Foundation: native Neo4j 5.x vector index for schema embeddings

[peter7775]

Problem Statement

Planned AI features (semantic schema search, transformation rule suggestions, text-to-Cypher) need to store and query vector embeddings. The application currently has no vector infrastructure at all (no vector/embedding references anywhere in the codebase).

Proposed Solution

• After the Neo4j and driver upgrades land, add a Cypher migration/setup step that creates a VECTOR INDEX on nodes representing schema elements (columns/tables), using Neo4j 5.13+ syntax.
• Extend the domain model (internal/domain/models/) with an embedding vector field for schema nodes.
• Add repository methods wrapping db.index.vector.queryNodes() in internal/infrastructure/persistence/neo4j/repository.go.

Alternative Solutions

Store embeddings outside Neo4j (pgvector, in-memory). Simpler to start with, but prevents combined graph+vector queries in a single Cypher statement.

Use Case

Foundation for: semantic schema search, transformation rule suggestions, text-to-Cypher.

Priority Level

Nice to have

Implementation Ideas

• Vector dimension depends on the chosen embedding model (see the EmbeddingProvider port issue).
• Decide between cosine and euclidean similarity in the Neo4j vector index config.

Additional Context

Depends on: "Upgrade Neo4j server from 4.4 to 5.13+" and "Migrate neo4j-go-driver from v4 to v5". Related to: "EmbeddingProvider port" issue.