HyperMemory ComfyUI Extract Text Node

ComfyUI's Extract Text node, with HyperMemory brand context and OpenRouter prompt refinement behind the scenes.

What It Does

Extract Text - Hypermemory keeps the same simple shape as ComfyUI's built-in Extract Text node:

• visible inputs: string, regex_pattern, mode
• advanced regex settings: case_insensitive, multiline, dotall, group_index
• output: one STRING

The node first extracts text using the normal regex behavior. If HyperMemory and OpenRouter are configured outside the workflow, it uses the extracted text to recall brand context from HyperMemory and asks OpenRouter to turn it into an on-brand prompt.

If credentials are not configured, it falls back to plain extracted text.

Install

Clone this repo into your ComfyUI custom nodes folder:

cd ComfyUI/custom_nodes
git clone https://github.com/RunStack-AI/hypermemory-cumfyui-extract-text-node.git

Restart ComfyUI. The node appears as:

text/HyperMemory -> Extract Text - Hypermemory

No extra Python packages are required.

Configuration

Do not put API keys into workflow nodes. Set them in the environment before starting ComfyUI:

export HYPERMEMORY_API_KEY="hm_..."
export OPENROUTER_API_KEY="..."

Optional environment settings:

export HYPERMEMORY_API_URL="https://api.hypermemory.io"
export HYPERMEMORY_MAX_RESULTS="10"
export OPENROUTER_MODEL="minimax/minimax-m3"
export OPENROUTER_TEMPERATURE="0.4"
export OPENROUTER_MAX_TOKENS="900"

No graph ID is required. HyperMemory resolves the graph from the hm_... API key.

Brand details should live in HyperMemory. Store voice, audience, palette, product rules, campaign history, and visual constraints in the graph instead of entering them in the node UI.

How To Use

1. Add The Node

In ComfyUI, right-click the canvas and add:

text/HyperMemory -> Extract Text - Hypermemory

2. Add Source Text

Paste or connect the rough text into string.

Example:

Campaign brief: 15 second teaser for the new running shoe line.
Focus on speed, city lights, and premium performance.

3. Extract The Useful Part

Set regex_pattern and mode.

Example regex:

Campaign brief:\s*(.*)

Recommended mode:

First Group

Extracted text:

15 second teaser for the new running shoe line.

Common patterns:

Goal	Regex	Mode
Text after a label	Brief:\s*(.*)	First Group
Every hashtag	#(\w+)	All Groups
First quoted phrase	"([^"]+)"	First Group
First URL	https?://\S+	First Match
Numbered lines	^\d+\.\s*(.*)	All Groups with multiline on

4. Wire The Output

The node has one output. Use it where prompt text belongs:

Extract Text - Hypermemory -> CLIP Text Encode.text

With HYPERMEMORY_API_KEY and OPENROUTER_API_KEY configured, that output is the refined brand prompt. Without them, it is the extracted text.

Inputs And Output

Name	Type	Notes
string	input	Source text.
regex_pattern	input	Regex pattern.
mode	input	Defaults to First Group.
STRING	output	Extracted text or refined prompt.

Advanced regex controls match the source Extract Text node: case_insensitive, multiline, dotall, and group_index.

Failure Behavior

• Regex does not match: output is empty, matching the source node.
• HyperMemory is not configured or fails: the node continues without memory context.
• OpenRouter is not configured or fails: the node returns the best local output available.

Development

Run the local checks:

python3 -m py_compile core.py nodes.py __init__.py tests/test_core.py
python3 -m unittest discover -s tests -v

Project Structure

.
├── __init__.py        # ComfyUI exports
├── core.py            # extraction, HyperMemory, OpenRouter, prompt helpers
├── nodes.py           # ComfyUI node definition
├── requirements.txt   # no external dependencies
└── tests/
    └── test_core.py

License

MIT. See LICENSE.