The source contract for data orchestration.
Use the Data Connector Protocol to define how external systems are modeled, executed, fingerprinted, cached, validated, and admitted into Vuon data workflows.
Declare the plugin and sources
Define plugin identity, source keys, input schemas, auth, permissions, schema strategy, and cache policy.
Normalize, fingerprint, and run
The connector canonicalizes input, renders provenance-safe identity, resolves runtime secrets, and fetches source data.
Materialize validated rows
Vuon validates the result, composes cache keys, records provenance, and makes the output available to downstream workflow nodes.
Connector model
Manifest, connector, fingerprint, cache
A contract before data enters the workflow
DCP is the boundary between source-specific code and platform orchestration. Plugin code knows how to reach a system. Vuon handles retries, cache key composition, result validation, provenance, permissions, sandboxing, and UI behavior.
Plugin manifest
Declare identity, version, runner strategy, trust model, auth model, permissions, and the sources a plugin provides.
Source connector
Validate input, normalize it into a deterministic shape, render a non-secret source description, fingerprint freshness, and return rows plus schema.
Runtime boundary
Resolve secrets only at execution time and enforce network, filesystem, environment, subprocess, IPC, and secret permissions.
Stable identity for changing sources
External APIs and workspace tools are not naturally shaped like warehouse tables. DCP gives each source a repeatable identity and freshness signal before its output is materialized for downstream DAG work.
Schema strategy
Use declared, inferred, or dynamic schemas so each source can describe the rows it returns, including multi-source connectors such as different Notion databases.
Fingerprint strength
Expose strong, weak, TTL, or none freshness guarantees using source-specific signals such as hashes, snapshot ids, ETags, timestamps, or expiry windows.
Capability metadata
Tell Vuon whether a source is SQL, API, streaming, parameterized, cancellable, schema-discoverable, or dry-run capable so platform behavior can adapt.
How it works
A clear path from context to action.
The same operating rhythm runs through the product: gather trusted context, show the work, then ship an output that can be reviewed.
Declare the plugin and sources
Define plugin identity, source keys, input schemas, auth, permissions, schema strategy, and cache policy.
Normalize, fingerprint, and run
The connector canonicalizes input, renders provenance-safe identity, resolves runtime secrets, and fetches source data.
Materialize validated rows
Vuon validates the result, composes cache keys, records provenance, and makes the output available to downstream workflow nodes.
Next step