NEKOVA docs
NEKOVA / Docs / AI-Native Features

retry & fallback

Structured retry logic for flaky operations — an AI call, a network request, anything that can transiently fail.

Basic usage

retry 3:
    let x = think "hello"
    show x
fallback:
    show "could not get a response"

retry <n> attempts the block up to n times. If every attempt raises, the fallback block runs instead of letting the error propagate.

Why this exists

AI calls (and network calls generally) fail transiently more often than most code — a timeout, a rate limit, a dropped connection. retry/fallback makes handling that a language-level pattern instead of hand-rolled loop-and-try/catch boilerplate every time.

Combined with raise

task fetch_data():
    raise "simulated failure"

retry 3:
    fetch_data()
fallback:
    show "all 3 attempts failed"