fix(rivetkit): treat empty actor input bytes as absent

[abcxff]

No description provided.

[abcxff]

Stack for rivet-dev/rivet

• #5548 feat(container-runner): log version and git sha at actor sta …
• #5549 feat(container-runner): setup instance memory and cpu usage …
• #5546 fix(rivetkit): keep actor event loop alive through shutdown …
• #5539 feat(container-runner): report actors as crashed on unexpect …
• #5536 feat(container-runner): log unexpected platform SIGTERM as a …
• #5530 fix(rivetkit-core): cancel driver alarm before sqlite teardo …
• #5531 feat(container-runner): keep instance warm instead of self-e …
• #5533 chore(container-runner): default stop grace to 10s
• #5528 fix(rivetkit): log and surface actor startup failures
• #5529 fix(rivetkit): treat empty actor input bytes as absent 👈
• main

Get stack: forklift get 5529
Push local edits: forklift submit
Merge when ready: forklift merge 5529

_{change ouvoqtns}

[railway-app]

🚅 Deployed to the rivet-pr-5529 environment in rivet-frontend

Service	Status	Web	Updated (UTC)
frontend-cloud	😴 Sleeping (View Logs)	Web	Aug 7, 2026 at 1:13 am
website	😴 Sleeping (View Logs)	Web	Aug 2, 2026 at 7:07 pm
kitchen-sink	😴 Sleeping (View Logs)	Web	Aug 2, 2026 at 7:58 am
frontend-inspector	😴 Sleeping (View Logs)	Web	Aug 2, 2026 at 7:30 am
ladle	✅ Success (View Logs)	Web	Jul 31, 2026 at 4:13 am
mcp-hub	✅ Success (View Logs)	Web	Jul 31, 2026 at 4:13 am

[claude]

Review

Small, well-targeted fix: Input::decode() / decode_or() now treat a zero-length input buffer the same as an absent one, mirroring the existing Snapshot::decode() behavior. The rationale in the comment (engine omits input entirely, but an empty base64 string round-trips to Some(vec![]) rather than None) is clear, and the added tests (input_decode_or_default_treats_empty_bytes_as_missing, input_decode_treats_empty_unit_as_unit, run_actor_invalid_input_fails_to_start) cover the fix and guard the "invalid non-empty bytes still error" case.

Bug: is_present() was not updated to match the new semantics

present_bytes() is now the source of truth for "does this input have real bytes", used by both decode() and decode_or(). But is_present() (start.rs:44-46) still does the old check:

pub fn is_present(&self) -> bool {
    self.bytes.is_some()
}

Since is_present() is pub on a type reachable from actor authors via Start::input, this creates an observable inconsistency: for an empty (but Some) byte buffer, is_present() reports true while decode()/decode_or() now behave as if the input were absent (defaulting, or erroring with MissingInput when required). A caller branching on is_present() to decide whether to trust decode()s result vs. apply its own fallback would get the wrong signal for exactly the case this PR fixes.

Suggested fix:

pub fn is_present(&self) -> bool {
    self.present_bytes().is_some()
}

Minor

• The fix is scoped to rivetkit-rusts Input decode path. Worth double-checking whether the equivalent input path in rivetkit-typescript (or wherever the empty-base64-vs-omitted distinction can arise) has the same latent issue, since rivetkit (Rust) is meant to mirror rivetkit-typescript per the layer-architecture doc. Not blocking if TS already normalizes this elsewhere.

Overall a solid fix with good regression tests; the is_present() inconsistency should be addressed before merge.