Skip to content

What is AetheriusDB?

OverviewConcepts

AetheriusDB is a high-performance database that serves SQL, vector search, graph traversal, and AI workloads from one engine — and lets you run your own logic inside the database as WASM compute cells, instead of shipping rows out to an application server.

Most stacks bolt together a relational database, a vector store, a graph database, and a temporal/audit store. AetheriusDB serves all of those over the same data:

  • Relational SQL — strictly typed tables and a broad ANSI SQL surface.
  • Vector search — native fixed-dimension TENSOR(n) columns with HNSW and IVFFLAT indexes for k-nearest-neighbour queries.
  • Graph traversal — relationships are followed by direct link hops, not rebuilt with join tables on every query.
  • Time-travel — query a table as of an earlier point in time, and replay history.

You don’t choose between a SQL database, a vector store, and a graph store — they are different views over the same tables.

Instead of an application server that pulls rows over the network and reshapes them, you can deploy a small WASM compute cell into the database. The cell runs next to the data and returns only the final result — turning many round-trips and serialization passes into a single call.

flowchart LR
  classDef wire fill:#1f2640,stroke:#00d4ff,color:#e8eaf0
  classDef core fill:#1f2640,stroke:#7c5cff,color:#e8eaf0
  classDef store fill:#0d1322,stroke:#4ade80,color:#e8eaf0

  subgraph Wires["Connect over any protocol"]
    W1[Native client]:::wire
    W2[PostgreSQL pgwire]:::wire
    W3[Arrow Flight SQL]:::wire
  end

  subgraph Engine["One engine"]
    Q[SQL planner]:::core
    C[WASM compute cells]:::core
  end

  subgraph Data["One set of tables"]
    S1[Relational + vector + graph + temporal]:::store
  end

  Wires --> Engine --> Data
If you build…AetheriusDB gives you…
An AI / RAG app needing SQL + vector + graph in one queryNative tensor columns, kNN indexes, and link traversal in one engine
A real-time analytics pipeline ingesting millions of rows/secA high-throughput bulk-ingest path that projects raw bytes straight into columns
A multi-tenant SaaS with row-level security on every queryRLS predicates compiled into the query plan
An edge / IoT fleet with intermittent connectivityEach node is a full database; changes reconcile when they reconnect
A team that wants “database-as-code”Branchable schemas, a time-travel debugger, and deterministic test runs
A mixed workload: short writes next to long analytical scansHTAP — writes land in a fast mutable tier with snapshot isolation, and are merged into optimized columnar storage in the background
Opt-in Implemented · off by default — enable with enable_htap = true · primary-key lookups use a fast-reject filter to skip scans

HTAP (mixed transactional + analytical workloads on the same table) is implemented and enabled with a flag. When on, writes land in a mutable Delta tier with snapshot isolation and write-conflict detection, and queries read both the Delta and the sealed analytical tiers. See the configuration matrix for the exact flags and their defaults.