API reference
Media privacy
How prompts, references, account credentials and generated media are handled.
Provider processing
Klymi sends generation prompts and reference inputs to the configured upstream provider with a server-only credential. The selected model processes the request. Provider retention and copies follow that provider's policies.
Studio history
Drafts, reference IDs, pending prompt snapshots and likes use account-scoped browser storage. Normalized image blobs are cached in IndexedDB. When sync is available, settings and likes also live in the private account metadata service. Durable job records preserve progress and output keys across sessions. Another account cannot read your ledger or media.
Output links
Generated media is copied to private storage. Studio reads require a verified same-account session. API reads require an active same-account key and return temporary signed output URLs, normally valid for 30 minutes. Anyone with a temporary URL can read that object until expiry. If copying fails, the output may retain a provider URL and a warning. Download important results; no retention guarantee or generated-output deletion API is offered.
Credentials
Only a SHA-256 digest of a Klymi API key is stored; its secret is displayed once. Billing records and key digests are unavailable to browser database roles. Provider and payment credentials stay in the server environment. Revoke a Klymi key from the dashboard if it is exposed.