Skip to content

Port FM payload loader to kumo.fm - #391

Draft
am-porter wants to merge 13 commits into
codex/pytorch-existing-snapshotsfrom
codex/fm-loader-migration
Draft

am-porter wants to merge 13 commits into
codex/pytorch-existing-snapshotsfrom
codex/fm-loader-migration

Conversation

@am-porter

@am-porter am-porter commented Nov 18, 2026 •

Copy link
Copy Markdown
Contributor

Summary

  • Add experimental kumo.fm payload loading over connector-created GraphSnapshots. The loader adopts a public kumo.engine snapshot/index, derives table and relationship metadata, resolves context and query occurrences, and builds target and graph plans through the public PyTorch API. It does not re-ingest the base graph.
  • Retain the frozen bundle/workload path and HTTP payload contract. A loaded batch keeps its Arrow tables and can be encoded as arrays or Arrow IPC without resampling. Validate occurrence alignment, target visibility, supported temporal plans, and bounded graph fanout.
  • Add a concise connector-to-SDM prediction example, integrated snapshot tests, source/wheel CI, and a frozen DuckDB comparison runner.

This PR is stacked on #403 (codex/pytorch-existing-snapshots), which provides the public snapshot API. It also includes the small latest #404 native follow-up because #403 has not yet absorbed it. Source baseline for the previous loader: fm-dataloader b4516394e98929c5b4752745866d5e3138276255.

Change map

Review the complete change map.

Testing done

  • Built and installed matching DiskGraph, engine, PyTorch and FM wheels outside the checkout. FM tests: 14 passed, including connector snapshot export/reload, schema aliases, occurrence alignment, 2/3/4-layer graph selection, and arrays/Arrow IPC parity.
  • Ran connector_to_prediction.py end to end on CPU against the shop connector and pretrained KumoRelational classifier. It produced predictions and both payload encodings from one retained batch.
  • The frozen four-event smoke case previously matched the DuckDB reference; raw timings and limits are in benchmark/fm-payloads/reports/omnistation-smoke.json.
  • Repository docs checker, locked uv resolution, and git diff --check passed. GitHub Actions pending.

Scope

kumo.fm is new and experimental. Supported future targets and graph plans, plus deliberate exclusions, are documented in packages/experimental/kumo-fm/README.md. No SDM dependency is added to the production package. A40 and HTTP parity runs have not been repeated on this stacked head.

@copy-pr-bot

copy-pr-bot Bot commented Nov 18, 2026

Copy link
Copy Markdown

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

@am-porter
am-porter temporarily deployed to gha-kumo-workflow-tests November 18, 2026 18:31 — with GitHub Actions Inactive
@coderabbitai

coderabbitai Bot commented Nov 18, 2026 •

Copy link
Copy Markdown

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

  • Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true
Pre-merge checks: Passed 1 · Failed 2

Failed Failed checks: 2 warnings

Check name Status Explanation Resolution
Change Map Design Review Warning Summary: The change map (PR comment 6987830022, begins with <!-- kumo-pr-change-map -->, author am-porter) states an explicit requirement and a plausible overall design. The map still has proven… Correct the change map comment (6987830022) before release review: 1. Restate the graph.rs row. Say the change alters the missing-extension error text (adds namespace and path to the canonicalize context) and that the paired test assert…
Pr Size Warning PR size check: FAIL. The PR changes more than 40 files. The best current estimate is 67 files. Evidence - The provider record gives the target as codex/pytorch-existing-snapshots. That branc… The PR changes 67 files by the best current estimate. The limit is 40. Split the PR into smaller reviewable units, for example: 1. The kumo.fm package source and its packaging files. 2. The kumo.fm tests and the DuckDB reference compari…
Passed Passed checks: 1 passed

Check name Status Explanation
Public Resource Access Passed Summary: The PR adds or modifies a small set of external resource references. Each reference I could test is publicly reachable without credentials, so the check passes. Scope: I treated commits 1fb5b…
Full details: Public Resource Access

