Supports cloud-based LLM APIs and local Ollama models. Provides translation for prompt nodes, Markdown nodes, and node documentation; features prompt enhancement, image/video captioning, common tag collection, and history logs. An all-in-one prompt assistant plugin!

📣 Updates

[2026-04-21] 🔥 V2.0.6

Changes:

V3 Architecture Upgrade: Fully refactored node backend to align with ComfyUI V3 API standards for enhanced stability and performance.

Fixes:

Video Caption Fix: Fixed an issue where certain models (e.g., Qwen3.5-Plus) were incorrectly identified as not supporting multi-image analysis. Added intelligent frame-limit inference and automatic truncation.
Ollama Fix: Fixed empty results for Qwen3.5 and other thinking models on Ollama caused by missing think parameter.
Subgraph Mounting Optimization: Fixed assistant initialization and stability on Subgraph nodes across Node 2.0 (Vue) and LiteGraph modes.

V2.0.5

Changes:

Node Random Seed: Added a unified random seed widget to all nodes to handle repetitive execution, replacing the old "[R]" trigger keyword mechanism.
Enhanced Frontend UI Multi-language Support: Special thanks to @rafek1241. Added ui-i18n functionality, supporting (Chinese, English, Japanese, Korean, French, Spanish, Russian, German, etc.).

Fixes:

Node Width Locked: Fixed an issue where node width could not be adjusted in Node 2.0.
Greying Out Native Service baseUrl: Prevented accidental modification of built-in service URLs to avoid request failures.
Network Exception Errors: Fixed a network exception bug for proxy-based (e.g., xflow) requests caused by forced direct-connection logic.
Image Node ✨ Icon Moved: Moved the icon to the right side to prevent overlap with node IDs in Node 2.0.

V2.0.4

Bug Fixes: Fixed issues where tag and history functions were unavailable.

V2.0.3

Assistant UI: Fixed instability of the assistant creator in subgraphs and cases where creators failed to appear when image nodes lacked images.
Ollama: Fixed HTTP 502 errors caused by proxy settings.

V2.0.2

Tag Module: Fixed formatting issues; users can now freely create categories and manage tags. Fixed preset creation and migration errors.
Assistant UI: Optimized mounting methods in Node 2.0, fixed subgraph stability issues, and improved performance.
Interaction: Added streaming input effects and refined UI details.
Translation: Added mixed-language translation rule parameters (default to Chinese/English) and improved node doc translation.
Built-in Rules: Fixed issues with mixed Chinese/English output and missing Kontext translations.
API Requests: Fixed Gemini-1.5-Pro request issues and Ollama 404 errors.
Node Optimization: Improved the Video Caption node.
Console Logs: Optimized logs and fixed a infinite loop bug in progress logs.
Dependencies: Prevented boot failures due to missing dependencies.

V2.0.0

Core Refactoring: Completely rebuilt the assistant for better stability and response speed via API/Ollama.
UI Refresh: Rebuilt frontend components for better stability and Node 2.0 support, including customizable positions and button sorting.
Tag Module: New CSV-based tag mechanism with on-the-fly switching and collection features.
Rule Module: New configuration window supporting categories and custom rule displays; many built-in rules added.
API Service: New API configuration UI with custom service support and multiple model fallback options. Independent service selection for enhancement, translation, and captioning.
Node Overhaul: All nodes rebuilt with multi-language support. Added Video Caption node (Beta).
Migration: User configs moved to \user\default\prompt-assistant to prevent data loss during reinstalls.
New Features: Node doc translation and Markdown node translation.

V1.x.x

V1.2.x

[2025-11-12] V1.2.3

Fixed empty results for Ollama and custom services.
Switched Ollama to native interface for better Qwen2-VL support.
Added HTTP API fallback for better stability.

[2025-10-14] V1.2.2

Dropped support for ComfyUI < 0.3.27 to prevent UI issues.
Fixed errors when using 302.AI and issues with Ollama auto-unload.
Added independent Ollama unload options to all nodes.
Removed forced direct-connection for LLM/VLM to avoid request errors; added a setting toggle for direct-connection.
Optimized console log formatting.

[2025-10-14] V1.2.1

Improved UI responsiveness.
Enhanced API request retry mechanism.
Added settings for auto-removing redundant spaces, dots, and punctuation conversion.
Tag window now remembers size, category, and scroll position.
API config UI now supports auto-fetching model lists.
Ollama auto-vram-unload option added.
Fixed a bug where helpers weren't created for every textbox in list views.

[2025-09-16] V1.2.0

Added Prompt Expand node.
Added 302.AI and Ollama service support.
Tag panel memory feature added.
Right-click menu for quick service switching.
Support for disabling CoT (Chain of Thought) for mainstream models.
Optimized caption and translation nodes.
Added community feedback badge.
Fixed various UI and request bugs.

