UltraRAG

UltraRAG is a self-hosted framework for researchers and developers building AI systems that answer questions using their own documents. Its Python application and Docker deployment provide a browser workspace for knowledge bases, workflow design and evaluation. Generation can use local models through vLLM or Hugging Face. Embeddings can use local Sentence Transformers or a configured OpenAI service; the OpenAI cloud connection sends embedding inputs to that service.

The visual editor keeps canvas and code views synchronized, so developers can inspect a workflow and adjust its prompts and parameters together. A built-in AI assistant helps with pipeline design, prompt writing and parameter tuning. Knowledge base management supports document Q&A, and the framework also supports multimodal workflows.

Pipelines can use conditional branches and loops to repeat retrieval before producing an answer. Retrieval and generation components are separate Model Context Protocol (MCP) servers, which developers can reuse across workflows or extend with their own tools.

For research comparisons, UltraRAG includes evaluation workflows, benchmark datasets, shared metrics and baseline integration. Its case analysis interface exposes intermediate outputs to help identify where a pipeline went wrong. Developers can also turn a workflow into a conversational web demo without building a separate interface.

UltraRAG is open source under Apache 2.0. It offers CPU and GPU base Docker images and a full GPU image. Its Deep Research pipeline combines AgentCPM-Report with multiple retrieval passes to produce research reports.