Release date: 2025.11.25

Highlights

This release introduces a refreshed UltraRAG front-end UI, fixes several logical issues, and adds new ToolCall and PipelineCall capabilities for directly invoking UltraRAG tools or pipelines from your own code. The retriever server has been further optimized to support full deployment without repeated corpus/index initialization, significantly improving experimental efficiency. We also refine the GPU/CPU configuration logic for more stable and flexible retriever performance.

What's Changed

1. Added full deployment support for the retriever server. by @mssssss123 #136
2. Updated the UltraRAG demo front-end interface. by @mssssss123 #139
3. Optimized GPU/CPU configuration logic for the retriever server. by @mssssss123 #142
4. Introduced new ToolCall and PipelineCall functionality. by @hm1229 @mssssss123 #143

Release date: 2025.11.13

Highlights

This release improves the stability and flexibility of UltraRAG. Retriever and index are now fully decoupled, with added support for Milvus and Faiss. The VisRAG 2.0 pipeline is newly supported, and compatibility with chonkie 1.4.0+ has been updated. Several pipelines, including Search-o1 and WebNote, have been fixed. User experience is improved through progress-bar support during document parsing, a script for saving retrieval results, and zoom-in support for images in the case-study UI. The README now includes a clearer Docker usage guide.

What's Changed

1. Decoupled Retriever and Index with Milvus & Faiss Support: The retriever and vector index are now fully decoupled, with native support for both Milvus and Faiss, providing greater flexibility for large-scale corpus construction and high-performance retrieval. by @mssssss123 #122
2. Added Full Pipeline Support for VisRAG 2.0: UltraRAG now natively supports the VisRAG 2.0 Pipeline, enabling stronger multimodal document understanding and integrated image–text reasoning. by @mssssss123 #115
3. Upgraded Compatibility for chonkie ≥ 1.4.0: Updated document-processing modules to be fully compatible with chonkie 1.4.0+, enhancing stability in structured extraction and parsing workflows. by @yinyannlp @mssssss123 #119 #120
4. Fixed the Search-o1 Pipeline: Resolved critical issues in the iterative search-generation loop, restoring stable functionality for the Search-o1 pipeline. by @mssssss123 #123
5. Fixed Accuracy Metric Calculation Bug: Corrected inaccuracies in the acc metric computation, ensuring more reliable and consistent evaluation results across benchmarks. by @mssssss123 #127
6. Added Script for Saving Retrieval Results: Introduced a new script to persist retrieved content, improving debugging workflows, visualization, and error case analysis. by @mssssss123 #129
7. Fixed Bugs in the WebNote Pipeline: Addressed multiple issues in the WebNote pipeline to improve robustness when processing webpage-like or long-form content. by @mssssss123 #128
8. Enhanced Case Study UI with Image Zoom-In Support: The case-study interface now supports zoomable images, offering a significantly improved multimodal inspection and visualization experience. by @mssssss123 #130
9. Added Progress Bar for Document Parsing: Document parsing now displays a progress bar, providing clearer feedback during long document processing. by @mssssss123 #133
10. Updated README with Complete Docker Usage Guide: The README has been updated with a comprehensive Docker usage section, making deployment easier and more accessible. by @xhd0728 #131

Release date: 2025.10.22

Highlights

1. Comprehensive Multimodal Upgrade: Both the Retriever and Generation Servers now support multimodal inputs, enabling a complete end-to-end multimodal workflow from retrieval to generation.
2. Corpus Parsing and Chunking Redesign: The Corpus Server adds multi-format file parsing with deep MinerU integration, supporting token-level, sentence-level, and customizable chunking strategies to flexibly adapt to diverse corpus structures.
3. Unified Deployment and Efficient Inference: The Retriever and Generation Servers are fully compatible with standardized deployment frameworks such as vLLM, supporting offline inference, multi-engine adaptation, and accelerated experimentation.
4. Enhanced Evaluation and Experimentation Workflow: Introduced TREC-based retrieval evaluation and significance testing modules, supporting parallel experiment execution and multimodal result visualization to optimize research assessment and experimental workflows.

What's Changed

• Corpus Server supports plain text extraction from .txt, .md, .pdf, .xps, .oxps, .epub, .mobi, and .fb2 files. @mssssss123
• Corpus Server adds simple per-page image conversion for .pdf files. @mssssss123
• Corpus Server integrates MinerU for high-precision PDF parsing. @mssssss123
• Corpus Server introduces a new chunking strategy supporting token-level (word/character) segmentation. @mssssss123
• Corpus Server supports sentence-level chunking. @mssssss123
• Corpus Server supports customizable chunking rules (default rule recognizes Markdown sections; other rules can be extended via config files). @mssssss123
• Retriever Server supports three retrieval engines: Infinity, Sentence-Transformers, and OpenAI. @mssssss123
• Retriever Server supports multimodal retrieval. @mssssss123
• Retriever Server adds BM25 sparse retrieval. @xhd0728
• Retriever Server supports hybrid retrieval (dense + sparse). @mssssss123
• Retriever Server provides standardized deployment based on vLLM, unified under the OpenAI-compatible API. @xhd0728
• Retriever Server supports online retrieval via Exa, Tavily, and ZhipuAI. @xhd0728
• Reranker Server supports Infinity, Sentence-Transformers, and OpenAI ranking engines. @xhd0728
• Generation Server supports multimodal inference. @mssssss123
• Generation Server introduces vLLM offline inference, significantly improving experimental efficiency. @mssssss123
• Generation Server supports Hugging Face inference for local debugging. @xhd0728
• Evaluation Server supports TREC retrieval evaluation. @xhd0728
• Evaluation Server supports TREC significance testing. @xhd0728
• VisRAG Pipeline enables an end-to-end workflow from local PDF ingestion to multimodal retrieval and generation. @mssssss123
• RAG Client supports running multiple experiments in parallel under the same pipeline through custom parameter files. @mssssss123
• UltraRAG Benchmark adds six new VQA datasets, including wiki2024 and corresponding VQA corpora. @mssssss123 @xhd0728 @hm1229
• Case Study UI adds multimodal result visualization support. @mssssss123

