[PR #3353] Register Arquinovatos tools node pack #3221

New issue

Open

opened 2026-10-28 21:16:12 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-28 21:16:12 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/Comfy-Org/ComfyUI-Manager/pull/3353
Author: @danielpatillo
Created: 10/6/2026
Status: 🔄 Open

Base: main ← Head: register-arquinovatos-prompt-enhancer

📝 Commits (3)

b3b1cbb Register Arquinovatos Prompt Enhancer node pack
d1e891e Describe combined Arquinovatos model acquisition node and retained compatibility
fba382c Update Arquinovatos tools Manager registration for v0.0.04

📊 Changes

1 file changed (+10 additions, -0 deletions)

View changed files

📝 custom-node-list.json (+10 -0)

📄 Description

Registers Arquinovatos_Prompt_Enhancer as one Manager node pack.

The v0.0.03 pack adds (Down)load LLM model by Arquinovatos (Arquinovatos_LLM_Model): select a catalogue model or an existing local GGUF. On execution it reuses installed files or acquires a missing catalogue GGUF and pinned Windows CUDA engine. A missing/invalid manual file reports an error without downloading a substitute. Importing the pack does not download files or start a server.

The prompt enhancer accepts the model connection and outputs improved text, run information, and the exact LLM messages. Prompt attributes default blank. Optional generation controls default off (768 output tokens, temperature 0.15, context 4096); enabled controls apply only to that execution. The previous loader and downloader classes remain for compatibility.

Python uses the standard library. Automatic engine installation targets Windows x64 with NVIDIA CUDA; other setups need a compatible CUDA-enabled llama.cpp server configured manually.

Only this pack's entry is added to custom-node-list.json; its description now includes the combined node. All 5,986 pre-existing entries, their order, and their bytes are preserved.

Validation of v0.0.03:

Official json-checker.py passes.
Anonymous public ZIP download matches all 24 source files; 48 tests pass on downloaded code.
22 Chromium DOM checks pass (generation transport simulated).
14 live ComfyUI functional cases pass: 10 generations across the three catalogue models and manual file, 3 existing model preparations, and one expected manual-path error.
Effective parameters, advanced OFF reset, server release, and real UI generation/copy roundtrip verified.
Independent semantic review accepts 9/10 generated samples; one invented condition conflicts with the original house scene. This limitation is documented; functional checks do not establish universal LLM fidelity.

Published pack commit: aa20d38.

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/Comfy-Org/ComfyUI-Manager/pull/3353 **Author:** [@danielpatillo](https://github.com/danielpatillo) **Created:** 10/6/2026 **Status:** 🔄 Open **Base:** `main` ← **Head:** `register-arquinovatos-prompt-enhancer` --- ### 📝 Commits (3) - [`b3b1cbb`](https://github.com/Comfy-Org/ComfyUI-Manager/commit/b3b1cbb4ea3d4bd13764e160a4c606a57ac0c9a8) Register Arquinovatos Prompt Enhancer node pack - [`d1e891e`](https://github.com/Comfy-Org/ComfyUI-Manager/commit/d1e891e122028b7f2ff7e87cd9a63ca270b35de3) Describe combined Arquinovatos model acquisition node and retained compatibility - [`fba382c`](https://github.com/Comfy-Org/ComfyUI-Manager/commit/fba382c5fa7f0fc5b975585f8baacd8dbcaa095b) Update Arquinovatos tools Manager registration for v0.0.04 ### 📊 Changes **1 file changed** (+10 additions, -0 deletions) <details> <summary>View changed files</summary> 📝 `custom-node-list.json` (+10 -0) </details> ### 📄 Description Registers [Arquinovatos_Prompt_Enhancer](https://github.com/danielpatillo/arquinovatos) as one Manager node pack. The v0.0.03 pack adds **(Down)load LLM model by Arquinovatos** (`Arquinovatos_LLM_Model`): select a catalogue model or an existing local GGUF. On execution it reuses installed files or acquires a missing catalogue GGUF and pinned Windows CUDA engine. A missing/invalid manual file reports an error without downloading a substitute. Importing the pack does not download files or start a server. The prompt enhancer accepts the model connection and outputs improved text, run information, and the exact LLM messages. Prompt attributes default blank. Optional generation controls default off (768 output tokens, temperature 0.15, context 4096); enabled controls apply only to that execution. The previous loader and downloader classes remain for compatibility. Python uses the standard library. Automatic engine installation targets Windows x64 with NVIDIA CUDA; other setups need a compatible CUDA-enabled llama.cpp server configured manually. Only this pack's entry is added to `custom-node-list.json`; its description now includes the combined node. All 5,986 pre-existing entries, their order, and their bytes are preserved. Validation of v0.0.03: - Official `json-checker.py` passes. - Anonymous public ZIP download matches all 24 source files; 48 tests pass on downloaded code. - 22 Chromium DOM checks pass (generation transport simulated). - 14 live ComfyUI functional cases pass: 10 generations across the three catalogue models and manual file, 3 existing model preparations, and one expected manual-path error. - Effective parameters, advanced OFF reset, server release, and real UI generation/copy roundtrip verified. - Independent semantic review accepts 9/10 generated samples; one invented condition conflicts with the original house scene. This limitation is documented; functional checks do not establish universal LLM fidelity. Published pack commit: [aa20d38](https://github.com/danielpatillo/arquinovatos/commit/aa20d38cba758d57c98c8aee1b6fa40d33e30416). --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi added the

pull-request

label

2026-10-28 21:16:12 +00:00