Benchmarks
Evaluating Ragie against real-world applications.
Recall measures how effectively a RAG system retrieves all relevant information from a knowledge base without missing any critical insights.
Precision ensures a RAG system retrieves only the relevant information, filtering out noise.
Accuracy measures how effectively a RAG system provides correct, contextually relevant answers based on retrieved information.
How Ragie achieves this
Advanced Extraction
Ragie uses a multi-step extraction process to capture data from any format. Texts are efficiently extracted during ingestion, tables and figures are processed with advanced OCR, while LLM vision models generate contextual descriptions for images and charts to enrich the extracted data.
Specialized Table Chunking
Ragie's specialized table chunking prepend table headers to each chunk and ensures that row data is never split mid-record. This feature helps Ragie retrieve precise data better from dense tables, especially in the financial sector where critical data often live in tables.
Advanced Retrieval Modes
Ragie has advanced retrieval modes like recency bias which prioritizes fresh data, hi_res mode for extracting images and tables, and fast mode which is optimized for speed. Fast mode extracts only text and can be up to 20x faster than hi_res mode.