Samples¶
Nine complete applications to read and copy from, in samples/. Each one runs against a disposable Neo4j database and is covered by the repository's tests. New to GraphBridge? Take the workshop first.
Choose by how you author the model¶
All three use the same provider and end in an ordinary EF Core model; they differ only in where the model comes from.
| Sample | You start from | It shows | Keep in mind |
|---|---|---|---|
| 01 Blogging | C# entities and EF configuration | Async CRUD, optimistic concurrency and rollback, native HAS_POST references with Include |
Required references only; no foreign-key reassignment |
| 02 Organization | A Turtle vocabulary plus a YAML persistence policy | Generated C# and EF mappings, separate SHACL validation, parallel employment relationships, a v1 → v2 model migration | No OWL reasoning; generation never deploys schema |
| 03 Movies | An existing graph | Bounded inspection, a reviewed key policy, deterministic scaffolding, preserved native payloads and a backed-up identity backfill | Sampling cannot prove global integrity or infer durable keys |
Tooling and query samples¶
| Sample | It shows |
|---|---|
| 04 Design-time commands | Context discovery and graph migration commands (list, preview, status) around the authoring workflows; see design-time commands |
| 05 Fluent Cypher gallery | Nine reusable graph queries: co-actors, shared-movie scores, optional matches, nested EXISTS, lists and text, aggregates and bounded paths. The same source feeds the integration tests, Doctor, Verify and Query Lab |
| 06 Vector search through VectorData | A search class that knows only IVectorSearchable<T>, bound to a Neo4j vector index by the optional VectorData adapter |
GraphRAG samples (experimental)¶
They use the experimental GraphRAG toolkit through project references, so they run the repository's current source. They run without a model, with stand-ins for it, but always against a Neo4j 2026.09 or later database: here "offline" means no model, not no database. Give each one an empty database of its own: its schema is the database's first migration. Samples 07 to 09 have also run with a real model, GLM-5.3-Flash, and replay its recorded calls offline with GRAPHBRIDGE_AI_MODE=replay. Sample 09's offline extraction replies for its fictional articles are written by hand.
| Sample | It shows |
|---|---|
| 07 GraphRAG quick start | GraphBridge's own docs ingested with exact chunk positions, and answers whose quotes are checked against the stored bytes, or a refusal |
| 08 GraphRAG novel | Alice's Adventures in Wonderland with the novel ontology: characters, kin, places and who says what, read through typed LINQ |
| 09 GraphRAG articles | The article ontology on fictional articles and on NASA's Artemis II news releases: authors, topics, jobs and quotations, the source's own metadata, an Agent Framework agent, and the toolkit's streaming endpoints |
Running them¶
You need the .NET SDK pinned in global.json, PowerShell 7 and Docker for the database steps. Each README lists its exact commands; most start from the repository's disposable test database. Settings come from environment variables, never from files in the repository, and no sample upgrades existing data implicitly.
