (Windows / macOS)
|
||||
|
||||
One-click installer for non-technical users. Currently Beta.
|
||||
|
||||
**Docs:** https://docs.comfy.org/installation/desktop
|
||||
- **Windows (NVIDIA):** https://download.comfy.org/windows/nsis/x64
|
||||
- **macOS (Apple Silicon):** https://comfy.org
|
||||
|
||||
Linux is **not supported** for Desktop — use Path D.
|
||||
|
||||
---
|
||||
One-click Beta installer for non-technical users. Not available on Linux —
|
||||
use Path D. **Docs:** https://docs.comfy.org/installation/desktop
|
||||
|
||||
### Path C: ComfyUI Portable (Windows Only)
|
||||
|
||||
**Docs:** https://docs.comfy.org/installation/comfyui_portable_windows
|
||||
|
||||
Download from https://github.com/comfyanonymous/ComfyUI/releases, extract,
|
||||
run `run_nvidia_gpu.bat`. Update via `update/update_comfyui_stable.bat`.
|
||||
|
||||
---
|
||||
**Docs:** https://docs.comfy.org/installation/comfyui_portable_windows
|
||||
|
||||
### Path D: comfy-cli (All Platforms — Recommended for Agents)
|
||||
|
||||
|
|
@ -459,26 +435,15 @@ python main.py
|
|||
### Post-Install: Download Models
|
||||
|
||||
```bash
|
||||
# SDXL (general purpose, ~6.5 GB)
|
||||
# SDXL (general purpose, ~6.5 GB). Same pattern for SD 1.5 / Flux fp8 —
|
||||
# any direct HuggingFace .safetensors URL works.
|
||||
comfy model download \
|
||||
--url "https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/resolve/main/sd_xl_base_1.0.safetensors" \
|
||||
--relative-path models/checkpoints
|
||||
|
||||
# SD 1.5 (lighter, ~4 GB, good for 6 GB cards)
|
||||
comfy model download \
|
||||
--url "https://huggingface.co/stable-diffusion-v1-5/stable-diffusion-v1-5/resolve/main/v1-5-pruned-emaonly.safetensors" \
|
||||
--relative-path models/checkpoints
|
||||
|
||||
# Flux Dev fp8 (smaller variant, ~12 GB)
|
||||
comfy model download \
|
||||
--url "https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors" \
|
||||
--relative-path models/checkpoints
|
||||
|
||||
# CivitAI (set token first):
|
||||
comfy model download \
|
||||
--url "https://civitai.com/api/download/models/128713" \
|
||||
--relative-path models/checkpoints \
|
||||
--set-civitai-api-token "YOUR_TOKEN"
|
||||
# CivitAI needs a token:
|
||||
comfy model download --url "https://civitai.com/api/download/models/128713" \
|
||||
--relative-path models/checkpoints --set-civitai-api-token "YOUR_TOKEN"
|
||||
```
|
||||
|
||||
List installed: `comfy model list`.
|
||||
|
|
|
|||
|
|
@ -40,7 +40,7 @@ If a skill is missing from this list but present in the repo, the catalog is reg
|
|||
| [`ascii-video`](/docs/user-guide/skills/bundled/creative/creative-ascii-video) | ASCII video: convert video/audio to colored ASCII MP4/GIF. | `creative/ascii-video` |
|
||||
| [`baoyu-infographic`](/docs/user-guide/skills/bundled/creative/creative-baoyu-infographic) | Infographics: 21 layouts x 21 styles (信息图, 可视化). | `creative/baoyu-infographic` |
|
||||
| [`claude-design`](/docs/user-guide/skills/bundled/creative/creative-claude-design) | Design one-off HTML artifacts (landing, deck, prototype). | `creative/claude-design` |
|
||||
| [`comfyui`](/docs/user-guide/skills/bundled/creative/creative-comfyui) | Generate images, video, and audio with ComfyUI — install, launch, manage nodes/models, run workflows with parameter injection. Uses the official comfy-cli for lifecycle and direct REST/WebSocket API for execution. | `creative/comfyui` |
|
||||
| [`comfyui`](/docs/user-guide/skills/bundled/creative/creative-comfyui) | ComfyUI image/video/audio generation — local or Comfy Cloud. | `creative/comfyui` |
|
||||
| [`design-md`](/docs/user-guide/skills/bundled/creative/creative-design-md) | Author/validate/export Google's DESIGN.md token spec files. | `creative/design-md` |
|
||||
| [`excalidraw`](/docs/user-guide/skills/bundled/creative/creative-excalidraw) | Hand-drawn Excalidraw JSON diagrams (arch, flow, seq). | `creative/excalidraw` |
|
||||
| [`humanizer`](/docs/user-guide/skills/bundled/creative/creative-humanizer) | Humanize text: strip AI-isms and add real voice. | `creative/humanizer` |
|
||||
|
|
|
|||
|
|
@ -1,14 +1,14 @@
|
|||
---
|
||||
title: "Comfyui"
|
||||
title: "Comfyui — ComfyUI image/video/audio generation — local or Comfy Cloud"
|
||||
sidebar_label: "Comfyui"
|
||||
description: "Generate images, video, and audio with ComfyUI — install, launch, manage nodes/models, run workflows with parameter injection"
|
||||
description: "ComfyUI image/video/audio generation — local or Comfy Cloud"
|
||||
---
|
||||
|
||||
{/* This page is auto-generated from the skill's SKILL.md by website/scripts/generate-skill-docs.py. Edit the source SKILL.md, not this page. */}
|
||||
|
||||
# Comfyui
|
||||
|
||||
Generate images, video, and audio with ComfyUI — install, launch, manage nodes/models, run workflows with parameter injection. Uses the official comfy-cli for lifecycle and direct REST/WebSocket API for execution.
