feat(liveness): honest SSE states — thinking must be backed by a live stream (L1, #638) - #640
Merged
Conversation
…stream is live (L1, #638) The studio showed 'thinking' while the harness was hung: the SSE client swallowed every error, health was probed only at boot, and the pings the server already sends were discarded (sse_client's comment-only block handler returned without recording anything). The fix is an evidence-backed state machine: frames are the liveness signal (events AND comment-only pings — the discarded-ping hole closed); staleness (30s default, configurable) transitions LIVE→STALE; the STALE branch probes the server to split dead from half-open — a live server with a quiet stream tears down and reconnects (never wait for TCP to notice), a dead one says so. Every transition fires onStateChange and drives the status bar: 'Amicode' / 'stream stalled' / 'server unreachable' — never 'thinking' without a live stream behind it. The tracker is pure by injection (probe/clock/sinks) — sse_client is the production caller; T4a's serve daemon tests the same semantics. No protocol changes; works against the incumbent harness today.
aarontrowbridge
marked this pull request as ready for review
August 29, 2026 13:40
|
Important
This repository does not receive automatic reviews because it has fewer than 10 stars. ⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Plus Run ID: Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
…mutation isn't visible to control-flow analysis (L1, #638) CI's typecheck caught what local vitest couldn't: the stale-branch guard compared a type TS had already narrowed to 'live'. The guard was dead paranoia; the transition always lands. Also the local-gate lesson: run typecheck, not just vitest.
aarontrowbridge
deleted the
638-honest-liveness-the-sse-staleness-state-machine-thinking-must-be-backed-by-a-live-stream-l1
branch
August 29, 2026 13:52
aarontrowbridge
restored the
638-honest-liveness-the-sse-staleness-state-machine-thinking-must-be-backed-by-a-live-stream-l1
branch
August 30, 2026 02:46
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #638.
L1 of the harness-agnostic campaign — the fix for the recurring pain: the studio showing "thinking" while the harness is actually hung (worst across fleet connections).
What
src/sse_liveness.ts— a pure, injected state machine:connecting → live → stale → dead. Frames are the truth (ANY SSE block — events and the comment-only pings the old client discarded). No frames for the staleness threshold (30 s default) → STALE. The STALE branch probes the server: alive-but-quiet = half-open TCP → tear down + reconnect (never wait for TCP to notice); unreachable = DEAD. Transitions fire exactly once, with the evidence logged.sse_client.ts— the tracker wired into the connect/reconnect loop: frames note liveness, connections/disconnections note state, the half-open signal destroys the socket and reconnects, every transition reaches the status bar and anonSseStatehook. The "swallow errors" doctrine is narrowed to "don't crash the host" — the UI gets the truth.status_bar.ts— honest states:Amicode(live, as today) ·stream stalled — probing, reconnecting·server unreachable — work is not lost. Boot states untouched.Evidence
Director-drafted in the harness-agnostic campaign (worktree off main @ the #622 ADR merge).