• Home
  • Features
  • Pricing
  • Docs
  • Announcements
  • Sign In

kubernetes-sigs / inference-perf / 43930557778
87%

Build:
DEFAULT BRANCH: main
Ran 16 Nov 2026 06:48PM UTC
Jobs 1
Files 142
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst

16 Nov 2026 06:36PM UTC coverage: 86.498%. Remained the same
43930557778

push

github

web-flow
test(e2e): cover conversation_replay across three stage boundaries (#1059)

Fixes #929

Old behavior: there was no multi-stage `conversation_replay` e2e
coverage for
the stage-boundary regressions addressed in #701.

New behavior: a sim-backed e2e test runs three concurrent stages of
`conversation_replay`. It checks completion within a 180-second timeout,
per-stage request counts, conversation turn counts against the
blueprint,
and first-turn input token counts against stage 0's median.

Changes:
- Extend `_concurrent_stage_config` to accept per-stage request counts
while
  retaining single-stage support for an integer argument.
- Run three stages with four conversation slots, three turns per
conversation,
two nominal rounds per stage, and two workers: 24 requests per stage, 72
total.
- Key conversations by `(stage, session)` to distinguish sessions
re-primed
  across stage boundaries.
- Keep the three-turn upper bound strict, allowing only each slot's
final
  conversation in a stage to be short. Check the 24-request stage budget
  independently rather than requiring slots to advance in lockstep.
- Check first-turn tokens in every stage, including stage 0's rollover
  conversations, against stage 0's median.
- Add helper self-tests for the token-band and conversation-length
assertions.

Testing:
- `pdm run pytest e2e/tests/test_load_shape_accuracy_sim.py`
- `pdm check`
- Before the review fixes, stage-0 rollover outliers failed with
`Failed: DID NOT RAISE AssertionError`, and the synthetic 7/5/6/6
request split
failed with `every conversation must turn exactly 3 times` for every
choice
of leading slot. Both regressions pass after the fixes; existing
rejection
  tests remain intact.
- The original PR commit passed Unit Tests, Code Coverage, Python
Linting and
Type Checks, and [E2E Test on
change](https://github.com/kubernetes-sigs/inference-perf/actions/runs/43064322644).
CI verification of the review fixes and repeated `e2e-tests` runs are
pending.

S... (continued)

13838 of 15998 relevant lines covered (86.5%)

0.86 hits per line

Jobs
ID Job ID Ran Files Coverage
1 43930557778.1 16 Nov 2026 06:48PM UTC 142
86.5
GitHub Action Run
Source Files on build 43930557778
  • Tree
  • List 142
  • Changed 0
  • Source Changed 0
  • Coverage Changed 0
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #43930557778
  • 332fad88 on github
  • Prev Build on main (#43485826608)
  • Next Build on main (#43934581025)
  • Delete
STATUS · Troubleshooting · Open an Issue · Sales · Support · CAREERS · ENTERPRISE · START FREE TRIAL · SCHEDULE DEMO
ANNOUNCEMENTS · TWITTER · TOS & SLA · Supported CI Services · What's a CI service? · Automated Testing

© 2026 Coveralls, Inc