RAGFlow is an open-source Retrieval-Augmented Generation (RAG) engine that integrates advanced RAG techniques with agentic capabilities to provide a high-fidelity context layer for Large Language Models (LLMs). The system utilizes a converged context engine and an orchestratable ingestion pipeline to extract knowledge from unstructured data, including Word, Excel, PDF, and scanned images, through deep document understanding. Technically, the platform supports multiple recall mechanisms paired with fused re-ranking and offers visualization of text chunking for manual intervention. It is compatible with various LLMs, including OpenAI's GPT-5 series, Gemini 3 Pro, and DeepSeek v4, and supports data synchronization from sources like Confluence, Notion, and Google Drive. Deployment is primarily managed via Docker, requiring a minimum of 4 CPU cores, 16 GB of RAM, and 50 GB of disk space. While Elasticsearch is the default storage engine for text and vectors, users can switch to Infinity. For advanced agentic features, such as the Python/JavaScript code executor, the system requires gVisor for sandboxing. The architecture supports various chat channels, including Discord and Telegram, and allows for customizable embedding models and LLM factories.