Native AI runtime

Planned

A native generative AI runtime built into AthenaDB: models load from encrypted, signed packages into protected memory and run inside the database security boundary. Deployment shapes:

• In-process with the database
• Dedicated model machine over an encrypted, mutually authenticated link
• Opt-in external provider (classification-restricted and audited)

Available for Early Access evaluation today

Vector search and embeddings (provider abstraction, indexes, re-embedding) — not the generative runtime. See Retrieval & embeddings.

Product: Native AI runtime