Ontology Engine
Define your business as a typed model — object types, link types, and actions — in a single YAML schema. Tapestry materializes it across Neo4j, Postgres, and Qdrant in one coordinated pass.
MVP Stage
Tapestry
The Context Fabric Platform for Canadian Enterprise
Generic LLMs reason over text. Tapestry reasons over your business — modeled as a typed graph of objects, links, and actions, with every AI answer cited to a specific record.
Request a Briefing
Product Preview
The Tapestry Workspace
A live walkthrough of the analyst workspace — Object Explorer, Graph Visualizer, and AI Chat with citations — all running on the same Context Fabric.
Object Explorer · Graph Visualizer · AI Chat · Data Catalog · Pipeline · Ontology Editor · Actions Queue
Capabilities
What Tapestry Does
Data Connections and Lineage
Native connectors for CSV, Excel, PostgreSQL, MySQL, REST APIs, and S3. Every dataset is registered with schema, freshness, and full lineage from raw source to materialized object.
Transform Pipeline
Python functions decorated with @transform convert raw datasets into typed ontology objects. Versioned, schedulable, and re-runnable with full provenance preserved.
AI Query Layer (GraphRAG)
A LangGraph agent layer answers natural-language questions by composing graph traversals and vector lookups. Every answer cites the exact object IDs it was built from — no hallucination.
Action Execution Engine
Ontology actions mutate the Context Fabric under a configurable approval workflow. The AI recommends, humans decide, and every effect lands in an append-only audit log.
Multi-Tenancy and Governance
Schema-per-tenant on Postgres, DB isolation on Neo4j, collection prefix on Qdrant. JWT RS256 auth, RBAC roles, parameterized queries — compliance is structural, not retrofitted.
Domain Templates
Ready-to-Run Ontologies
Energy & Mining
Assets, maintenance records, technicians, sites — pre-loaded with Canadian energy data and a working schedule_maintenance approval flow.
Financial Services
Customers, loans, collateral, risk assessments, covenants — with breached-covenant detection and an approve_loan workflow.
Retail & CPG
Products, suppliers, stores, inventory, purchase orders, promotions — with critical-stock alerts and a create_purchase_order action.
The Problem
Generic AI Cannot Reason on Enterprise Data
Enterprise data lives in dozens of disconnected systems. The same business entity — a vendor in ERP, a supplier in procurement, an account in CRM — appears under different names, different keys, different schemas. A generic LLM querying any one of them has no way to know they're the same, no way to traverse to the others, and no way to cite specific records.
Tapestry inverts the architecture. Entity resolution happens at ingestion, not query time. Relationships are explicit, not inferred. Every object has an ID, every link has a type, and every AI answer references the specific objects it was built from. Decisions become defensible. Outputs become auditable.
Who This Is For
Built for Canadian Mid-Market Enterprise
Tapestry adapts to any domain. It ships with working templates for three.
Energy and Resource Operators
Connect asset registries, maintenance, and field data into a unified graph. Surface overdue maintenance and route work-order approvals — grounded in real records.
Financial Services
Map customers, loans, collateral, and covenants into a typed graph. Detect breached covenants automatically and route loan decisions through structured approval queues.
Retail and Consumer Goods
Connect products, suppliers, stores, and inventory into one operational picture. Identify critical stock-outs and supplier traceability gaps without stitching exports together.
Request a Briefing
Tapestry is in MVP with three working domain templates. We're talking to design partners in energy, finance, and retail — bring us the problem.
Get in Touch