Group: GitHub Aindriub

Sort by:Popular▼

✓Popular✓Newest

The privacy model, field metadata, the scrubbing engine, the SPI interfaces every other module implements, and the audit contract (events, sinks, the per-writer hash chain). Carries no business domain.

Independent check that nothing prohibited reaches the model.

Classification metadata applied to application models. Deliberately dependency-free: consuming applications annotate their own DTOs without taking on any of the platform.

Distributed scope state: synthetic identity caching, the re-identification index and the shared read budget. An optimisation and a store, never a source of truth for a synthetic value.

Fans out to sources, correlates, applies privacy, validates, audits.

Deterministic synthetic identity generation. Correct without any cache.

AuthorizationService and purpose validation, built on the InvestigationContext type core already carries. Empty until task 03.

The MCP tool surface. Sees orchestration only, never a connector.

The one controlled path from a pseudonym back to a subject id: authenticated, purpose-bound, audited, with optional four-eyes approval. A library with no transport; never reachable from an MCP tool.

Validated deployment binding and privacy-pipeline wiring.

Data Prism is an open-source privacy layer for Java/Spring teams putting LLM agents or MCP clients in front of internal APIs holding customer data. It pseudonymises personal data per privacy scope, refuses anything unclassified, and can keep a ...

HTTP sources over Spring RestClient. Nothing in the base standalone server or Spring Boot starter compiles against this module: the configuration-driven JSON REST mode it implements is a separately reviewed adapter extension, loaded only ...

Compile-time enforcement that every field of an LLM-exposed model was classified. Belongs on annotationProcessorPaths, never on the compile classpath.

Dependency-only entry point for embedding the protected HTTP MCP endpoint.

Client Library for IRC

Probabilistic entity resolution for Java 8: match a source object against candidate objects on fuzzy field evidence and return a ranked, explainable decision. No framework, JSON, database or logging dependency.

The resolution engine: field comparison, normalization, similarity metrics, alias equivalence, scoring and explainable decisions. Knows nothing about any particular domain.