Explanation

Summary: The PR adds or modifies a small set of external resource references. Each reference I could test is publicly reachable without credentials, so the check passes. Scope: I treated commits 1fb5bba through eab3312 (the PR-specific range on top of 4a68baa) as the PR change. The contributor description names kumo-fm, the benchmark, the example, and the CI workflow. Diffing from the earlier base 3153cf5 gave the same external references. Added or modified external references found: 1. packages/experimental/kumo-fm/examples/local_sdm/requirements.lock (new file, from 1fb5bba): structured-data-models @ git+https://github.com/NVIDIA/structured-data-models.git@fd2c29d035a165a998ea4fdb2b08a81fd89f71e1. The example README says this pinned SDM checkout is needed to run connector_to_prediction.py, so it is a required package source. Anonymous checks: https://github.com/NVIDIA/structured-data-models returned HTTP 200. git ls-remote on that URL worked without credentials. The GitHub commits API for fd2c29d... returned the commit object anonymously. Access: PASS. 2. packages/experimental/kumo-fm/src/kumo/fm/_dataset_support/relbench.py:57 (from 1fb5bba): https://huggingface.co/datasets/{REPOSITORY}/resolve/{REVISION}/{relative}, with REPOSITORY = "stanford-star/relbench-v1" and a pinned REVISION. Anonymous checks: the dataset API and a resolve/&lt;pinned revision&gt;/README.md URL both returned HTTP 200. Access: PASS. 3. Model reference nvidia/Kumo-Relational (v1.0.1) in the example README. This is a Hugging Face model id, not a literal URL. Anonymous checks: https://huggingface.co/api/models/nvidia/Kumo-Relational and the model tree page both returned HTTP 200. Access: PASS. 4. Repository-relative links in the PR docs and README (for example reports/omnistation-smoke.json, ../packages/experimental/kumo-pytorch/README.md, examples/local_sdm/connector_to_prediction.py). These resolve inside NVIDIA/kumo-engine, which returned HTTP 200 for anonymous GitHub access. Access: PASS. Excluded: the Apache license URLs in the unchanged LICENSE file, the pypi/files.pythonhosted.org URLs in uv.lock, the apache.org links, and code identifiers that only look like URLs. The uv.lock change only adds editable workspace entries for kumo-fm, and the pyproject.toml change is a single workspace member line. No added or modified reference was confirmed inaccessible. The check therefore passes.

Full details: Change Map Design Review

Explanation

Summary: The change map (PR comment 6987830022, begins with &lt;!-- kumo-pr-change-map --&gt;, author am-porter) states an explicit requirement and a plausible overall design. The map still has proven misleading explanations and omitted material behavior, so the check fails. These findings concern the map text. Uncertain design questions are listed separately. Scope and size: The PR has 6,797 additions and 2 deletions across 67 files. That is above 1,000 changed lines, so file-level scope descriptions are allowed. What holds up: - The requirement and the observed failure are explicit: "the first #391 version re-ingested a bundle and made callers orchestrate labels and graphs." - The public-plan claim holds. src/kumo/fm/_pytorch.py imports ColumnarGraph from kumo.pytorch.api. That module exports it through contract.batch.__all__. Proven issues: 1. Misdescribed native change. The map row for packages/diskgraph/diskgraph-index/src/graph.rs says "Merge current #404 path-identity fix." The PR diff for that file is 2 insertions and 2 deletions. It only changes the canonicalize error context to include namespace and subdir_rel. The paired test (test_runtime_batch_iterator_e2e.py) now expects the text "extension stage". No path-identity logic change appears in the diff. This is a user-visible error-text change for all diskgraph callers, and the map does not say so. 2. Omitted packaging behavior. src/kumo/fm/native_json.py (encode_rows) imports ._native_json and fails if the C++ extension is not built. Its comment reads "Fail explicitly if a source checkout was not built." batch.py uses this path for encoding == "arrays". The map row ("Port native array serialization | Preserve HTTP arrays encoding") does not say that arrays payloads need a compiled extension. It also does not justify C++ over the existing Arrow or Python options. 3. Omitted verification gap. The PR description says: "A40 and HTTP parity runs have not been repeated on this stacked head." The map's "Focused verification" lists only passed checks. Its stopping condition defers to "CI results" and does not name the parity gap. The stated requirement includes preserving the frozen FM request contract, and HTTP parity is the direct evidence for that. 4. Copy-pasted and inaccurate rows. - benchmark/fm-payloads/make_fixture.py only writes a fixture from test_integration.fixture. - benchmark/fm-payloads/README.md only gives instructions. - benchmark/fm-payloads/runner.py is the comparison runner. - All three rows say "Add fixture, comparison runner, and instructions." - src/kumo/fm/__init__.py is described as "Port FM contract and validation logic." It actually defines the new public API (Context, Query, FMLoader, PreparedRequest, PayloadBatch, prepare). The map never names that API. 5. Undescribed runtime dependencies. pyproject.toml adds duckdb==2.0.0, pyarrow==24.0.0 (exact pins), and kumo-pytorch as runtime dependencies. The row says only "Define source and wheel packaging." Uncertain design question (not a proven defect): src/kumo/fm/_dataset_support/ ships generation-side code in the runtime package. That includes generate.py, relbench.py (603 lines, RelBench source handling), presets/marketo.json (504 lines), tasks.py, and config.py. The runtime reaches this code only for dataset-manifest bundles. datasets.py imports Config, schema_for, and storage helpers. validate.py imports relbench only for RelBench-sourced bundles. No public API exposes generation. The map gives only "Preserve frozen dataset contract." It does not say whether generation is needed for connector-snapshot loading. The map should justify this scope or move the generation code out. Not addressed: The map does not say which fm-dataloader behavior was inspected before porting.

