🐳Prompt管理器星球版

The 🐳Prompt管理器星球版 (CustomPromptManagerEnhanced) is the "planet edition" - the grown-up version of the pack's base prompt manager. It takes a JSON array of prompts with per-item toggles and weights, sorts them into positive and negative, applies the weights, and - here's the new bit - filters out any tag you list. If you maintain prompts as data (which is the only sane way once a workflow has dozens of them), this is the node that turns that data into actual prompt strings.

The core mechanism is a JSON parser that's surprisingly forgiving. You feed it a list of objects, each with a text, an enabled flag, a weight, and a type of positive or negative. The node skips anything disabled, wraps weighted items in (tag:1.2) syntax when use_weights is on, and joins each polarity group with your chosen separator. Output comes as two strings: positive_prompt and negative_prompt, ready to feed CLIP Text Encode.

What "Enhanced" adds

The base version (CustomPromptManagerWithNegative) is the same JSON-in/two-strings-out deal. This one layers on:

• enable_tags_filter - turn on, and any comma-separated tag in filter_tags gets dropped from the output entirely. This is the "oops, this workflow shouldn't produce cleavage" or "remove all mentions of X across every variant" switch - a global veto that's far less fiddly than editing each prompt entry.
• filter_tags - the list those tags live in.

The tag filter alone justifies the upgrade if you ever run the same prompt bank across different content policies or models. Everything else is the same solid plumbing.

The inputs

• prompts_json - the JSON array (default []). Format: [{"text": "...", "enabled": true, "weight": 1.2, "type": "positive"}, ...].
• separator - default ", "; how tags join.
• use_weights - boolean, default on. Wraps non‑1.0 weights in (tag:weight).

No required inputs at all - every field is optional, so an empty node is harmless. Outputs: positive_prompt and negative_prompt (STRING).

Install

From comfy_Pond_Nodes (Manager search, or git clone https://github.com/Pondowner857/comfy_Pond_Nodes + pip install -r requirements.txt). No models, no heavy deps - it's a pure JSON/string node.

Where people get stuck: it's not a dynamic-prompt randomizer, despite the name "管理器" implying more. It doesn't generate, it organizes. If you expected wildcard syntax, this isn't it - but if you've got a big prompt bank that needs a clean positive/negative split with weights and a global tag veto, this is the node in this pack that does it.