RIDE Lab builds agent-native systems on vLLM-HUST and stewards the core repositories of SAGE.

SAGE — Streaming-Augmented Generative Execution — is the shared flagship product and technical vision of the IntelliStream research ecosystem: applying streaming-computing principles to LLM inference and agent execution.

Our research spans agent programming and orchestration, tool use, RAG, memory, evaluation, application-level state, real workloads, and reproducible end-to-end benchmarks. RIDE is the lab identity, not an additional runtime layer: products and research projects call vLLM-HUST directly or through SAGE.

SAGE product website: sage.org.ai

The SAGE core series lives under the RIDE-Lab GitHub namespace because RIDE Lab is its principal engineering steward. That namespace is an operational home for code, reviews, releases, and product documentation; it does not make SAGE a RIDE-exclusive product. DataSys and vLLM-HUST keep their own technical ownership while contributing the data-system and inference capabilities that SAGE integrates.

Sage Mate and other applications
              |
              v
        SAGE and research systems
              |
              v
          vLLM-HUST

RIDE repositories own caller-side agent systems, policies, workloads, and evaluations. Model execution, KV-cache implementation, decode scheduling, kernels, compilation, and hardware backends remain in vLLM-HUST.

• IntelliStream incubates early research ideas across the ecosystem.
• RIDE Lab conducts agent-native systems research and turns mature results into open-source products.
• SAGE is the ecosystem's shared flagship product for streaming-augmented LLM and agent execution; RIDE Lab stewards its core repositories and agent-facing surfaces.
• DataSys provides framework-neutral stream, graph, vector, index, update, query, and benchmark infrastructure.
• vLLM-HUST develops model execution, KV/cache scheduling, compilation, kernels, and hardware acceleration.

• SAGE - the shared product core, stewarded by RIDE Lab.
• SAGE Docs - product documentation and the source for sage.org.ai.
• SAGE Agentic
• SAGE RAG
• SAGE Eval
• SAGE Studio
• SAGE Tutorials

• SAGE Agent for VS Code
• SAGE Agentic Tool Use
• SAGE Examples
• SAGE Finetune
• SAGE GitHub Manager
• SAGE Intent Library
• SAGE Refiner
• SAGE PyPI
• SAGE Wiki

• Sage Mate - a continuously running agent application built with SAGE and backed by vLLM-HUST.

• SAGE Benchmark
• Agentic Tool-use Benchmark
• Agentic Tool-use SIAS
• RAG Benchmark
• Refiner Benchmark
• Wiki-link Retrieval
• Neuromem Benchmark

• Complete publication timeline - published work, public evidence, corresponding repositories, and explicitly labeled author-reported acceptances.
• SAGE - ICML 2026 dataflow-native framework for modular, controllable, and transparent LLM-augmented reasoning.
• Neuromem - ICML 2026 benchmark and lifecycle decomposition for external memory in LLM systems.
• FlowRAG - WWW 2026 continual retriever adaptation for evolving corpora.
• Demonstrating SAGE - ICPP 2026 demo acceptance reported by the authors; the public program or proceedings record is pending.

Projects graduate from IntelliStream when they have a clear public abstraction, named maintainers, reproducible evaluation, documentation, tests, licensing, and a sustainable release path.

RIDE Lab is not a data-engine or model-runtime umbrella. Framework-neutral data systems belong in DataSys, while model-runtime and hardware-execution projects belong in vLLM-HUST.

See our contribution guide, security policy, and support guide.