PromptDeMerde.com · Application · GitHub

Local AI prompt generator — Ollama, privacy, open source

PromptDeMerde.com (v1.21.0) reformulates sloppy instructions into clear prompts for ChatGPT, Claude, or your stack. Processing runs on Ollama on your machine — no cloud LLM, no account, zero telemetry.

Why PromptDeMerde

Unlike cloud optimizers (PromptForge, Prompt Guide), PromptDeMerde is fully local Ollama, MIT open source, unlimited, and telemetry-free. It reformulates your draft — ideal after dictation or paste.

Browser STT via Vosk, Whisper, or Parakeet. 12 UI languages, 50 themes, unlimited #Tag context prompts.

How it works

  1. Paste or dictate in the Workspace.
  2. Enable system and/or #Tag context prompts.
  3. Click Clean — Ollama reformulates.
  4. Copy the sniper prompt elsewhere.

On promptdemerde.com, use your local Ollama — no clone required. Try now →

Backup — ZIP profile archive

Settings → Profile JSON → Export downloads {profile}-promptdemerde-profile-v1.24.3.zip with pdm-config (49 keys), Markdown prompts and session. Import accepts .zip only.

Open source

GitHub — PHP self-host, no build step. After clone: cd install && bash restore-large-assets.sh for STT models.

FAQ

Is it free?

Yes. MIT open source, no subscription, no commercial quota. LLM is local Ollama.

Does data leave my network?

No cloud LLM. On the prod site, the browser talks to your local Ollama (OLLAMA_ORIGINS required). No server persistence, zero telemetry.

How to backup?

Settings → Profile JSON → Export: ZIP profile archive v1.21.0. Re-import via Import (.zip only).