|
||||
ComfyUI image/video/audio generation — local or Comfy Cloud.
|
||||
|
||||
## Skill metadata
|
||||
|
||||
|
|
@ -80,30 +80,27 @@ SDXL inpaint, ESRGAN upscale, AnimateDiff video, Wan T2V. See
|
|||
|
||||
## Route First: Comfy Cloud MCP vs. This Skill
|
||||
|
||||
Hermes can also drive Comfy Cloud through the `comfy-cloud` MCP server from
|
||||
the Nous-approved catalog (`hermes mcp install comfy-cloud`) — a hosted
|
||||
remote server with browser OAuth sign-in, no API keys to copy around, and no
|
||||
local compute. Decide which surface to use **before** running anything from
|
||||
this skill:
|
||||
For **remote/cloud** generation, prefer the `comfy-cloud` MCP server from
|
||||
the Nous-approved catalog over this skill's API-key path:
|
||||
|
||||
1. **Default to the Comfy Cloud MCP.** It is the easiest path — one-command
|
||||
install, OAuth on first connect, nothing to run locally. Recommend it
|
||||
especially when the machine is a Mac or the local GPU has less than
|
||||
24 GB VRAM.
|
||||
2. **Offer a choice only when local is genuinely viable**: the local GPU has
|
||||
24 GB+ VRAM AND a local ComfyUI setup already exists. Then ask the user
|
||||
whether they want to run Comfy locally or on cloud.
|
||||
3. **If the user chooses local**, remind them it takes a bit longer, and use
|
||||
this skill to handle the rest — it will find or set up the ComfyUI
|
||||
instance (`comfyui_setup.sh`) and install any missing models/nodes
|
||||
(`auto_fix_deps.py`) if they don't have them already.
|
||||
4. **The MCP and this skill are alternatives, not layers.** Pick one for the
|
||||
task at hand; don't mix MCP tool calls with this skill's CLI/REST scripts
|
||||
against the same job.
|
||||
```bash
|
||||
hermes mcp install comfy-cloud # browser OAuth on first connect, no API keys
|
||||
```
|
||||
|
||||
The API-key Cloud path in this skill (**Path A** below) remains the right
|
||||
choice for headless / CI / scripted use where a browser OAuth flow isn't
|
||||
available.
|
||||
Hosted, zero local compute, a curated tool subset enabled by default
|
||||
(`hermes mcp configure comfy-cloud` to adjust). Its tool schemas add ~10k
|
||||
tokens to every request while enabled — right for regular Comfy work, heavy
|
||||
for a one-off generation.
|
||||
|
||||
Use this skill instead when:
|
||||
|
||||
- The user wants to run ComfyUI **locally** — this skill handles setup
|
||||
(`comfyui_setup.sh`), models/nodes (`auto_fix_deps.py`), and execution.
|
||||
- **Headless / CI / scripted** cloud use where browser OAuth isn't
|
||||
available — **Path A** below (API key).
|
||||
|
||||
Pick one surface per task; don't mix MCP tool calls with this skill's
|
||||
CLI/REST scripts against the same job.
|
||||
|
||||
## Architecture: Two Layers
|
||||
|
||||
|
|
@ -261,15 +258,10 @@ The scripts emit JSON to stdout describing every output file:
|
|||
|
||||
## Setup & Onboarding
|
||||
|
||||
When a user asks to set up ComfyUI, **the FIRST thing to do is ask whether
|
||||
they want Comfy Cloud (hosted, zero install, API key) or Local (install
|
||||
ComfyUI on their machine)**. Don't start running install commands or hardware
|
||||
checks until they've answered.
|
||||
|
||||
(If you haven't already routed between the Comfy Cloud MCP and this skill,
|
||||
do that first — see **Route First** above. Interactive cloud users are
|
||||
usually better served by `hermes mcp install comfy-cloud` than by this
|
||||
skill's API-key path.)
|
||||
Route first (see **Route First** above) — interactive cloud users are usually
|
||||
better served by `hermes mcp install comfy-cloud` than by this skill. If the
|
||||
user is setting up *this skill*, **ask whether they want Comfy Cloud
|
||||
(API key) or Local before running any install commands or hardware checks**.
|
||||
|
||||
**Official docs:** https://docs.comfy.org/installation
|
||||
**CLI docs:** https://docs.comfy.org/comfy-cli/getting-started
|
||||
|
|
@ -278,20 +270,14 @@ skill's API-key path.)
