bharatrag 0.1.0

RAG evaluation library for Indian languages (Hindi, Marathi)

bharatrag

Open-source RAG evaluation library for Indian languages (Hindi, Marathi). pip install bharatrag

BharatRAG 🇮🇳

RAG Evaluation Library for Indian Languages

BharatRAG is the first open-source RAG evaluation library built specifically for Indian languages (Hindi and Marathi).

Existing tools like RAGAS are built and tested on English data. BharatRAG fills the gap — giving developers a reliable way to measure RAG quality in Indic languages.

---

The Problem

RAG (Retrieval Augmented Generation) systems are being deployed across India for:

• Government scheme chatbots (PM Kisan, Ayushman Bharat)
• Health information systems in regional languages
• EdTech platforms for vernacular learners
• Banking and insurance customer support

But there is no standard way to evaluate whether these systems are actually working correctly in Hindi, Marathi, or other Indian languages.

RAGAS — the most popular RAG evaluation tool — uses English-first embedding models that produce unreliable scores for Indic text.

BharatRAG solves this.

---

What it measures

BharatRAG computes the RAG Triad in Hindi and Marathi:

Metric	Question it answers
Context Relevance	Did we retrieve the right documents?
Groundedness	Is the answer based on the context, or hallucinated?
Answer Relevance	Does the answer actually address the question?

---

Benchmark Results

BharatRAG evaluated on 20 Hindi + Marathi QA pairs across government schemes, health, and agriculture domains:

Metric	Hindi ✅ Correct	Hindi ❌ Hallucinated	Marathi ✅ Correct	Marathi ❌ Hallucinated
Context Relevance	0.4793	0.4793	0.4327	0.4327
Groundedness	0.9167	0.7000	0.9667	0.2000
Answer Relevance	0.6221	0.5417	0.5072	0.2959
Overall	0.6727	0.5737	0.6355	0.3095

BharatRAG correctly scores hallucinated answers lower than correct answers in both languages. Marathi hallucination detection shows a 2x difference in overall score (0.6355 vs 0.3095).

---

Installation

pip install bharatrag

---

Quick Start

from bharatrag import evaluate

results = evaluate(
    questions=["पीएम किसान योजना में कितने रुपये मिलते हैं?"],
    contexts=[[
        "पीएम किसान सम्मान निधि योजना के तहत किसानों को",
        "प्रति वर्ष 6000 रुपये तीन किश्तों में मिलते हैं।"
    ]],
    answers=["पीएम किसान योजना में किसानों को 6000 रुपये मिलते हैं।"],
    language="hindi"
)

print(results)
# {
#   'context_relevance': 0.72,
#   'groundedness': 1.0,
#   'answer_relevance': 0.66,
#   'overall': 0.79,
#   'language': 'hindi',
#   'num_questions': 1
# }

Marathi support

results = evaluate(
    questions=["पीएम किसान योजनेत किती रुपये मिळतात?"],
    contexts=[[
        "पीएम किसान सन्मान निधी योजनेंतर्गत शेतकऱ्यांना",
        "दरवर्षी 6000 रुपये तीन हप्त्यांमध्ये मिळतात."
    ]],
    answers=["पीएम किसान योजनेत 6000 रुपये मिळतात."],
    language="marathi"
)

Individual metrics

from bharatrag.metrics.context_relevance import ContextRelevance
from bharatrag.metrics.groundedness import Groundedness
from bharatrag.metrics.answer_relevance import AnswerRelevance

cr = ContextRelevance(language="hindi")
score = cr.score(
    question="भारत की राजधानी क्या है?",
    contexts=["भारत की राजधानी नई दिल्ली है।"]
)
print(score)  # 0.61

---

Supported Languages

Language	Model Used
Hindi	sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2
Marathi	l3cube-pune/marathi-sentence-bert-nli
English	sentence-transformers/all-MiniLM-L6-v2

More languages coming soon — Tamil, Bengali, Gujarati.

---

Benchmark Dataset

BharatRAG ships with a hand-curated benchmark dataset of 20 Hindi + Marathi QA pairs across:

• Government schemes (PM Kisan, Ayushman Bharat, Jan Dhan, Ujjwala)
• Health (diabetes, sanitation)
• Agriculture (wheat sowing, crop insurance)
• Education (Mid Day Meal, Beti Bachao)

Each example includes a correct answer and a hallucinated answer for evaluation testing.

Dataset location: data/benchmark.json

---

Project Structure

bharatrag/ ├── bharatrag/ │ ├── init.py # evaluate() function │ ├── embeddings/ │ │ └── indic_embeddings.py # Indic embedding models │ └── metrics/ │ ├── context_relevance.py # Metric 1 │ ├── groundedness.py # Metric 2 │ └── answer_relevance.py # Metric 3 ├── data/ │ └── benchmark.json # 20 Hindi+Marathi QA pairs ├── tests/ │ └── test_metrics.py # 21 pytest tests └── examples/ └── run_benchmark.py # Benchmark runner

---

Running Tests

pip install pytest
pytest tests/ -v
# 21 passed ✅

---

Why BharatRAG?

Feature	RAGAS	BharatRAG
English RAG evaluation	✅	✅
Hindi RAG evaluation	❌ Unreliable	✅
Marathi RAG evaluation	❌ Not supported	✅
Indic benchmark dataset	❌	✅
Free, no API key needed	✅	✅

---

Roadmap

• Hindi support
• Marathi support
• 20-example benchmark dataset
• Tamil support
• Bengali support
• 100-example benchmark dataset
• LangChain integration
• LlamaIndex integration
• HuggingFace Spaces demo

---

Author

Pradnya Gundu B.E. Artificial Intelligence & Data Science, APCOER Pune

• GitHub: @pradnyagundu
• LinkedIn: pradnya-gundu

---

License

MIT License — free to use, modify, and distribute.

---

Contributing

Contributions welcome! Especially:

• New Indian language support
• More benchmark QA pairs
• Integration with LangChain/LlamaIndex

Open an issue or submit a PR on GitHub.