Prompt Optimizer — Web App

A lightweight Flask web app that helps users craft clearer, more effective prompts for AI tools (ChatGPT, Claude, Gemini, etc.). It offers a Quick mode for one‑shot optimization and an Interactive mode that guides users through a few questions to build a high‑quality final prompt.

---

Features

• Quick mode: Paste an idea, pick a tone (Standard / Technical / Informal or Custom), and get an optimized prompt with one‑click Copy.
• Interactive mode: Answer guided questions → receive tailored follow‑ups → get a final optimized prompt (with Copy).
• Privacy page: Clear, minimal policy (no account; no PII stored beyond what’s needed to generate the result).
• Feedback link: Optional post‑result form so users can share quick feedback.
• Chrome extension link: Homepage promotes the companion browser extension.

Tech stack

• Backend: Flask, flask‑cors
• AI layer: Agno agent + Groq model (e.g., qwen/qwen3-32b) calling an external MCP server over SSE
• Frontend: Jinja templates, Bootstrap‑style UI
• Process manager (deploy): gunicorn

Directory layout (example)

augustosouza8-prompt-optimizer-web-app/
├── README.md
├── LICENSE
├── requirements.txt
├── run.py
└── app/
    ├── __init__.py
    ├── agno_agent.py
    ├── routes.py
    └── templates/
        ├── base.html
        ├── index.html
        ├── quick.html
        ├── interactive_step1.html
        ├── interactive_step2.html
        ├── interactive_result.html
        └── privacy.html

Requirements

Install Python dependencies:

pip install -r requirements.txt

Key libraries: Flask, flask-cors, agno, groq, mcp, gunicorn.

Configuration

Create a .env file (or set environment variables):

# Required for the agent/model call
GROQ_API_KEY=your_key_here

# Optional
SECRET_KEY=change_me
PORT=8000

• The MCP SSE endpoint is configured in app/agno_agent.py (look for the _SSE_URL). Update it if you host your own MCP server.

Run locally

python run.py
# App binds 0.0.0.0 and uses PORT (default 8000)

Open: http://localhost:8000/

Common routes:

• Home: /
• Quick: /quick
• Interactive: /interactive → follow‑ups → result
• Privacy: /privacy

Deploy

With Gunicorn (e.g., Render/Heroku‑style):

gunicorn run:app --bind 0.0.0.0:${PORT:-8000}

How it works (high level)

1. The UI posts user input to Flask.
2. The server calls query_agent(...), which uses Agno with Groq (e.g., qwen/qwen3-32b) and an MCP tool over SSE to produce the optimized prompt.
3. The result is rendered back to the page with a Copy helper.

Privacy

• No accounts. No PII storage. Prompts are sent only to generate the optimized result. Session cookies are temporary. See templates/privacy.html.

License

MIT — see LICENSE.