RAG Evaluator

LLM-as-Judge for Persian RAG chatbots: scores each answer against a reference on accuracy, completeness, relevance and faithfulness (0–10). Runs entirely in your browser.

Your OpenRouter API key

Your key stays in your browser and is sent only to OpenRouter. No server of ours ever sees it. Usage is billed to your OpenRouter account. Get a key at openrouter.ai/keys.

Remember for this tab (sessionStorage, cleared when the tab closes)

Judge model

Any OpenRouter model id works; Persian answers judge best on larger models.

Dataset

Default: the synthetic sample below. Your CSV needs columns question, ground_truth, agent_response (or the Persian headers of the sample). Limits: 20 rows, 2000 characters per cell, 64 KB. The file is read locally, never uploaded.

Dataset

#	Question	Reference answer	Agent response

RAG Evaluator

Results

Dataset