fix(app): keep render state consistent - #7
Open
z2z23n0 wants to merge 3 commits into
Open
Conversation
z2z23n0
marked this pull request as ready for review
July 10, 2026 13:53
Author
|
PTAL when you have a chance, @zimoCCC @yhzx233 @gaoyang07. Thanks! |
z2z23n0
force-pushed
the
fix/render-state-consistency
branch
from
July 23, 2026 08:48
644ab6b to
d0ac1d2
Compare
AlphaSightInc
added a commit
to AlphaSightInc/MOSS-Transcribe-Diarize
that referenced
this pull request
Jul 29, 2026
Target: candidate 56 / Phase P items 60-63 (P1 monotonic measurement, P2 the degrade rule, P3 the real-seam regression, P4 the class sweep), authorized by the seventh prd.md amendment. P1: RunnerBoundedWavInference measures its own decode on time.monotonic() and never reads the runner's elapsed_sec. P2: trustworthy_duration_sec() states the rule once as a conversion - an untrustworthy duration is recorded null (elapsed and RTF) on canonical_processed, logged on moss_transcribe_diarize.live.decode, and never terminal; applied in InferenceTranscript and the coordinator, both of which used to raise. P3: four seam nodes putting the real runner's own duration measurement under the real coordinator for the first time. P4: vllm_runner, model_runner and jobs' progress throttle moved to monotonic; live_transport._request_now ruled deliberate (a timestamp, not a duration). Validation: red-before proved per half by semantic revert - each half has its own named node, and reverting both wall-clock halves reproduces iteration 28's probe record field for field (integrity/canonical_decode_failed, "runner result elapsed_sec must be finite and non-negative", retryable=false) from a test with no host. Green after: full Python suite 608 passed / 2 skipped / 368 subtests in 59 s (604 -> 608); live slice 136 passed / 329 subtests. Six nodes in test_live_vad.py asserted the superseded rule and were rewritten to the new one, not deleted. New candidate 58 recorded, deliberately not fixed: the replay evaluator counts a null (degraded) elapsed as an invalid measurement, so a declared absence fails the RTF summary. Needs its own authorization. Blocker: none. P5 remains - gate, advance expected_main to 77e0014 in merge-keeper.sh, merge OpenMOSS#7, push, redeploy the server (no Mac rebuild needed), then F1 and F3. Until the redeploy the deployed 77e0014 still carries the defect, so no certification run or offline probe speaks for this code.
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.
Problem
Subtitle rendering was not isolated from other job operations:
failed, even when its subtitle data was intact.Changes
renderingbefore the worker starts, reject duplicate render requests with HTTP 409, and restore the previous state if thread startup fails.waiting_reviewso users can edit or retry; other interrupted processing states still recover asfailed.The work is split into three focused commits, one for each failure mode.
Impact
Rendered output can no longer silently fall behind saved subtitle content, duplicate FFmpeg work is avoided, and a server restart no longer makes completed subtitle work inaccessible in the Web UI.
Validation
python -m pytest -q tests/test_app_api.py(6 passed)python -m pytest -q(30 passed)git diff --check origin/main...HEAD