Brief-Decoder: The Local Gemma AI That Translates "Make It Pop" Into Actionable Specs (So You Don't Lose Your Mind) 🛡️🦥
This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady
What I Built
Every freelancer knows this message:
"Bhai, design theek hai par thoda sa modern vibe aur pop chahiye. Colors vibrant kar do aur homepage pe ek naya banner daal do."
It is never small. Vague feedback like "make it pop" or "vibe mismatch" forces freelancers into endless, unpaid revision loops because guessing what a client means takes hours, and asking for clarification often feels awkward.
Brief-Decoder is a local-first, privacy-focused AI tool that acts as your professional buffer. You paste the client's vague feedback and select your current revision round, and it:
- Translates subjective jargon into concrete technical items (e.g., converting "make it pop" into contrast adjustments, font-weight changes, and specific padding updates).
- Generates an actionable checklist ready to copy directly into your IDE or Figma task board.
- Auto-drafts professional replies ranging from polite clarifications to graceful extra-charge notices.
- Sloth Mode Payoff: Saves you hours of frustrating back-and-forth so you can channel your inner sloth and protect your energy.
** Demo**
Brief-Decoder features a sleek dark-mode UI built with Streamlit, powered by Google's Gemma 2 (gemma2:2b) running locally via Ollama, and launched with a zero-friction Windows batch script (start.bat).
-
Local Execution Snapshot:
>
http://localhost:8501— Running 100% locally with zero cloud API latency or dependency.
Code
The project structure is minimal and self-contained:
BriefDecoder/
├── main.py # Streamlit UI & local Gemma 2 prompt handler
└── start.bat # Zero-command automated launcher script
How I Built It
Open-Weight Model: Google's Gemma 2 (gemma2:2b) running locally via Ollama. Gemma 2's compact parameter footprint allows lightning-fast instruction following and structured JSON/Markdown extraction on standard laptop hardware.
Frontend Framework: Streamlit for a responsive, modern dark-mode glassmorphism interface.
Zero-Terminal Setup: A custom Windows .bat launcher (start.bat) that verifies environment dependencies and launches the local web server with a single double-click—no CLI commands needed.
Prompt Engineering & Structuring:
The system prompt forces Gemma 2 to act as a strict freelance scope buffer.
It enforces clear, structured output sections: Vibe-to-Spec Translation, Actionable Checklist, and Client-Facing Reply Drafts.
Output parsing handles potential model hiccups gracefully, ensuring clean presentation every time.
_
Why Does Open Innovation Matter?
_
Client communications, contract details, and rough project briefs are strictly confidential. Relying on closed cloud APIs presents real challenges:
Absolute Data Privacy: Sending sensitive client messages and project specs to third-party cloud servers violates client trust and NDAs. Running Gemma 2 locally guarantees that private communications never leave your laptop.
Zero Operating Costs ($0 API Fees): Closed models charge per token. A local open-weight model lets freelancers parse hundreds of vague feedback notes without worrying about recurring cloud bills.
100% Offline Capability: Freelancers often work on the go from cafes or during travel. Brief-Decoder requires zero internet connection to decode feedback and generate responses.
My Agent Session
I built and refined Brief-Decoder collaboratively using my AI coding companion to rapidly prototype the UI layout, craft the local Ollama integration, and optimize the Gemma 2 prompts.
Key Iterations & Enhancements:
Designed a dark glassmorphism theme with custom CSS metric badges for high visual appeal.
Added a Revision Round Selector in the sidebar to dynamically adjust response tone based on contractual limits.
Streamlined the Windows launch script (start.bat) so non-technical users can run the app without touching the terminal.
The Lazy Payoff
Vague feedback like "make it pop" or "colors vibrant kar do" used to trigger 20-30 minutes of back-and-forth messaging or hours of guessing and re-doing work.
Time Saved: Brief-Decoder takes an unstructured client text and decodes it into precise technical tasks in under 5 seconds.
Energy Saved: You no longer need to spend emotional energy drafting awkward pushback emails or worrying about sounding confrontational when a client asks for unpaid extra work.
Real-World Impact: In testing with a freelance UI designer, it converted a vague request ("make the homepage feel more energetic") into three concrete design changes (contrast tweaks, font weight updates, CTA positioning) instantly—saving over 25 minutes of guesswork per revision.
Prize Categories
Best Use of Gemma: The core intelligence behind Brief-Decoder runs entirely on Google's Gemma 2 (gemma2:2b) locally via Ollama.
Top comments (0)