Quick Start

Quick Start — your first enhanced prompt in 2 minutes

Assumes Installation is done and Ollama is running with at least one model.

---

Path A — the Web UI (no terminal)

1. Double-click WildRoot.command (macOS), WildRoot.bat (Windows), or run ./WildRoot (Linux).
2. Your browser opens on http://localhost:7860.
3. Type your task, pick a model, press Generate.

Full walkthrough: Web UI.

---

Path B — the interactive menu

source .venv/bin/activate      # Windows: .\.venv\Scripts\Activate.ps1
python3 prompt_expert_enhance.py

You get a numbered menu — no flags to memorize:

==============================================================
   WILD_ROOT_PROMPT  —  Expert Prompt Enhancement Tool  v2.3
==============================================================
   Model A     : llama3:latest
   Model B     : qwen2.5:7b
   Synthesis   : qwen2.5:7b
   Temperature : 0.3
   Techniques  : 15 active / 173 available
   Internet    : ON   Web enrichment : ON   Streaming : ON
--------------------------------------------------------------

  1.  Single generation         (1 model, streaming)
  2.  Parallel generation       (2 models, split screen)
  3.  Full pipeline             (parallel + synthesis)
  4.  Synthesize 2 files

  5.  Configure models
  6.  Configure techniques
  7.  Browse available techniques
  8.  Advanced settings          (temperature, timeout, url, web, stream)

  9.  View memory
  10. Clear memory

  0.  Quit

Not sure which job you're doing? Use Cases maps common goals to commands. Start with 1. When it asks for a model, it lists everything installed locally with size and date — type a number, or type a name to pull a new one.

---

Path C — one-shot commands

# Quick mode: one enhanced prompt, ready to paste anywhere
python3 prompt_expert_enhance.py generate "Design a REST API for a bookstore" --mode quick

# Full mode: the complete 12-section instruction manifest
python3 prompt_expert_enhance.py generate "Design a REST API for a bookstore" --mode full

# Two models in parallel, split screen
python3 prompt_expert_enhance.py parallel "Design a REST API" --model-a llama3 --model-b qwen2.5:7b

# Parallel + synthesis in one go
python3 prompt_expert_enhance.py full "Design a REST API"

Every flag is documented in the CLI Reference.

---

Quick mode vs Full mode

	Quick	Full
Output	One enhanced prompt	A 12-section manifest
Length	A paragraph to a page	Several pages
Use it for	Pasting into another LLM	Handing a complete brief to an agent
Speed	Seconds on a 3B model	Minutes

Start in Quick while you're iterating on the wording of your task; switch to Full when the wording is right and you want the full deliverable. --draft gives you only sections 1–2 of a Full manifest, which is the fastest way to sanity-check direction before paying for the whole generation.

---

Three things worth trying immediately

1. A slash metacommand. Prefix your task with modifiers:

python3 prompt_expert_enhance.py generate "/expert /tableau /sources compare message queues" --mode quick

That's expert voice, table format, sourced claims. There are 60 of them — see Slash Metacommands.

2. A technique bundle. Instead of picking technique IDs by hand:

python3 prompt_expert_enhance.py generate "audit this auth flow" --techniques "bundle:Audit / securite"

See Templates and Bundles.

3. Auto-recommended techniques. Let the tool read your task and choose:

python3 prompt_expert_enhance.py generate "audit this codebase for vulnerabilities" --recommend-techniques

---

Where output goes

Generated files land in outputs/ (gitignored). Session history lands in memory/ and is fed back as context on later runs for cross-session coherence — clear it any time with python3 prompt_expert_enhance.py memory clear.

---

Next: Examples — real before/after runs · Slash Metacommands — reshape the output · CLI Reference — every flag.