The memory of your lab
Paramus WORLD
A documented observation is not yet a confirmed statement. A plausible interpretation is not yet established knowledge. Paramus keeps them apart — in four graphs, each answering one epistemic question — so your agents know not only what happened, but why a conclusion is justified, and where its limits are.
Four questions, four graphs
| Graph | Epistemic question | Kind of statement |
|---|---|---|
| Workflow graph | What happened? | Observation & provenance |
| Concept graph | What does it mean? | Interpretation & semantic abstraction |
| Knowledge graph | What is known? | Reference knowledge |
| Evidence graph | What supports what — and how sure are we? | Evidence, hypotheses, conclusions |
Every run, tool call and gate decision, recorded as it happens — who acted, with which inputs, producing which result.
- Run provenance and decision traces
- Written by the system, never by hand
- Observation, not interpretation
Derived from your workflows, artifacts and desktop activity: the concepts behind them, connected into a meaningful semantic structure.
- Provenance on every node and edge
- Deletable and re-derivable
- Proposes links you did not draw
The world model: substances, materials, methods and their established relations — the reference your work is grounded against.
- Curated reference knowledge
- Queryable with SPARQL
- Callable by assistants over MCP
Hypotheses and claims with the evidence for and against them: uncertainty, inference steps, confidence and the conditions under which a statement holds.
- Supporting and contradicting evidence
- Confidence and validity on every claim
- Every claim traced to its runs
What we mean by concept graph
“Concept graph” is an established term, but not a standardized graph category — the literature uses it in different ways. In Paramus it means one precise thing: a graph derived from your workflows, artifacts and desktop activity, which identifies the concepts behind them and connects these concepts into a meaningful semantic structure. It is not a hand-drawn concept map, and not a store of contextual metadata: every concept exists because specific runs and files gave rise to it, and the whole graph can be rebuilt from those sources at any time.
How the graphs feed each other
Your work lands in the workflow graph as it happens. The concept graph is derived from it, plus your files and desktop activity, and aligned with the knowledge graph. The evidence graph weighs it all: observations with their uncertainty, claims from the concept graph, reference data from WORLD — and turns them into statements with a confidence and a scope.
Solid: recorded as it happens. Dashed: derived, planned or in design.
One experiment, four graphs
A solubility series for aspirin in ethanol on a liquid handler, approved at a gate by a chemist. Each graph records something different — and only the last one says how far you can trust the conclusion.
Illustrative example. Same experiment: what happened · what it means · what is known · how sure we are.
One vocabulary: the Science Ontology
The four graphs share one vocabulary, the Paramus Science Ontology. It names the things research work produces — workspaces, chats, files, datasets, results, analyses, agents, hypotheses — and how they relate, built on established vocabularies: W3C PROV-O for provenance, schema.org, Research Objects, DCAT and the OBO ontologies IAO, OBI and SIO. The workflow graph is written in it today; its hypotheses already carry a status from draft through tentative to confirmed or refuted, where the evidence graph begins.
The Science Ontology describes the research work, not the chemistry. What a substance is belongs to the knowledge graph and to domain ontologies such as ChEBI. The concept graph links what it derives to those entities: the Science Ontology says what kind of statement something is, WORLD says what it is about.
Four layers, not four databases
The four graphs are epistemic layers, not necessarily four separate stores. The workflow graph is its own append-only record; the evidence graph may live as a qualified layer on top of the concept graph. What matters is that the distinction is explicit: an agent never mistakes a single measurement for a fact, and you can always follow a claim back to the runs it rests on.
Local by design
All graphs live on your Paramus OS installation, in open RDF — no cloud account, no external service, nothing leaves the box unless you decide so.
From Paramus 3.4, Paramus CENTRAL can consolidate these graphs across a team — scoped by each company’s own data-compliance rules.