Release date: 2025.08.28

Highlights

UltraRAG 2.0 introduces a complete redesign of the system architecture, fully adopting the MCP-based modular structure. This release significantly improves flexibility, extensibility, and developer experience, enabling researchers and engineers to rapidly build, customize, and reproduce complex RAG pipelines with minimal overhead.

The core of UltraRAG 2.0 is the new MCP Server–Client workflow, providing a clean separation between functional components such as Corpus, Retriever, Generation, Evaluation, Router, and Note. The pipeline engine has been rebuilt to support looping, branching, multi-server instances, and fully customizable data flow, enabling Search-o1, RankCoT, WebNote, Iter-RetGen, and other advanced RAG pipelines to run out-of-the-box.

Developer experience is further enhanced with streamlined YAML configuration, Functool-style tool invocation, automatic server building, and improved error handling. UltraRAG 2.0 marks a major step toward a unified, reproducible, and developer-friendly RAG research platform.

What’s Changed

MCP Architecture & Server–Client Redesign

• Unified MCP Server Architecture: All RAG components (Benchmark, Corpus, Retriever, Generation, Evaluation, Router, Note) are now implemented as standalone MCP Servers, enabling modular development and clean extensibility.
  by @hm1229 @xhd0728
• Full MCP Client Pipeline Engine: Introduced a new pipeline executor supporting loop, branch, router, multi-server aliases, early stopping, and nested pipelines, making complex iterative workflows natively supported.
  by @hm1229 @xhd0728

Corpus Server

• Added unified file-reading support (doc/docx/pdf/md/txt) and multiple chunking strategies (static/semantic/LLM).
  by @mssssss123
• Enhanced Chunk Pipeline: Improved stability and extensibility for large-document processing and JSONL corpus generation.
  by @mssssss123

Retriever & Reranker Server

• Retriever Server Rebuilt: Embedding, indexing, retrieval, and reranking fully modularized. by @mssssss123 @Kaguya-19 @gdw439
• Improved Vector Index Handling: Cleaner index-building API and improved fault tolerance during large-scale indexing.
  by @mssssss123 @Kaguya-19 @gdw439
• Reranker Server Added: Native support for passage reranking in modular form. by @mssssss123 @xhd0728

Prompt & Generation Server

• Unified Generation Server: Supports OpenAI-compatible API, local HF/vLLM models, customizable sampling params, and Prompt Server integration. by @methanechen @hm1229
• Prompt Server Support: Introduced dedicated Prompt Server for parameterized Jinja templates and reusable prompt logic.
  by @hm1229 @xhd0728

Benchmark & Evaluation Server

• Unified Benchmark Loader: Added support for multiple datasets simultaneously with standardized q/gt extraction.
  by @lifelsl @xhd0728
• Expanded Metrics: Added CoverEM, Acc, and improved EM/F1 calculation consistency.
  by @xhd0728 @lifelsl

Baseline Pipeline Implementations

• baselines added and tested on UltraRAG 2.0 MCP.
  by @xhd0728 @Mignonmy @lifelsl
• Multi-Benchmark Leaderboard Support: Standardized evaluation outputs to enable stable leaderboard reproduction.
  by @xhd0728 @mssssss123

Developer Experience Improvements

• ultrarag build Command: Automatically compiles servers, generates server.yaml & parameter.yaml, and validates parameters.
  by @hm1229
• Improved Logging & Debugging: Added log-level control, error propagation, and clearer exception tracing.
  by @hm1229
• Multi-Agent/Multi-Retriever Support: Added server aliasing for running multiple retrievers or generators within one pipeline.
  by @xhd0728

Documents

• Project Website: The official UltraRAG 2.0 website. by @mssssss123 @xhd0728
• Comprehensive Tutorial: A full end-to-end guide covering installation, MCP Server definitions, pipeline configuration, loop & branch logic, example baselines, and best practices for developing new servers and tools. by @mssssss123 @xhd0728
• README: The README has been redesigned to include a clearer introduction to UltraRAG 2.0. by @mssssss123 @xhd0728

Contributors

Thanks to everyone who contributed to the UltraRAG 2.0 beta and release:

• Project Leaders: @yanyk13 @EdwardZH @mssssss123
• Technical Leaders: @hm1229
• Core Contributors: @xhd0728 @hm1229 @mssssss123
• Contributors: @lifelsl @methanechen @Kaguya-19 @gdw439 @Mignonmy