Example proposal: portable agent vector memory with Neon and AgentsKit

Neon users building agents often need durable semantic memory without coupling the rest of the application to one framework or model provider. I would like to contribute a small TypeScript example that connects Neon's serverless Postgres driver to the pgvector adapter in @agentskit/memory.

The example would cover schema setup, a deterministic store/search/delete lifecycle, environment-variable documentation, and cleanup. The runnable path would not require a paid model API: fixed test embeddings would verify the database boundary, while the README would show where an application can supply embeddings from any hosted or local provider.

The proposed directory would follow this repository's naming and self-contained example conventions. I will not push the implementation until a maintainer gives the +1 requested by CONTRIBUTING.md.

Would ai/agentskit-vector-memory be a useful location and scope for this example?

Start with CONTRIBUTING.md and the repository’s self-contained example conventions, then review the proposed ai/agentskit-vector-memory location. A complete example should cover schema setup, deterministic store/search/delete behavior, environment-variable documentation, cleanup, and fixed embeddings without a paid model API; the README should explain where application-provided embeddings fit.