RAG Application with Gemma(Open-Source Contributor Guide)
About the project
π Nova: Open-Source Contributor Assistant (Problem Statement 2B)
Nova is an intelligent, RAG-powered developer assistant designed to help first-time open-source contributors easily navigate complex repositories, understand setup instructions, and start contributing during Hacktoberfest.
π οΈ The Problems We Faced (Challenges & Pain Points)
-
Scattered Documentation: Essential project setup instructions and contributing guidelines are fragmented across multiple files (
README.md,CONTRIBUTING.md), making it overwhelming for newcomers. - Starter Kit Limitations: Default starter kits only supported single PDFs and lacked concurrent multi-Markdown parsing.
- AI Hallucinations: Standard LLMs often generated fake setup commands or inaccurate project details when asked questions outside their training scope.
π‘ How We Solved Them (Technical Solutions)
-
Multi-File Ingestion & Dynamic Tagging: Upgraded the pipeline to parse multiple
.md,.txt, and.pdffiles concurrently with source file-tagging ([File: README.md]). -
Strict Hallucination Guardrails: Enforced rigorous system prompts (
ANSWER_RULES) restricting Google Gemma 4 to strictly use provided context. - Minimalist SaaS UI/UX: Built a clean, modern interface using Gradio with a monochrome theme.
- Automated Onboarding Checklist (Bonus 2B): Implemented a specialized βWhere do I start?β feature for instant contribution roadmaps.
π Tech Stack
- LLM Core: Google Gemma 4
-
Embeddings:
sentence-transformers(all-MiniLM-L6-v2) - Interface: Gradio (Python Web Framework)
- Environment: Google Colab
Built with
geminigemma
AI tools
Coded withCodexAntigravityGemini 3.8 FlashGemini 3.5 Flash Lite
Prize Categories Entered
- Submitted
Best Use of Gemma 4
- Submitted

Best Open-Source AI Project
- Event prizeSubmitted
Theme 2: RAG Application with Gemma