Task
A bounded unit of work performed by a person, service, or controlled capability.
Academy Platform / academic lifeSystem / 01
Product
One platform for teaching, research, and collaboration.
ScienzaOS connects academic work through shared projects, sources, documents, review, approval, rights, and reuse. Research Agent for papers and funding proposals is the first offer on this foundation.
Research first
Sources stay visible
People retain approval
Designed for reuse
Evidence linkedReview accountableAction controlled
- 01Teaching
- 02Research
- 03Business
- 04Shared sources
- 05Human review
- 06Responsible reuse
Architecture overview
A layered system for governed research operations.
Stable objects connect people, roles, projects, sources, documents, versions, review, approval, rights, and reuse across the platform.
Review evidence governance
01 / Experience layerAcademic projects · source library · review room · workflow surfaces
02 / Orchestration layerWorkflow engine · planning · drafting · review issues · approval state
03 / Intelligence layerRetrieval · grounded assistance · citation checks · consistency review
04 / Data and knowledge layerSources · documents · versions · rights context · reusable artifacts
05 / Runtime layerCapability routing · policy checks · adapters · sandboxed execution
06 / Infrastructure layerDatabase · object storage · outbox · keys · observability · recovery
Workflow primitives
Composable building blocks for every research process.
The same durable primitives connect Research Agent, Teaching Studio, and Business Collaboration through reviewed knowledge and responsible reuse.
Plan
A structured sequence of steps, evidence requirements, gates, and accountable roles.
Tool
A replaceable capability invoked with explicit parameters and operating constraints.
Retrieval
Find and rank relevant evidence while preserving source and access context.
Evaluate
Assess quality, rights, policy, uncertainty, and decision readiness.
Decide
Issue a governed decision bound to exact evidence, criteria, authority, and conditions.
Adaptive runtime
Observe. Adapt. Execute with guardrails.
The runtime can refine plans and route capabilities while policy, decision state, and accountable human authority remain explicit.
Observe
Collect evidence, case state, policy context, and outcome signals.
Adapt
Refine plans, capability choices, and next evidence within the approved boundary.
Execute
Run exact approved workflow steps and capture deterministic receipts.
Decision intelligence surfaces
Provider-neutral routing
Select an approved model, tool, data source, or human workflow according to purpose, quality, cost, latency, policy, and current availability.
01 / Approved capability set02 / Policy and purpose match03 / Recorded routing decision
Federated retrieval
Search institutional, partner, and public sources while preserving access decisions, citation, provenance, and freshness context.
Institutional sourcesPartner collectionsPublic literature
Evidence graph
Traverse relationships among entities, evidence, rights, methods, decisions, actions, and outcomes without replacing transactional records.
Claims linked to sourcesRights and method contextDecision-ready provenance
Institutional controls
Ready for accountable adoption.
Decision Ledger
Preserve the decision, rationale, evidence, authority, conditions, dissent, and outcome plan.
Human Approval Gates
Put qualified people at significant decision points and bind approval to exact payloads.
Release Gates
Require automated checks, quality thresholds, policy results, and operating readiness.
Enterprise Controls
Support identity, privacy, audit, observability, isolation, exports, and AI-off continuity.
Start with a real research workflow.
Evaluate one paper or funding-proposal workflow with a responsible reviewer, then expand only when the work proves useful.
Request access