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

kubeflow / mcp-server / 42975773254
95%

Build:
DEFAULT BRANCH: main
Ran 11 Nov 2026 06:56AM UTC
Jobs 3
Files 52
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

11 Nov 2026 06:55AM UTC coverage: 89.38% (+1.5%) from 87.843%
42975773254

push

github

web-flow
feat(pipelines): add pipelines client module (#468)

* feat(pipelines): add pipelines client module

Adds the tools listed in #182 on top of kubeflow.pipelines.PipelinesClient
(SDK 0.5.0), so agents can find registered pipelines, start runs and
follow them.

- read: list_pipelines, get_pipeline, list_pipeline_runs,
  get_pipeline_run, get_pipeline_logs
- write, preview first: create_pipeline_run, stop_pipeline_run,
  retry_pipeline_run

Runs start only from a registered pipeline, passed to the SDK as a
Pipeline object: given a string ending in .yaml or .zip, the SDK reads
that file from the server's disk. The namespace is resolved, checked
against policy and then passed to the client, so the checked namespace
is the one the SDK uses. KFP runs take no labels, so stop and retry
cannot check MCP ownership and sit at ml-engineer.

The pipelines extra is not added to pyproject: kfp 2.17 needs
kubernetes<31 and this repo pins kubernetes>=35. kfp is imported lazily
and the tests mock the SDK; sdk_contracts_test runs where kfp is
installed.

Fixes #182

Signed-off-by: pratik naik <naikp0241@gmail.com>

* fix(pipelines): address review feedback on #468

- add the missing space in the --clients help text
- reject tail_lines=True, since bool is an int in Python
- log at debug level when a pod has no main container
- match ApiException subclasses when mapping HTTP errors

Signed-off-by: pratik naik <naikp0241@gmail.com>

* test(pipelines): cover the audit trail and run timestamps

Review on #468 asked for audit trails and finished_at. Both already exist
(every tool call goes through _audit_wrap, and runs carry finished_at),
but nothing tested them for the pipelines tools. Add tests that tool calls
are audited with secret params masked, and that run times come back as ISO
strings. Also spell out pipeline (blueprint) vs run (execution) in the
tool descriptions and the workflow guide.

Signed-off-by: pratik naik <naikp0241@gmail.com>

* fix(pipelines): fix kfp_clien... (continued)

4082 of 4567 relevant lines covered (89.38%)

2.68 hits per line

Coverage Regressions

Lines Coverage ∆ File
39
90.83
3.68% trainer/api/training.py
35
78.8
1.22% core/dynamic_tools.py
32
80.4
-1.48% common/utils.py
9
80.3
4.38% trainer/api/platform.py
7
92.57
0.0% core/server.py
2
99.17
-0.24% trainer/api/monitoring.py
Jobs
ID Job ID Ran Files Coverage
1 python-3.10 - 42975773254.1 11 Nov 2026 06:56AM UTC 52
89.38
GitHub Action Run
2 python-3.11 - 42975773254.2 11 Nov 2026 06:56AM UTC 52
89.38
GitHub Action Run
3 python-3.12 - 42975773254.3 11 Nov 2026 06:56AM UTC 52
89.38
GitHub Action Run
Source Files on build 42975773254
  • Tree
  • List 52
  • Changed 9
  • Source Changed 0
  • Coverage Changed 9
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #42975773254
  • b5b54f8c on github
  • Prev Build on main (#42797333173)
  • Next Build on main (#42987238269)
  • 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