Pliable Pixels pliablepixels

From https://github.com/pliablepixels/slop-mop

Personal rules

Last updated: 2026-10-07. Sources: 1,601 GitHub issue comments by pliablepixels on ZoneMinder/zmNinja, 2015 to 2021, about 57,000 words after removing code blocks and quoted replies. All are support replies, so these rules fit issue comments, PR comments, and casual email best, and may fit READMEs and long docs less well.

Rules

Keep replies short: median comment 22 words, median sentence 13 words. A diagnosis or explanation runs 60 to 110 words, like the samples below. A sentence can run long when it explains a chain of causes. Example: "Closing as ZM updates fixed it."
Use a spaced hyphen ( - ) as a dash, often more than once per sentence. Never use an em dash. Example: "Folks, my push server is down - I am on travel till 17th of Jan - not quite sure what happened". [overrides: No dashes] [overrides: One idea per sentence]
Use --> to mean "which shows" or "so". Example: "I found the bug --> it was to do with my http injector code

1. Install

a) Same box (ZM and gateway together) One install, with the serve extra:

sudo /opt/zoneminder/venv/bin/pip install --force-reinstall --no-deps \
  "pyzm[serve] @ git+https://github.com/ZoneMinder/pyzmNg.git@master"

b) Separate boxes ZM box — client only, no extra:

Build instructions

1. Prerequisites

brew install cmake pkg-config numpy

2. Clone repos

git clone https://github.com/opencv/opencv

These are the things I did to get NVIDIA GeForce GTX 1050 Ti working on Ubuntu 24. Given its an older GPU card, I had to use a lower CUDA version and a lower gcc version. Your system may be different. Don't try and figure it out yourself. Easier to use claude CLI and debug together:

IMPORTANT: Don't copy paste - find out your GPU specs, and then download the right nvidia driver/CUDA/CuDNN/

wget https://developer.download.nvidia.com/compute/nvidia-driver/590.48.01/local_installers/nvidia-driver-local-repo-ubuntu2404-590.48.01_1.0-1_amd64.deb
sudo dpkg -i nvidia-driver-local-repo-ubuntu2404-590.48.01_1.0-1_amd64.deb
wget https://developer.download.nvidia.com/compute/cuda/11.8.0/local_installers/cuda_11.8.0_520.61.05_linux.run

API V3 Migration Review

1. Architecture Overview

Current API (v1/CakePHP)

Base path: /api/ with .json suffix on all endpoints
Auth: Token passed as query parameter (?token=xxx)
Request format: Form-encoded POST bodies (Monitor[Function]=Modect)
Response wrapping: Objects nested inside { monitors: [{ Monitor: {...}, Monitor_Status: {...} }] }
Naming: PascalCase field names (MonitorId, StartDateTime, CaptureFPS)

	1. The 7pm flood is the rules file I gave him, working as written

	My file only muted non-person events between 7:00 am and 6:59 pm. Nothing muted anything after 7 pm. So at 7:00 pm the mute window closed and every event — cars, animals, plain motion — started notifying. I flagged this in passing ("a car at 10pm still sends a normal notification") but I should have just put the night mute in the file. That's my error, not a bug in his setup.

	His workaround, one all-day mute on non-person, is the right shape. It just drops the critical escalation.

	2. cause_has: "person" really does match "no Person"

	Rules.pm:131-132 lowercases the cause and matches the pattern with /i as a plain substring regex. There's no word boundary and no negation awareness. I tested the patterns against realistic cause strings:

	{"ts": "2026-04-30T02:00:51Z", "level": "INFO", "method": "GET", "path": "/api/orders/1292", "status": 200, "latency_ms": 90, "ip": "10.0.2.185", "user_id": "u_1039", "user_agent": "Mozilla/5.0 (iPhone; CPU iPhone OS 17_2 like Mac OS X) Safari/604.1", "response_bytes": 991}
	{"ts": "2026-04-30T02:02:00Z", "level": "INFO", "method": "GET", "path": "/api/health", "status": 200, "latency_ms": 2, "ip": "10.0.3.64", "user_id": null, "user_agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) Chrome/120.0.0.0", "response_bytes": 74}
	{"ts": "2026-04-30T02:03:34Z", "level": "INFO", "method": "GET", "path": "/api/products", "status": 200, "latency_ms": 49, "ip": "10.0.1.175", "user_id": "u_1024", "user_agent": "Mozilla/5.0 (iPhone; CPU iPhone OS 17_2 like Mac OS X) Safari/604.1", "response_bytes": 5214}
	{"ts": "2026-04-30T02:05:27Z", "level": "INFO", "method": "GET", "path": "/api/products/6066", "status": 200, "latency_ms": 55, "ip": "10.0.1.25", "user_id": "u_1017", "user_agent": "Mozilla/5.0 (Macintosh; Intel Mac OS X

	import numpy as np

	import sys
	import cv2
	import time
	import datetime
	import re
	from pyzm.helpers.Base import Base
	import portalocker
	import os

	#!/bin/bash

	# claude-runner.sh - Runs Claude CLI with automatic rate-limit handling
	# Uses tmux to maintain proper terminal environment

	set -euo pipefail

	SESSION_NAME="claude-worker"
	RESUME_FILE="$HOME/.claude-session-id"
	STATE_FILE="$HOME/.claude-runner-state"

	## 👤 User

	Analyze agents.md to make sure its in great shape. One key issue I see is you implement things, run tests, say its all ok, but it breaks when I try. Are you generating/updating tests always? If not, how shoulD AGENTS.md be improved? I really want to follow a "test for everything" philosophy - when tests pass, it really should mean the app should work and all options/buttons/screens tested

	---

	## 🤖 Claude

	I'll analyze AGENTS.md and identify the testing gaps. Let me first check what tests currently exist: