Installation

Install Wild_Root_Prompt on Linux, macOS, Windows or Android

How to install Wild_Root_Prompt, the free open-source local LLM prompt engineering tool. Three install paths — double-click, one terminal command, or fully manual — plus a headless option and a compiled standalone app.

Wild_Root_Prompt needs two things: Python 3.8+ and Ollama with at least one model pulled. The installers set up both, plus a virtual environment, the dependencies, and a one-click launcher.

---

Option 1 — Double-click, no terminal

Platform	Double-click this file
macOS	Install WildRoot.command
Windows	Install WildRoot.bat

Both detect what is already present and only install what's missing. The Windows wrapper calls PowerShell with -ExecutionPolicy Bypass scoped to that single launch — it does not change the system-wide script execution policy.

When it finishes you get a day-to-day launcher (WildRoot.command, WildRoot.bat, or WildRoot) that opens the Web UI in one click.

---

Option 2 — One line in a terminal

Linux / macOS

git clone https://github.com/TFD-42/Wild_Root_Prompt.git && cd Wild_Root_Prompt && chmod +x install.sh && ./install.sh

Windows (PowerShell)

git clone https://github.com/TFD-42/Wild_Root_Prompt.git; cd Wild_Root_Prompt; powershell -ExecutionPolicy Bypass -File install.ps1

Android (Termux)

git clone https://github.com/TFD-42/Wild_Root_Prompt.git && cd Wild_Root_Prompt && chmod +x install_termux.sh && ./install_termux.sh

---

Option 3 — Manual install

If you'd rather control every step:

git clone https://github.com/TFD-42/Wild_Root_Prompt.git
cd Wild_Root_Prompt
python3 -m venv .venv
source .venv/bin/activate          # Windows: .\.venv\Scripts\Activate.ps1
pip install -r requirements.txt

Then install Ollama from ollama.com and pull a model:

ollama pull llama3.2:3b

Minimal / headless install

The CLI never imports Flask or the web server module unless you actually run the web subcommand. On a server, in CI, or on a constrained device, install only what the CLI needs:

pip install -r requirements-light.txt

That's just requests. Two dependencies stay optional and degrade gracefully:

Package	Needed for	If missing
flask	the local Web UI (web subcommand)	the web command exits with a one-line message; the CLI is unaffected
cryptography	at-rest encryption of session memory (encrypt_memory)	memory is stored in plaintext and the tool warns once

---

Choosing your first model

Any Ollama model works. Rough guidance:

Model size	RAM needed	Good for
3B (e.g. llama3.2:3b)	~4 GB	Fast iteration, Quick mode, low-power devices, Termux
7–8B (e.g. llama3, qwen2.5:7b)	~8 GB	The sweet spot for Full manifests
13B+	16 GB+	Deepest single-pass output, slowest

For parallel dual-model generation, pick two models with different temperaments — a systematic one and a creative one — so the synthesis pass has genuinely different material to merge.

On first run, if no model is installed, Wild_Root_Prompt offers a guided pull with RAM requirements shown per model.

---

Building a standalone app (single icon, no terminal)

After the installer has run once:

source .venv/bin/activate           # Windows: .\.venv\Scripts\Activate.ps1
pip install -r requirements-build.txt
python3 build_app.py

Platform	Produces
macOS	dist/Wild_Root_Prompt.app — drag to Applications
Windows	dist/Wild_Root_Prompt.exe
Linux	dist/Wild_Root_Prompt — a single binary; add a .desktop file for a menu entry

The compiled app needs no Python and no virtual environment at runtime. It starts Ollama if needed and opens the Web UI. Because app bundles are read-only (and code-signed on macOS), its writable data — settings, memory, cache, outputs — lives in the standard per-OS user data directory instead of next to the binary.

On macOS the first launch is blocked by Gatekeeper because the app isn't notarized — see Troubleshooting.

---

Verify the install

python3 prompt_expert_enhance.py generate "test" --mode quick --offline

If that prints an enhanced prompt, everything works. Next: Quick Start.

---

Next: Quick Start — your first output in two minutes · Examples — see what it produces · Troubleshooting if the install misbehaved.