Samples
Start small, then add semantic complexity. The samples are teaching applications; benchmark fixtures remain separate experimental artifacts.
The learning path
| Sample | Purpose | Start with |
|---|---|---|
| Foundgine.SupplyChain (.NET) | Minimal realistic application: application boundary, semantics, authorization, MCP and PostgreSQL. | Starter tutorial |
| Foundgine.SupplyChain.Advanced (.NET) | Full semantic proving ground: grounding, retrieval, richer authorization, high-assurance mutation and agent-facing execution. | Tutorial index |
| foundgine-supply-chain (Java) | Domain, semantic annotations and authorization shape — the smallest path to a compiling Foundgine Java application. | Starter tutorial |
| foundgine-supply-chain-advanced (Java) | Claims/authorization, high-assurance mutation, MCP transport and PostgreSQL, and adversarial security scenarios. | Tutorial |
| Supply Chain PenTest (.NET) | Deterministic security regression across MCP and GraphQL. | PenTest page |
Why separate samples?
The starter teaches the basic application shape without requiring the reader to absorb
every semantic feature at once. The advanced sample then adds ambiguity, retrieval,
authorization depth and adversarial behavior. Benchmark-only fixtures stay under
benchmarks/ so experimental evidence does not become confused with
reference architecture.
Go from sample to source
Starter source (.NET)Smallest path to a working application.Advanced source (.NET)Full semantic and security proving ground.Starter source (Java)Domain, semantics and authorization shape in Maven.Advanced source (Java)MCP, PostgreSQL and high-assurance mutation in Java.PenTestAttack the same boundary from MCP and GraphQL.EvidenceSee the application-level E2E workload.