Resolution

Correct the change map comment (6987830022) before release review: 1. Restate the graph.rs row. Say the change alters the missing-extension error text (adds namespace and path to the canonicalize context) and that the paired test asserts it. Remove "path-identity fix" unless the diff shows that logic change. Name the user-visible effect on diskgraph callers. 2. State the native-extension requirement for arrays encoding. Say the source and wheel builds must compile _native_json, and that no Python fallback exists. Justify C++ over the existing options, or name the alternative considered. 3. Add the unrepeated A40 and HTTP parity runs to the verification section. Make them an explicit release condition, with the required parity evidence named. 4. Replace the copy-pasted rows with accurate per-file descriptions for make_fixture.py, README.md, runner.py, and __init__.py. Name the public kumo.fm API exported from __init__.py. 5. Add the new runtime dependencies and their exact pins to the pyproject.toml row. 6. Justify the generation-side _dataset_support modules (generate.py, relbench.py, marketo.json, tasks.py, config.py) against the stated requirement. Otherwise propose moving them out of the runtime package. 7. Replace the stopping condition with checkable criteria. Name the required CI jobs and parity runs, and state when the change is complete. 8. Add a sentence on which fm-dataloader behavior was inspected before porting.

Full details: Pr Size

Explanation

PR size check: FAIL. The PR changes more than 40 files. The best current estimate is 67 files. Evidence - The provider record gives the target as codex/pytorch-existing-snapshots. That branch is not in the local clone. The cited head 4b5f885e from the standing failure is also missing. The checkout is detached at eab33121. - The local history shows that 4a68baad ("Address snapshot reuse review findings") is the parent of the first FM commit 1fb5bbae ("Add experimental FM payload loader over DiskGraph"). It is the most likely tip of the stacked #403 base. - git diff --name-only 4a68baad HEAD lists 67 files. Rename detection does not change this count, because --name-only lists destination paths. - The 67 files include the packages/experimental/kumo-fm/ package, tests, examples, and packaging. They also include the CI workflows, benchmark files, root pyproject.toml, uv.lock, ARCHITECTURE.md, and docs/README.md. The PR description says it includes the #404 follow-up, and those changes are part of this count. - Other plausible bases give higher counts. 3b92023e gives 81 files. The standing run's base 9637e9cc gave 46 files for its own head. That head is not present here, so I did not re-verify it. - Against origin/main, the merge-base is b53781b1. The diff there is 123 files, but that base is not the PR's stacked base. Prior failure disposition The standing FAIL cited 46 files at head 4b5f885e against base 9637e9cc. I could not reproduce that head or base locally. The current revision still includes the same kumo-fm, CI, benchmark, and root-config areas. The current count is above 40 under every plausible base I could reconstruct. The prior failure therefore stands. Limitation The exact provider-reported base is not in the clone. The count of 67 assumes 4a68baad is the #403 tip. Under any base that yields a plausible stacked diff, the count stays above 40. Only the PR's own commit (5960636b) gives 15 files, and that is not a valid base. Check criteria The check fails when the count exceeds 40. Line counts were ignored.

