Skip to main content

SemanticFed: The Data Federation Platform

SemanticFed is a data federation platform that unifies your databases, warehouses, lakes, and SaaS APIs behind one governed semantic layer — so you can write a single query and get a single answer, with access policy, lineage, and audit built in, without copying the data anywhere.

Most analytics stacks answer questions by moving data: a pipeline copies a source into a warehouse, another reshapes it, a third feeds a dashboard, and every copy becomes a thing to schedule, monitor, reconcile, and pay for. SemanticFed takes the opposite path — it queries data where it already lives. A cost-aware planner pushes filters, projections, and aggregations down to each source, joins only what it must in memory, and returns one result. You move bytes, not terabytes, and you read live operational data instead of a stale copy.

Pre-launch

SemanticFed is launching in 2026 and is currently in waitlist mode. This documentation describes the platform as it exists in the codebase today and the direction it is committed to. Capabilities not yet built are clearly marked as roadmap, and there are no customer, revenue, or performance claims.

🎯 Why SemanticFed​

  1. Query in place, don't copy A cost-aware federated query engine pushes work down to each source and returns live results — zero-copy by default, copies only as an opt-in optimization.

  2. One governed semantic layer Model entities, dimensions, and metrics once. "Active revenue" becomes a documented, versioned object that means the same thing in every dashboard, notebook, API, and AI agent.

  3. Governance at the query edge Row filters, column masking, entity-level access, end-to-end lineage, and a full audit trail are defined once and enforced on every federated query — whatever source the data came from.

  4. Trustworthy context for AI Point agents and copilots at the semantic layer so they reason over documented metrics, with the same access policy as a human — grounded answers, not hallucinated schemas.

  5. One SQL and GraphQL contract Apps and tools integrate against a single stable interface over the whole federated model instead of a dozen source drivers.

  6. Workspace-native and enterprise-ready Every connector, data source, model, query, and policy belongs to a workspace. Role-based access control, audit, billing, files, notifications, and internationalization are included out of the box.

🚀 Quick Links​

📚 Core Concepts​

These are the objects SemanticFed works with — the vocabulary of the product:

  • Connector — the driver for a kind of source (PostgreSQL, MySQL, MongoDB, Snowflake, BigQuery, Redshift, Databricks, ClickHouse, REST/GraphQL APIs, S3/Parquet) that advertises which operations it can push down.
  • Data source — a concrete, registered instance of a connector with an endpoint, a credential reference (never an inline secret), a push-down toggle, and a connectivity lifecycle.
  • Connection — a recorded health probe (health, latency, message) used to reconcile a source's status.
  • Semantic model — the typed, versioned model of your business that sits above your sources (DRAFT → PUBLISHED → DEPRECATED → ARCHIVED).
  • Entity, dimension & metric — modeled business objects, typed attributes to slice by, and aggregations defined once and resolved consistently everywhere.
  • Relationship — a typed join between entities so the engine can plan cross-entity and cross-source joins correctly.
  • Saved query & query run — governed query text over the model, and the auditable record of each federated execution (status, plan summary, rows, duration, sources touched, cache status).
  • Access policy — an ALLOW/DENY rule scoped to ROW, COLUMN, or ENTITY, enforced on every federated query.
  • Workspace — the container that owns everything above and provides multi-tenancy, role-based access control, and governance out of the box.

🤝 Support & Community​