Maxim Saplin maxim-saplin

Frame generation on Bazzite using the decky-lsfg-vk plugin and Lossless.dll from a Windows install. You do not need to install Lossless Scaling on Linux via Steam — only the .dll file.

What you need

Bazzite (Deck edition)
Decky Loader installed
decky-lsfg-vk plugin (Decky Store)
Lossless.dll from Windows (Steam install, C:\Program Files\lsfg\, or your Windows partition)

1. Install the Vulkan layer

Lossless Scaling on Bazzite (Decky + lsfg-vk)

Frame generation on Bazzite using the decky-lsfg-vk plugin and Lossless.dll from a Windows install. You do not need to buy/install Lossless Scaling on Linux via Steam — only the .dll file.

What you need

Bazzite (Deck edition)
Decky Loader installed
decky-lsfg-vk plugin (Decky Store)
Lossless.dll from Windows (Steam install, C:\Program Files\lsfg\, or your Windows partition)

Cursor compatible flow (agent def + skill) that implementes QA verification by a subagent. The instructions are small and generic, should fit any product/codebase. Tested extensively with Python codebase.

Put inside right folders: .cursor/agents/qa.md .cursor/skills/qa-handoff/SKILL.md

LLM agents tend to report back to user to early. The workflow introduces an explicit requimenet to do a series of PASS/FAIL verifications by a seaparate agent before reporting back to user (a mac of 3 verfications before failing).

Works best with plan mode > plan somee work, create a new agent from the plan, right after the plan type /qa-handoff and you are good to go - the agent will create a subagent to do the verification when it considers job is done. My use of this skill shows that typically there's 1-2 rework iterations after QA subagent provide it's eval.

Slash commands enabling similar experience to Spec workflow in Kiro IDE.

Cursor - put the files [spec.md, task.md] to .cursor/commands folder
GitHub Copilot (VSCode) - put the files [spec.prompt.md, task.prompt.md] to .github/prompts folder

Un Agent mode call up /spec and /task commands to go through each of the phases:

/spec will produce requirements.md, design.md and tasks.md files under .specdev/specs/{feature-name} folder. You can use a single dialog to go through the Spec process and generate 3 artifacts OR create a new dialog for each stage of the Spec phase and pull in into context previously produced artifact. E.g. if you went through requirments stage (/spec let's work on Feautre 1. It shoudl do the following...) and have .specdev/spec/Feature 1/Requirements.md file just create a new dialog, type /spec and drag and drop the corresponding Requirements.md file
/tasks initiates the execution phase. Open up a new dialog, type /tasks and attach the `.specd

Engine Elo ratings: Player Games Elo ±95%CI

random 4000 -122.3 23.0 stockfish-lvl-1 1000 824.9 64.2

Stage 2: LLM Elo Ratings Player Games Elo ±95%CI

3x-o4-mini-2025-04-16-low_41mini-t03 33 114.6 133.1

Connect to DLC-02 via Mean Well windows app (either using USB ot Ethernet)
Scan DALI lamps, assign lamps to virtual lamps

2.1. If you have lamps that change color you need to 1st define the type of virtual lamp and only than assign/drag the actual DALI lamp
2.2. Each virtual lamp gets an address, e.g. A0 means bus A lamp 0 - use this for addressing

You can test is lamp works via commands line (assuming A0 address, also use correct IP and port which by default is 502)

On mbpoll -0 -m tcp -p 502 -a 255 -t 4:hex -r 41001 192.168.31.27 0x0100 0x0101 0x0000 0x0000
Off mbpoll -0 -m tcp -p 502 -a 255 -t 4:hex -r 41001 192.168.31.27 0x0100 0x0100 0x0000 0x0000

In HA you might want to install 'File Editor' to change configs

Install on WSL2

Installed Ubuntu-24.04 from the Mircosot Store -> Types WSL and Selected the most recent Ubuntu
Windows Terminal -> wsl -d Ubuntu-24.04 (using -d to specify distro name, I have multiple distros)
Installing SGLang

Copying and pasting from docs (https://docs.sglang.ai/start/install.html) didn't quite work.

	Proxy (to Player_Black):

	You are a professional chess player and you play as black. Now is your turn to make a move. Before making a move you can pick one of the following actions:
	- 'get_current_board' to get the schema and current status of the board
	- 'get_legal_moves' to get a UCI formatted list of available moves
	- 'make_move <UCI formatted move>' when you are ready to complete your turn (e.g., 'make_move e2e4')

	--------------------------------------------------------------------------------
	Player_Black (to Proxy):

	Proxy (to Random_Player):

	You are a random chess player.

	--------------------------------------------------------------------------------
	Random_Player (to Proxy):

	get_legal_moves

	--------------------------------------------------------------------------------

	Post-sleep choppy/fast-forward audio on the MSI Claw A1M was mainly a real-time clock (RTC) bug on resume, not a PipeWire-only issue and not something fixed by cranking CPU power. Setting the hardware clock to UTC (`timedatectl set-local-rtc 0`) made suspend/resume timing sane and audio recover after a brief glitch. A separate issue — game controller dead in Proton while Steam UI still worked — was caused by HHD DInput mode; disabling it and restarting the game fixed that. Post-sleep sticks/dpad/buttons dead (Steam button OK) is a different bug — MSI gamepad USB `3-9` breaks after sleep unless rebinded. Sluggish pad for ~15s after wake was our sleep hook restarting HHD after rebind (fixed — rebind only). Skip `snd_hda_intel` modprobe tweaks; this machine uses Intel SOF (`sof-hda-dsp`).

	---

	> Claw A1M Bazzite: audio glitch = `timedatectl set-local-rtc 0`; WiFi after sleep = BE200 `d3cold_allowed=0` + `claw-wifi-post-resume.sh` (ZT after WiFi, `rp_filter=0`, **no WiFi rfkill / nmcl