RISE / AI integration / updated October 2026

Your model, your connection.

RISE spends no SyberLabs inference credential. An AI reading request runs on a connection the reader owns: hosted Jev through the reader’s own OpenRouter account, or a pinned Kev-4B running on the reader’s own computer. The RISE server calls no model and holds no model key. Reading and manual settings need neither.

Read the design note ↗

CURRENT STATUS

What the evidence says.

01 / ESTABLISHED
Bounded decisions, one contract
RISE builds finite choice questions from its public catalog, admits an answer only if every choice was offered, and maps it to reading settings in deterministic code. A model can pick an offered value and nothing else. The same contract serves hosted Jev, local Kev and the evaluation harness.
02 / SHIPPED
Shared inference retired
Since RISE #294 the Cloudflare Worker serves the app and a static decision catalog only. The six former inference routes answer 410 with instructions to connect. Connecting OpenRouter uses OAuth PKCE; the key lives in the tab’s memory and reaches only openrouter.ai. Local RISE starts the reader and a pinned Kev-4B on loopback with a per-run key the page never sees.
03 / NOT ADOPTED
Kev on RISE requests
Kev runs only on the reader’s own computer; RISE has no production Kev deployment, by design. In RISE’s internal checks, local Kev did not meet the bar for reading requests; those results are not published (RISE task SCR-004). Jev’s earlier 48 of 49 preference result belongs to the historical evaluation below and is not a Kev result. The local Kev path is tested on Windows with an RTX 5080; the Apple Silicon path is untested.