V1.1.x

[2025-08-28] V1.1.3

Optimized Assistant UI to automatically avoid scrollbars and prevent overlapping accidental triggers.
Fixed an issue where the tag popup lacked a scrollbar, causing incomplete content display.

[2025-08-23] V1.1.2

Refactored nodes to resolve multiple queues and duplicate execution issues during runtime.
Added model parameters to the API config interface; some errors can be resolved by adjusting the max token count.
Simplified image captioning workflow to improve captioning speed.
Fixed an issue where unloaded tags couldn't be searched during on-demand loading.

[2025-08-10] V1.1.1

Fixed image caption node errors.

[2025-08-10] V1.1.0

Modified UI interactions.
Supported all OpenAI SDK compatible APIs.
Added custom rules.
Added custom tags.
Added Image Caption, Kontext Preset, and Translation nodes.

V1.0.x

[2025-06-24] V1.0.6

Fixed some UI bugs.

[2025-06-24] V1.0.5

Fixed assistant UI anomaly when creating a kontext node using the selection toolbar.
Fixed Zhipu service unavailability likely caused by network conditions.
Fixed workflow loading failures caused by instance clearing errors.
Fixed an issue where the AIGODLIKE-COMFYUI-TRANSLATION plugin caused the tag popup to freeze.
Added resizable tag panel.
Optimized UI resource loading mechanism.

[2025-06-24] V1.0.3

Refactored API request service to prevent API keys from being exposed to the frontend.
Modified config save/load mechanism to fix config saving issues.
Fixed minor bugs.

[2025-06-21] V1.0.2

Fixed minor bugs.

[2025-06-15] V1.0.0

One-click tag insertion.
LLM expansion support.
Toggle between Baidu Translate and LLM Translate.
Image to prompt captioning.
History, undo, and redo.

✨ Features

💡 Prompt Optimization + Translation

Supports multiple preset rules (Expand, Qwen-edit, Kontext instructions, etc.) No need to set target language; automatic translation between Chinese and English with caching to prevent meaning drift.

🖼 Image Captioning

Quickly caption images as prompts on image nodes. Supports Chinese/English and various styles (Natural, Tag, etc.).

🔖 Tags, Phrase Presets & Collection

Collect common tags, phrases, and Lora triggers for quick insertion. Supports collection, custom sorting, and multi-set switching.

🕐 History, Undo, Redo

Sentence-based records (triggered on focus loss). Supports undo/redo and cross-node history viewing.

📜 Markdown & Node Documentation Translation

Translates Note and Markdown nodes while maintaining formatting.

Translates node documentation (Beta: Button appears only for nodes with English documentation).

📒 Node Introduction

Category: ✨Prompt Assistant

🔹 Translation Node

✨Prompt Assistant → Prompt Translation

🔹 Prompt Optimization Node

✨Prompt Assistant → Prompt Optimization

🔹 Image Caption Node

✨Prompt Assistant → Image Caption Prompt Captions images and optimizes editing instructions via visual models.

🔹 Video Caption Node

✨Prompt Assistant → Video Caption Prompt

📦 Installation

⚠️ Legacy Migration Warning

If you installed Prompt Assistant before V2.0.0, please backup the "config" folder in the plugin directory to avoid losing API keys, custom rules, and tag data!

If you installed via Manager, just update. For manual installation, it is recommended to delete the old plugin folder (backup "config" first!!), place the new folder in custom_nodes, and restore your config files.

Install via ComfyUI Manager

Search for Prompt Assistant in Manager and click Install.

Clone from Repository

Navigate to your custom nodes folder:
```
cd ComfyUI/custom_nodes
```

Clone the repo:

git clone https://github.com/yawiii/ComfyUI-Prompt-Assistant.git

Restart ComfyUI.

Download Zip

Download from Releases
Extract to ComfyUI/custom_nodes ⚠️ Note: It is recommended to rename the folder to "prompt-assistant" for ComfyUI compatibility.

Data Migration

The new version automatically upgrades and migrates API configs, custom rules, and tags. You can place the files you want to migrate in prompt-assistant/config. New configs are stored in ComfyUI\user\default\prompt-assistant.

⚙️ Configuration

Configuring API Keys and Models

Service Description

You can add custom providers or use built-in ones. ⚠️ Disclaimer: This plugin is a tool for API calls; responsibility for third-party services is independent of this plugin. User configs are stored locally.

Baidu Translate (MT): Baidu Translate Portal Fast but average quality. May require special network handling; 5M chars free per month.
Zhipu (LLM): Zhipu API Portal Fast and unlimited quota; Note: Strict censorship may return empty results.
xFlow-API Aggregation: xFlow API Portal Aggregates various models (Gemini, Grok, ChatGPT...) with a single API key; no networking issues.

🎀 Acknowledgments

Special thanks to our community for V2.0.0 rule templates: Adan, CJL, Normandy