[limen GEN-a-organvm-metasystem-master-test-coverage-0620] Raise test coverage in a-organvm/metasystem-master - #33
Conversation
limen task GEN-a-organvm-metasystem-master-test-coverage-0620
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 28010d631a
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| const hasVariance = deviations.some(deviation => deviation > 0.001); | ||
| if (!hasVariance) return inputs; | ||
|
|
||
| return inputs.filter(input => Math.abs(input.value - medianValue) <= 0.15); |
There was a problem hiding this comment.
When the median absolute deviation collapses to zero, this branch ignores the caller-provided threshold/config.outlierThreshold and hard-codes a 0.15 cutoff. In any deployment that raises the threshold to relax or disable outlier filtering, e.g. mostly identical inputs like [0.5, 0.5, 0.5, 0.7] with outlierThreshold: 100, computeConsensus still discards 0.7, so the documented configuration no longer controls consensus behavior.
Useful? React with 👍 / 👎.
| if (std < 0.001) return inputs; // No variance | ||
|
|
||
|
|
||
| const medianValue = median(inputs.map(input => input.value)); |
There was a problem hiding this comment.
This median/MAD calculation ignores the weight field even though computeConsensus passes in inputs that were already weighted by spatial proximity, recency, and agreement. In a live window where several stale or far-away low-weight values outnumber a smaller set of recent/front-row high-weight values, the unweighted median can classify the high-weight values as outliers and remove them before weightedMean, effectively overriding the weighting model that is supposed to give those inputs more influence.
Useful? React with 👍 / 👎.
|
Disposition: evolving Verified locally and against GitHub checks. The branch contains real core-engine test work: on the PR checkout, pnpm --filter @omni-dromenon/core-engine test passes (3 files / 60 tests), and it fixes the local current-master core test failures around crypto.v4/randomUUID and outlier filtering. But the PR is stale against current master: its direct diff would also touch stale.yml/README/pnpm-lock.yaml, and GitHub required checks are red because pnpm install --frozen-lockfile sees an outdated lockfile. Smallest next step: rebase/cherry-pick only the core-engine files onto current master, regenerate lock if needed, rerun CI, then merge. |
4444J99
added
the
lifecycle:blocked
1 participant
Autonomous limen dispatch of task
GEN-a-organvm-metasystem-master-test-coverage-0620.Find the largest source module in a-organvm/metasystem-master with little or no test coverage and add a focused, PASSING test suite for it. Run the repo's own test command and confirm green. No placeholder tests. [auto-generated 2026-06-20 to keep the stream endless]
Produced in an isolated worktree off origin — review before merge.