maheshsmc/evaluateday5

title: Evaluate Day 5 emoji: 📊 colorFrom: gray colorTo: blue sdk: gradio app_file: app.py pinned: false license: mit ---

Day 5 — RAG Evaluation Harness (Gradio)

This Space provides a small web UI to run the Day‑5 evaluator:

• —Upload your gold_qa.json (list of objects with question, answers)
• —Upload your corpus.csv (columns: doc_id,text)
• —Click Run evaluation
• —See per‑query Recall@1/3/5, MRR, nDCG and overall metrics

File formats

gold_qa.json

[
  {"question": "Who is the CEO of Tesla?", "answers": ["Elon Musk", "Musk"]},
  {"question": "What is the capital of France?", "answers": ["Paris"]}
]

corpus.csv

doc_id,text
doc_001,Elon Musk is the CEO of Tesla and SpaceX.
doc_002,Paris is the capital of France, famous for the Eiffel Tower.

Notes

• —If you don't upload files, the app will fall back to the tiny toy dataset bundled in data/ (good for a quick demo).
• —Metrics are computed with a BM25‑lite scorer implemented in pure Python (no heavy deps).
• —Results are shown in the table and can be downloaded from the UI.