Resolution

The PR changes 67 files by the best current estimate. The limit is 40. Split the PR into smaller reviewable units, for example: 1. The kumo.fm package source and its packaging files. 2. The kumo.fm tests and the DuckDB reference comparison. 3. The local SDM example and its notebook. 4. The benchmark files and the wheel and source CI workflow. 5. Shared docs and root pyproject.toml and uv.lock changes. 6. The #404 native follow-up, if it can be reviewed on its own. Each unit should stay at or below 40 files. Confirm the provider base branch codex/pytorch-existing-snapshots and the exact head before recounting.

  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

@am-porter

am-porter commented Nov 18, 2026 •

Copy link
Copy Markdown
Contributor Author

Change map

Requested release requirement: move the FM payload loader into kumo-engine as kumo.fm, adopt connector-created snapshots through the public PyTorch engine, and preserve the frozen FM request contract. The prior loader lives outside this repository; the first #391 version re-ingested a bundle and made callers orchestrate labels and graphs.

Smallest sufficient change: add the experimental package, a snapshot-backed FM loader over the existing index, source and wheel validation, an optional SDM example, and a frozen comparison runner. The production path uses public kumo.pytorch plans and retains Arrow tables for both encodings. Merge the current #404 native head because #403 has not yet absorbed its two-file follow-up.

Focused verification: 14 FM tests against installed wheels, including connector snapshot export/reload and integrated 2/3/4-hop sampling; a 72-line connector-to-SDM CPU example ran with a pretrained classifier; source and wheel builds, arrays and Arrow IPC checks, docs checker, and locked resolution passed. CI remains pending until the PR runs.

Stopping condition: no merge in this task; reviewers assess compatibility exclusions and CI results before release.

