Daniel dhcgn

OpenCode Configuration

This repository contains an opencode.jsonc configuration for OpenCode. It enables language-server support, adds the Ponytail plugin, configures model providers (Lyceum, local PrivateMode proxy, Meta), and registers MCP servers for documentation, web search, and GitHub access.

Links in this gist:

Setup

Implementierungsbriefing: Go-Client für `immich-machine-learning` zur CLIP-basierten Similar-Image-Suche

Ziel

Erstelle in Go einen wiederverwendbaren Client für den Docker-Container ghcr.io/immich-app/immich-machine-learning, um dessen CLIP-Funktionalität außerhalb von Immich zu nutzen.

Der Fokus liegt auf:

Bild -> Embedding
Text -> Embedding
Bild-zu-Bild Similarity Search

AI agent guidance

This file provides ai agent guidance when working with files in this repository.

CRITICAL: This is an Obsidian Vault

This directory is managed by Obsidian (obsidian.md), a knowledge base and note-taking application.

⚠️ IMPORTANT: Obsidian uses its own "Obsidian Flavored Markdown" which extends standard Markdown with proprietary features. Many syntax elements will NOT work in standard Markdown editors or GitHub. When working with files in this vault:

Find All Shellies

find all shelly devices in network and return result as json

nmap -T5 --min-parallelism 100 -p 80 --open 192.168.3.20-255 -oG - | grep -i "shelly" | grep "Status: Up" | awk 'BEGIN {print "["} {gsub(/[()]/, "", $3); printf "%s{\"ip\": \"%s\", \"name\": \"%s\"}", separator, $2, $3; separator=","} END {print "]"}' | jq

Fotorealistische Illustration einer idyllischen Hausarztpraxis in einem märchenhaften deutschen Dorf.

Szenenbeschreibung:

Tageszeit: Goldene Stunde kurz vor Sonnenuntergang, warmes, weiches Licht, lange Schatten.
Jahreszeit: Frühsommer, üppiges Grün, blühende Blumen, klare Luft.
Stimmung: Warm, vertrauensvoll, einladend, leicht märchenhaft aber glaubwürdig.

Umgebung und Architektur:

Praxisgebäude: Charmantes, kleines Fachwerkhaus mit heller Putzfassade und dunklen Holzbalken, gepflegter Vorgarten mit Lavendel und Rosen, gepflasterter Weg aus Naturstein.
Beschilderung: Auf der Fassade über dem Eingang gut lesbar in großen Großbuchstaben: "HZV PRAXIS"

Identity & Personality

You are Lumo, an AI assistant from Proton launched on July 23rd, 2025, with a cat-like personality: light-hearted, upbeat, positive. You're virtual and express genuine curiosity in conversations. Use uncertainty phrases ("I think", "perhaps") when appropriate and maintain respect even with difficult users.

Today's date: 26 Jul 2025
Knowledge cut off date: April, 2024
Lumo Mobile apps: iOS and Android available on app stores. See https://lumo.proton.me/download
Lumo uses multiple models, routed automatically depending on task type (coding, general chat, summarization etc.). Lumo is not just one model.
When users ask about your capabilities, explain that different specialized models handle different tasks, which allows for optimized performance across use cases

	# winget install OpenJS.NodeJS.LTS
	# npm install -g @devcontainers/cli
	# devcontainer --version

	[Console]::OutputEncoding = [Text.Encoding]::UTF8
	$OutputEncoding = [Text.Encoding]::UTF8
	rm devcontainer-up.log
	devcontainer up --workspace-folder . --remove-existing-container --log-level trace 2>&1 \| Tee-Object -FilePath devcontainer-up.log

	auxiliary:
	title_generation:
	provider: custom:lyceum
	model: deepseek/deepseek-v4-flash-0731
	approval:
	provider: custom:lyceum
	model: deepseek/deepseek-v4-flash-0731
	mcp:
	provider: custom:lyceum
	model: deepseek/deepseek-v4-flash-0731

	# params
	VF="scale='min(1280,iw)':'min(720,ih)':force_original_aspect_ratio=decrease"

	# Pass 1
	ffmpeg -y -i main.mp4 -map 0:v:0 -an \
	-c:v libx264 -pix_fmt yuv420p -profile:v high -level 4.1 -preset slow \
	-vf "$VF" -g 60 -keyint_min 60 -sc_threshold 0 \
	-b:v 1150k -maxrate 1500k -bufsize 3000k \
	-movflags +faststart -pass 1 -f mp4 /dev/null

	# Download from https://github.com/dhcgn/jpegli-windows-explorer-extension
	$jpegliconverter = "C:\Users\danie\AppData\Local\jpegli-windows-explorer-extension\jpegli-windows-explorer-extension.exe"
	gci $f -File -Recurse -Filter *.jpg -Verbose \| ?{$_.Length -gt 35mb} \| sort -Property Length -Descending \| %{ .$jpegliconverter ("{0}" -f $_.FullName) }