OpenRAG 2.0 is a document-centric Retrieval-Augmented Generation (RAG) platform developed by LINAGORA. According to the provided Discovery records, it was released on July 8, 2026, and offers a React console for managing documents, partitions, processing jobs, and models, along with an OpenAI-compatible API. Internally, it adopts a Hexagonal architecture that separates layers into core, services, api, and di. To use a library analogy: documents correspond to books, partitions to subject-specific shelves, and indexing/search presets to the librarian’s classification rules. This allows researchers to swap out specific components without rebuilding the entire model or storage layer. Typical RAG implementations require assembling document collectors, parsers, embedding models, vector stores, retrievers, generation models, and management UIs from disparate projects. During this process, differing API formats and data models make configuration changes and fault tracing difficult. Additionally, relying on managed services may expose documents and queries to external systems. OpenRAG 2.0 differentiates itself through a sovereign-by-design structure that enables running the entire stack on self-hosted infrastructure, along with partition-specific indexing and search pipeline presets. Just as GPT handles various types of text via a common interface, OpenRAG adopts an approach that organizes distinct document sets and search configurations into partitions as operational units. However, supported document formats, built-in embedding models, vector store types, Reranking capabilities, and authentication methods must be verified against the official documentation. Life science researchers can utilize the platform by loading papers, experimental protocols, and equipment manuals into separate partitions and isolating search policies. For example, one could index 50 new daily papers into a literature partition, request the top 10 source documents per query via the OpenAI-compatible API, and configure the generation model to summarize abstracts and source locations for five candidate papers. Note that this is a hypothetical operational scenario configurable by the researcher rather than an official performance metric; actual request parameters and response schemas require verification via the API reference. Another use case involves isolating Standard Operating Procedures (SOPs), safety guidelines, and quality documents into department-specific partitions to control query scope. By dividing 1,000 documents across research, quality, and safety partitions and designing queries to search only designated partitions, the inclusion of irrelevant documents in results can be minimized. While the OpenAI-compatible API aims for an interface that facilitates easy integration with existing clients, the full scope of compatibility and model replacement procedures remain unverified. Therefore, before processing sensitive research data, access control, transmission and storage encryption, audit logs, and deletion policies must be confirmed in the official deployment documentation.

OpenRAG 2.0 is a document-centric Retrieval-Augmented Generation (RAG) platform developed by LINAGORA. According to the provided Discovery record, it was released on July 8, 2026, and offers a React console for managing documents, partitions, processing jobs, and models, along with an OpenAI-compatible API. Its internal architecture adopts a hexagonal architecture that separates the core, services, api, and di layers. To use a library analogy: documents are books, and partitions are subject categories.

🔬 New Life Sciences Paper Evidence Search: Load 50 papers daily into the literature partition, query via an OpenAI-compatible API, and retrieve the top 10 results. The generative model then compiles five candidate papers with supporting evidence. These figures represent virtual operational targets; actual throughput must be measured in the deployment environment.