File What changed Why
.github/workflows/kumo-fm.yml Build and test source and finished wheels in CI Catch packaging and integration regressions
ARCHITECTURE.md Link the FM package from canonical docs Make the new component discoverable
benchmark/fm-payloads/README.md Add fixture, comparison runner, and instructions Reproduce correctness and timing comparisons
benchmark/fm-payloads/make_fixture.py Add fixture, comparison runner, and instructions Reproduce correctness and timing comparisons
benchmark/fm-payloads/reports/omnistation-smoke.json Record raw parity and timing draws Review measured smoke behavior
benchmark/fm-payloads/runner.py Add fixture, comparison runner, and instructions Reproduce correctness and timing comparisons
docs/README.md Link the FM package from canonical docs Make the new component discoverable
packages/experimental/kumo-fm/LICENSE Define source and wheel packaging Ship a standalone experimental package
packages/experimental/kumo-fm/MANIFEST.in Define source and wheel packaging Ship a standalone experimental package
packages/experimental/kumo-fm/README.md Document API, scope, and compatibility Guide use and review
packages/experimental/kumo-fm/examples/local_sdm/README.md Add optional local SDM adapter and executed notebook Demonstrate retained Arrow conversion and inference
packages/experimental/kumo-fm/examples/local_sdm/adapter.py Add optional local SDM adapter and executed notebook Demonstrate retained Arrow conversion and inference
packages/experimental/kumo-fm/examples/local_sdm/fixture/bundle/_COMPLETE.json Include the frozen validated example fixture Run the SDM notebook without external fixture setup
packages/experimental/kumo-fm/examples/local_sdm/fixture/bundle/manifest.json Include the frozen validated example fixture Run the SDM notebook without external fixture setup
packages/experimental/kumo-fm/examples/local_sdm/fixture/bundle/tables/categories/part-00000.parquet Include the frozen validated example fixture Run the SDM notebook without external fixture setup
packages/experimental/kumo-fm/examples/local_sdm/fixture/bundle/tables/interactions/part-00000.parquet Include the frozen validated example fixture Run the SDM notebook without external fixture setup
packages/experimental/kumo-fm/examples/local_sdm/fixture/bundle/tables/items/part-00000.parquet Include the frozen validated example fixture Run the SDM notebook without external fixture setup
packages/experimental/kumo-fm/examples/local_sdm/fixture/bundle/tables/users/part-00000.parquet Include the frozen validated example fixture Run the SDM notebook without external fixture setup
packages/experimental/kumo-fm/examples/local_sdm/fixture/bundle/tasks/binary_classification/context.parquet Include the frozen validated example fixture Run the SDM notebook without external fixture setup
packages/experimental/kumo-fm/examples/local_sdm/fixture/bundle/tasks/binary_classification/labels.parquet Include the frozen validated example fixture Run the SDM notebook without external fixture setup
packages/experimental/kumo-fm/examples/local_sdm/fixture/bundle/tasks/binary_classification/predict.parquet Include the frozen validated example fixture Run the SDM notebook without external fixture setup
packages/experimental/kumo-fm/examples/local_sdm/fixture/bundle/tasks/regression/context.parquet Include the frozen validated example fixture Run the SDM notebook without external fixture setup
packages/experimental/kumo-fm/examples/local_sdm/fixture/bundle/tasks/regression/labels.parquet Include the frozen validated example fixture Run the SDM notebook without external fixture setup
packages/experimental/kumo-fm/examples/local_sdm/fixture/bundle/tasks/regression/predict.parquet Include the frozen validated example fixture Run the SDM notebook without external fixture setup
packages/experimental/kumo-fm/examples/local_sdm/fixture/bundle/validation.json Include the frozen validated example fixture Run the SDM notebook without external fixture setup
packages/experimental/kumo-fm/examples/local_sdm/fixture/workload.json Include the frozen validated example fixture Run the SDM notebook without external fixture setup
packages/experimental/kumo-fm/examples/local_sdm/requirements.lock Add optional local SDM adapter and executed notebook Demonstrate retained Arrow conversion and inference
packages/experimental/kumo-fm/examples/local_sdm/walkthrough.ipynb Add optional local SDM adapter and executed notebook Demonstrate retained Arrow conversion and inference
packages/experimental/kumo-fm/pyproject.toml Define source and wheel packaging Ship a standalone experimental package
packages/experimental/kumo-fm/setup.py Define source and wheel packaging Ship a standalone experimental package
packages/experimental/kumo-fm/src/kumo/fm/__init__.py Port FM contract and validation logic Preserve bundle, workload, and wire behavior
packages/experimental/kumo-fm/src/kumo/fm/_assembly.py Build and retain FM Arrow batches over public graph API Sample once and encode without resampling
packages/experimental/kumo-fm/src/kumo/fm/_contract.py Port FM contract and validation logic Preserve bundle, workload, and wire behavior
packages/experimental/kumo-fm/src/kumo/fm/_dataset_support/__init__.py Port bundle generation and validation support Preserve frozen dataset contract
packages/experimental/kumo-fm/src/kumo/fm/_dataset_support/config.py Port bundle generation and validation support Preserve frozen dataset contract
packages/experimental/kumo-fm/src/kumo/fm/_dataset_support/generate.py Port bundle generation and validation support Preserve frozen dataset contract
packages/experimental/kumo-fm/src/kumo/fm/_dataset_support/presets/marketo.json Port bundle generation and validation support Preserve frozen dataset contract
packages/experimental/kumo-fm/src/kumo/fm/_dataset_support/relbench.py Port bundle generation and validation support Preserve frozen dataset contract
