Intelligent systems ecosystem / 01

SAGE

Compose intelligence into dependable services.

Agents, RAG, workflows, evaluation, studio tooling, and application-level state semantics in one open orchestration layer.

Start building Open core repository

24: public projects
05: capability groups
L1–L5: learning path

Agent orchestration RAG + memory Workflow runtime Evaluation Studio + tooling

Platform surfaces

From intent to running workflow.

SAGE owns application semantics and orchestration. Data engines stay in DataSys; model execution stays in vLLM-HUST.

SAGE Core

Execution model, operators, services, and shared lifecycle semantics.

Repository ↗ 02 / AGENT

Agentic

Planning, tool selection, coordination, and composable agent workflows.

Repository ↗ 03 / RETRIEVAL

RAG

Document, chunking, retrieval, refinement, and knowledge workflow primitives.

Repository ↗ 04 / EXPERIENCE

Studio

Visual workflow construction and an inspectable playground for SAGE pipelines.

Repository atlas

Public projects by capability.

The directory separates framework surfaces, reusable libraries, evaluation assets, and developer tools.

SAGEFramework core and execution modelcore↗ SAGE-DocsOfficial product and API documentationdocs↗ sage-agenticPlanning, tools, workflows, coordinationframework↗ sage-ragDocument and retrieval pipelinesframework↗ sage-evalMetrics, profilers, and judgesevaluation↗ sage-studioVisual workflow builder and playgroundproduct↗ sage-tutorialsLayered L1–L5 learning pathlearning↗ sage-agent-vscodeAgent development in VS Codetool↗ sage-agentic-tooluseTool retrieval and rankinglibrary↗ tooluse-benchmarkAgent capability evaluationbenchmark↗ tooluse-siasImportance-aware continual selectionresearch↗ sage-benchmarkAI pipeline evaluation suitebenchmark↗ sage-examplesRunnable framework examplesexamples↗ sage-finetuneLightweight model fine-tuning toolkittoolkit↗ sage-github-managerAI-assisted repository operationstool↗ sage-libs-intentIntent recognition librarylibrary↗ sage-mateAlways-on faculty-twin applicationapp↗ sage-rag-benchmarkRetrieval workflow evaluationbenchmark↗ sage-refiner-benchmarkRefinement component evaluationbenchmark↗ sage-wikiKnowledge and ecosystem wikiknowledge↗ sageRefinerRetrieval result refinementlibrary↗ sagepypiMonorepo package compiler and publishertooling↗ wiki-link-retrievalCurated link-graph expansionresearch↗ neuromem-benchLong-running memory evaluationbenchmark↗

Start with the core

Install. Compose. Observe.

The homepage routes to the maintained learning path; detailed APIs and examples stay in SAGE Docs.

01 / INSTALLPYTHON

pip install isage

Installation guide → L1–L5 tutorials ↗

Shared architecture

One ecosystem. Clear ownership.

SAGE is the application and orchestration layer. It consumes DataSys services and requests model execution from vLLM-HUST.

00IntelliStreamCross-layer research incubator 01SAGEApplications, agents, RAG, workflows 02DataSysStreams, graphs, vectors, indexes 03vLLM-HUSTRuntime, scheduling, kernels, hardware