Hosted addons: Postgres with pgvector + PostGIS, and gobed GPU vector search
Attach managed services to any app or agent in one command — gobed GPU CAGRA vector search, PostgreSQL with HNSW vectors, PostGIS and graph traversal, Mongo, cache, auth, and AI monitoring — with env vars injected into every runtime.
Every app eventually needs a database, a cache, search, auth, and something watching for errors. On app.nz those are now addons: managed services that clip onto any repo or deploy in one command, inject their connection details as environment variables, and bill the same prepaid credit balance as everything else.
The demo above is the whole workflow. Two commands provision a Postgres database and a GPU search index; app addons env shows what got injected; then it's just psql and curl.
The lineup
- **gobed** — our first-party flagship: the GPU-accelerated vector search engine that powers character, art, and repo search on app.nz itself. Int8 static embeddings (a 15 MB model that embeds ~150K texts/sec on CPU), flat/IVF/HNSW indexes, and NVIDIA cuVS CAGRA graph search on GPU. Three HTTP endpoints:
/search,/batch_search,/index. - **Postgres** — managed PostgreSQL with pgvector (HNSW) and PostGIS enabled by default, graph traversal via recursive CTEs, and hybrid retrieval that mixes vector similarity, full-text rank, and geo in one SQL statement.
- Mongo, Memcache, Auth, Monitor Agent — document store, shared cache namespace, Supabase-style auth + realtime, and AI-assisted error intake that dedupes and opens repo tasks. All on /addons.
Env vars your agents can use
Addon variables merge into the environment in a fixed order — account, then project, then addon — and land in every supported runtime: deploys, coding agents, schedulers and queue workers, and notebooks.
DATABASE_URL # postgres, with --attach-default
APP_ADDON_POSTGRES_URL/_HOST/… # always
GOBED_SEARCH_URL / GOBED_API_KEY # gobed
APP_ADDON_MONGO_URL, APP_ADDON_AUTH_*, APP_ADDON_MONITOR_*The part we care most about: agents see the same environment your app does. Ask a coding agent to "add vector search to the docs table" and it can run the migration, build the HNSW index, and verify the query against the real addon — no credential handoff, no pasting connection strings into prompts. Secret-marked values are injected at runtime and scrubbed from build and CI logs.
Declarative too
Addons can live in appnz.yaml next to your runtime config, so a fresh clone provisions its own infrastructure:
name: my-app
runtime: static
addons:
- type: postgres
plan: starter
- type: gobed
plan: sharedWhy gobed first-party matters
Most platforms bolt on a third-party vector database and mark it up. gobed is ours end to end — model quantization (int8, 512-dim, 7.9× smaller), SIMD kernels, IVF/HNSW/CAGRA index code, CUDA memory management — which is why it can be priced as a metered addon instead of a cluster subscription, and why the same engine runs in-process, as a sidecar, or as a dedicated GPU pod. Measured numbers and the honest roadmap (sub-millisecond CAGRA at 100K+ docs is a target, not yet a shipped benchmark) are on the addon page.
Start with curl -fsSL https://app.nz/cli.sh | sh, then app addons create --type postgres --attach-default. The CLI reference has the full command set.