packages/experimental/kumo-fm/src/kumo/fm/_dataset_support/storage.py Port bundle generation and validation support Preserve frozen dataset contract
packages/experimental/kumo-fm/src/kumo/fm/_dataset_support/tasks.py Port bundle generation and validation support Preserve frozen dataset contract
packages/experimental/kumo-fm/src/kumo/fm/_dataset_support/validate.py Port bundle generation and validation support Preserve frozen dataset contract
packages/experimental/kumo-fm/src/kumo/fm/_encoding.py Port FM contract and validation logic Preserve bundle, workload, and wire behavior
packages/experimental/kumo-fm/src/kumo/fm/_native_json.cpp Port native array serialization Preserve HTTP arrays encoding
packages/experimental/kumo-fm/src/kumo/fm/_phase.py Port FM contract and validation logic Preserve bundle, workload, and wire behavior
packages/experimental/kumo-fm/src/kumo/fm/_pytorch.py Build and retain FM Arrow batches over public graph API Sample once and encode without resampling
packages/experimental/kumo-fm/src/kumo/fm/_validation.py Port FM contract and validation logic Preserve bundle, workload, and wire behavior
packages/experimental/kumo-fm/src/kumo/fm/batch.py Build and retain FM Arrow batches over public graph API Sample once and encode without resampling
packages/experimental/kumo-fm/src/kumo/fm/bundle.py Port FM contract and validation logic Preserve bundle, workload, and wire behavior
packages/experimental/kumo-fm/src/kumo/fm/datasets.py Port FM contract and validation logic Preserve bundle, workload, and wire behavior
packages/experimental/kumo-fm/src/kumo/fm/loader.py Build and retain FM Arrow batches over public graph API Sample once and encode without resampling
packages/experimental/kumo-fm/src/kumo/fm/native_json.py Port native array serialization Preserve HTTP arrays encoding
packages/experimental/kumo-fm/src/kumo/fm/native_selection.py Port FM contract and validation logic Preserve bundle, workload, and wire behavior
packages/experimental/kumo-fm/src/kumo/fm/workload.py Port FM contract and validation logic Preserve bundle, workload, and wire behavior
packages/experimental/kumo-fm/tests/_fixture.py Cover contract, encoding, sampling, and SDM conversion Verify migration behavior
packages/experimental/kumo-fm/tests/test_batch.py Cover contract, encoding, sampling, and SDM conversion Verify migration behavior
packages/experimental/kumo-fm/tests/test_contract.py Cover contract, encoding, sampling, and SDM conversion Verify migration behavior
packages/experimental/kumo-fm/tests/test_integration.py Cover contract, encoding, sampling, and SDM conversion Verify migration behavior
packages/experimental/kumo-fm/tests/test_sdm_adapter.py Cover contract, encoding, sampling, and SDM conversion Verify migration behavior
pyproject.toml Register and lock the workspace package Resolve the package in the engine workspace
uv.lock Register and lock the workspace package Resolve the package in the engine workspace
packages/diskgraph/diskgraph-index/src/graph.rs Merge current #404 path-identity fix Use the latest native prerequisite while #403 awaits that follow-up
packages/diskgraph/diskgraph-pybind/python/tests/test_runtime_batch_iterator_e2e.py Merge current #404 regression assertion Cover the native path-identity fix in the stacked head
packages/experimental/kumo-fm/examples/local_sdm/connector_to_prediction.py Add concise connector-to-model workflow Make the snapshot-backed FM path runnable without manual FM metadata or anchors
packages/experimental/kumo-fm/src/kumo/fm/_snapshot.py Derive metadata and orchestrate labels/graphs over adopted snapshots Share retained Arrow assembly without a second base ingest
packages/experimental/kumo-fm/tests/test_connector_snapshot.py Test connector export/reload, alias, labels and wire payloads Verify installed-wheel cross-stack behavior and immutable base
packages/experimental/kumo-fm/tests/test_snapshot_api.py Test occurrence, schema and label alignment rules Catch duplicates, missing labels and unsupported plans
packages/experimental/kumo-fm/tests/test_snapshot_integration.py Test integrated 2/3/4-layer sampling Verify selected edges and per-occurrence membership
tests/public_setup/Dockerfile.dockerignore Include the new workspace manifest in the public setup build context Keep connector CI's workspace metadata fixture complete
scripts/check_python_license_metadata.py Register the FM distribution and its license artifact Include the new package in repository licensing checks
scripts/gen_third_party.py Classify FM as a first-party distribution Keep dependency notices accurate for the workspace package
test/scripts/test_python_license_metadata.py Update the expected distribution inventory count Verify the new manifest participates in licensing validation

@am-porter
am-porter temporarily deployed to gha-kumo-workflow-tests November 18, 2026 18:52 — with GitHub Actions Inactive
@am-porter
am-porter force-pushed the codex/fm-loader-migration branch from 60b2f25 to eab3312 Compare November 19, 2026 01:09
@am-porter
am-porter had a problem deploying to gha-kumo-workflow-tests November 19, 2026 01:09 — with GitHub Actions Error
@am-porter
am-porter changed the base branch from main to codex/pytorch-existing-snapshots November 19, 2026 01:09
@am-porter
am-porter had a problem deploying to gha-kumo-workflow-tests November 19, 2026 01:12 — with GitHub Actions Error
@am-porter
am-porter had a problem deploying to gha-kumo-workflow-tests November 19, 2026 01:15 — with GitHub Actions Error
@am-porter
am-porter deployed to gha-kumo-workflow-tests November 19, 2026 01:21 — with GitHub Actions Active

This branch was successfully deployed

1 active deployment
gha-kumo-workflow-tests — e998187d Deployed Nov 19, 2026 by am-porter via import #2665
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant