Local Slides RAG

This repository contains only the minimal local RAG app and the indexed Markdown chunks needed to run it.

What Is Included

• rag_app/: Python Streamlit RAG app.
• data/rag-chunks/: Markdown chunks used as the retrieval corpus.
• requirements.txt: Python dependencies.
• run_rag_app.ps1: Windows helper script.

Local Setup

Install dependencies:

pip install -r requirements.txt

Install Ollama, then pull the default small model:

ollama serve
ollama pull qwen3:1.7b

Build The Vector Index

python rag_app/indexer.py --build

Expected current count:

3101 chunks

Check status:

python rag_app/indexer.py --status

Run The App

streamlit run rag_app/app.py

Then open:

http://localhost:8501

On Windows you can also run:

powershell -ExecutionPolicy Bypass -File rag_app/run_rag_app.ps1

Smoke Tests

Test indexing and retrieval:

python rag_app/smoke_test.py

Test retrieval plus Ollama generation:

python rag_app/smoke_test.py --with-ollama

كيف تسحبه و تشغله عندك مباشرة

بس اعمل copy و paste لهاي جوى الterminal ملاحظة افتح الterminal من جوى الfolder تبع المشروع

git clone https://github.com/any777777/RAG.git
cd RAG
pip install -r requirements.txt
ollama serve
ollama pull qwen3:1.7b
python rag_app/indexer.py --build
streamlit run rag_app/app.py

بعدين رح يفتح معك هون

http://localhost:8501

لازم يكون مثبت عندك

• Python مثبت.
• Ollama مثبت.
• اتصال إنترنت في أول تشغيل لتحميل:
  • نموذج Ollama qwen3:1.7b
  • نموذج embeddings من Hugging Face.
• بعد أول بناء للفهرس، التطبيق يعمل محلياً من data/rag-chunks. و غالبا انهم عندك

Notes

• The app builds Chroma locally in rag_app/.chroma/.
• rag_app/.chroma/, logs, and Python caches are intentionally ignored by Git.
• The default model is qwen3:1.7b because it is light enough for an 8GB RAM machine.
• Arabic questions work in retrieval, but answer generation quality depends on the local Ollama model.