|
|||
|
||||
### Step 0: Ask Local vs Cloud (ALWAYS FIRST)
|
||||
|
||||
Suggested script:
|
||||
Offer the choice plainly:
|
||||
|
||||
> "Do you want to run ComfyUI locally on your machine, or use Comfy Cloud?
|
||||
>
|
||||
> - **Comfy Cloud** — hosted on RTX 6000 Pro GPUs, all common models pre-installed,
|
||||
> zero setup. Requires an API key (paid subscription required to actually run
|
||||
> workflows; free tier is read-only). Best if you don't have a capable GPU.
|
||||
> - **Local** — free, but your machine MUST meet the hardware requirements:
|
||||
> - NVIDIA GPU with **≥6 GB VRAM** (≥8 GB for SDXL, ≥12 GB for Flux/video), OR
|
||||
> - AMD GPU with ROCm support (Linux), OR
|
||||
> - Apple Silicon Mac (M1+) with **≥16 GB unified memory** (≥32 GB recommended).
|
||||
> - Intel Macs and machines with no GPU will NOT work — use Cloud instead.
|
||||
>
|
||||
> Which would you like?"
|
||||
- **Comfy Cloud** — hosted (RTX 6000 Pro), common models pre-installed, zero
|
||||
setup. Needs an API key; paid subscription required to run workflows
|
||||
(free tier is read-only). Best without a capable GPU.
|
||||
- **Local** — free, but requires: NVIDIA ≥6 GB VRAM (≥8 GB SDXL, ≥12 GB
|
||||
Flux/video), OR AMD ROCm (Linux), OR Apple Silicon ≥16 GB unified
|
||||
(≥32 GB recommended). Intel Macs / no-GPU machines will NOT work.
|
||||
|
||||
Routing:
|
||||
|
||||
|
|
@ -380,24 +366,14 @@ required for `/api/prompt`, `/api/upload/*`, `/api/view`, etc.
|
|||
|
||||
### Path B: ComfyUI Desktop (Windows / macOS)
|
||||
|
||||
One-click installer for non-technical users. Currently Beta.
|
||||
|
||||
**Docs:** https://docs.comfy.org/installation/desktop
|
||||
- **Windows (NVIDIA):** https://download.comfy.org/windows/nsis/x64
|
||||
- **macOS (Apple Silicon):** https://comfy.org
|
||||
|
||||
Linux is **not supported** for Desktop — use Path D.
|
||||
|
||||
---
|
||||
One-click Beta installer for non-technical users. Not available on Linux —
|
||||
use Path D. **Docs:** https://docs.comfy.org/installation/desktop
|
||||
|
||||
### Path C: ComfyUI Portable (Windows Only)
|
||||
|
||||
**Docs:** https://docs.comfy.org/installation/comfyui_portable_windows
|
||||
|
||||
Download from https://github.com/comfyanonymous/ComfyUI/releases, extract,
|
||||
run `run_nvidia_gpu.bat`. Update via `update/update_comfyui_stable.bat`.
|
||||
|
||||
---
|
||||
**Docs:** https://docs.comfy.org/installation/comfyui_portable_windows
|
||||
|
||||
### Path D: comfy-cli (All Platforms — Recommended for Agents)
|
||||
|
||||
|
|
@ -463,26 +439,15 @@ python main.py
|
|||
### Post-Install: Download Models
|
||||
|
||||
```bash
|
||||
# SDXL (general purpose, ~6.5 GB)
|
||||
# SDXL (general purpose, ~6.5 GB). Same pattern for SD 1.5 / Flux fp8 —
|
||||
# any direct HuggingFace .safetensors URL works.
|
||||
comfy model download \
|
||||
--url "https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/resolve/main/sd_xl_base_1.0.safetensors" \
|
||||
--relative-path models/checkpoints
|
||||
|
||||
# SD 1.5 (lighter, ~4 GB, good for 6 GB cards)
|
||||
comfy model download \
|
||||
--url "https://huggingface.co/stable-diffusion-v1-5/stable-diffusion-v1-5/resolve/main/v1-5-pruned-emaonly.safetensors" \
|
||||
--relative-path models/checkpoints
|
||||
|
||||
# Flux Dev fp8 (smaller variant, ~12 GB)
|
||||
comfy model download \
|
||||
--url "https://huggingface.co/Comfy-Org/flux1-dev/resolve/main/flux1-dev-fp8.safetensors" \
|
||||
--relative-path models/checkpoints
|
||||
|
||||
# CivitAI (set token first):
|
||||
comfy model download \
|
||||
--url "https://civitai.com/api/download/models/128713" \
|
||||
--relative-path models/checkpoints \
|
||||
--set-civitai-api-token "YOUR_TOKEN"
|
||||
# CivitAI needs a token:
|
||||
comfy model download --url "https://civitai.com/api/download/models/128713" \
|
||||
--relative-path models/checkpoints --set-civitai-api-token "YOUR_TOKEN"
|
||||
```
|
||||
|
||||
List installed: `comfy model list`.
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue