Launched this week
RAGReader
Benchmark|Compare. Find Best RAG Pipeline for Your Documents
2 followers
Benchmark|Compare. Find Best RAG Pipeline for Your Documents
2 followers
Open-source RAG research workspace with cited document chat, a configurable multi-LLM deep-analysis matrix, and 13 composable advanced retrieval modules (HyDE, RAG Fusion, RAPTOR, CRAG, Adaptive RAG and more) scored with retrieval and Ragas metrics.. Compare every retrieval method × model variant with auditable ground truth, module traces and Ragas scores — without notebooks or SaaS lock-in.
Launch Team
CloudflareDeploy serverless functions globally in seconds
Promoted
Hey Product Hunt! 👋 I'm from Nevatal, creator of RAGReader.
We built this project because RAG comparison usually stops at Dense, Sparse and Hybrid. Real answer quality also depends on query transformation, routing, evidence refinement and context construction — so RagReader makes those techniques composable and measurable on your own documents.
🚀 Highlights:
• 13 composable RAG modules: HyDE, RAG Fusion, Step Back, MemoRAG, RAPTOR, CRAG, Adaptive RAG and more
• Cited document workspace for PDF, DOCX, TXT, Markdown, web pages and pasted text
• Retrieval Precision@K / Recall@K / F1@K plus Ragas answer metrics
• 100% Free & Open-Source
• Modern Stack: Django ASGI / Channels, React / Vite / Tailwind, Celery + Redis, PostgreSQL, OpenRouter (any model; defaults GPT-4o mini, Claude Haiku 4.5, Gemini 3 Flash), BM25 + cosine similarity, Ollama cross-encoder reranker, Ragas + Instructor, WebSockets / SSE
• Live Demo: https://rag.nevatal.id
• Technical Benchmark & PRD: https://wordpress.nevatal.id/2026/11/10/ragreader-multi-llm-consensus-benchmark-free-alternative-3/
Who it's for: AI engineers and researchers evaluating retrieval methods, models and advanced RAG modules who need empirical metrics before deciding on architecture.
We would love to get your specific technical feedback on module composition, the evaluation metrics, and which pipelines you'd like benchmarked next!