Forced regeneration bypasses eligible cached bodies and requests fresh implementations. Compare it with ordinary offline reuse:

# Reuse accepted implementations; fail on a cache miss.
$ llmc++ -fllm-offline main.cpp -o main
 
# Request fresh bodies, even though offline is also set.
$ llmc++ \
    -fllm-offline -fllm-regenerate \
    main.cpp -o main

Regeneration takes precedence over offline mode, so the second command can contact a backend. Accepted results are saved unless cache writes are disabled. A function-level regenerate modifier scopes this behavior to one target.

A fresh body may differ even with the same prompt. Inspect the replacement and rerun its behavioral checks. Regeneration is a deliberate bypass of reuse, distinct from a miss caused by changed inputs.