dev-tools

Internal tooling for AI-assisted workflows: push-to-talk dictation, batch transcription/summarization, and Claude Code utilities.

Tools

Tool	Description
bootstrap	One-shot provisioning for a new Linux machine (Raspberry Pi / Ubuntu): base packages, Node.js, Tailscale, SSH key + git repos, and the dev-tools — all from one script
voice-dictation	Push-to-talk voice transcription via faster-whisper → auto-paste (Windows + Linux)
summarize-recording	Transcribe and summarize audio recordings via Azure OpenAI
claude-ctx-statusline	Shows context window usage in the Claude Code status bar
keep-awake	Keeps a Windows session awake (F15 tap + execution-state) so it never idle-logs-out; auto-starts at logon
termhub	Mobile-friendly web terminal over Tailscale — one server per machine; open/reattach shell sessions, incl. Claude Code (Windows + Linux)
disk-janitor	Safe, scheduled reclamation of rebuildable disk space (package caches + aged temp/cache); whitelist-only, never touches user data (Windows + Ubuntu/Debian)

Quick start

Clone:

git clone https://github.com/xvanov/dev-tools.git
cd dev-tools

Bootstrap a new Linux machine (Raspberry Pi / Ubuntu)

Provision a fresh box end-to-end — base packages, Node.js, Tailscale, an SSH key

• your git repos, and the dev-tools — with one self-contained script. Flash Ubuntu, then:

scp bootstrap/pi-setup.sh <user>@<pi>:~/
ssh <user>@<pi>
TS_AUTHKEY=tskey-auth-xxxx ./pi-setup.sh

See bootstrap/README.md for options. Re-runnable and idempotent.

Voice dictation (Windows)

cd voice-dictation
Set-ExecutionPolicy -Scope Process Bypass
.\windows\install.ps1

Press Ctrl+Alt+V to record. Recording auto-stops on silence; transcript is cleaned up and pasted at the cursor.

Voice dictation (Linux)

cd voice-dictation
./linux/install.sh

Bind ~/.local/voice-dictation/voice-toggle.sh to Ctrl+Alt+V in keyboard settings.

summarize-recording

cd summarize-recording
python -m venv .venv
source .venv/bin/activate   # Windows: .\.venv\Scripts\Activate.ps1
pip install -r requirements.txt
# Set AZURE_OPENAI_* or AZURE_FOUNDRY_* in repo-root .env
python summarize-recording.py run recording.mp3

Claude context status line

cd claude-ctx-statusline
.\install.ps1

Restart Claude Code to see Ctx: 30k/200k (15%) in the status bar.

keep-awake (Windows)

cd keep-awake
Set-ExecutionPolicy -Scope Process Bypass
.\install.ps1

Starts immediately and auto-starts at every logon (hidden). Remove with .\install.ps1 -Uninstall.

termhub (web terminal over Tailscale)

Install on each machine you want terminal access to:

cd termhub
./linux/install.sh        # Linux (systemd user service)
# Windows: .\windows\install.ps1  (elevated PowerShell)

Then open http://<machine-tailscale-ip>:7000 in a browser tab per machine (works on your phone). Open terminals, switch between them, scroll back, and reattach after disconnects. Mobile-friendly, with an on-screen key bar (Ctrl-C, arrows, Esc, Tab) and a one-click Claude Code launcher.

Repo layout

dev-tools/
├── bootstrap/                # provision a new Linux machine (Pi/Ubuntu) end-to-end
├── voice-dictation/          # real-time dictation (recorder + overlay + warm server)
├── summarize-recording/      # transcribe + Azure summarization CLI
├── claude-ctx-statusline/    # Claude Code statusLine helper
├── keep-awake/               # Windows keep-session-awake helper
└── termhub/                  # web terminal over Tailscale (agent + hub + xterm.js UI)

Each tool has its own README, requirements, and install scripts.