Expand regression coverage for pipeline edge cases

[giosilvi]

Summary

• add candidate generation regression tests for span distance, pair caps, and noun-phrase fallback behavior
• extend dedupe, export, and NER CLI suites to cover normalization gaps, attribute failures, and multi-sentence tagging
• harden relation scoring and inference tests for missing models, threshold filtering, and predicate type bypass paths

Testing

• pytest
• python tools/run_trace_coverage.py --min 80

---

https://chatgpt.com/codex/tasks/task_e_68fb3f383a70832dbe9560f9f3e11c3c

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

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".

[chatgpt-codex-connector]

Guard numpy stub against already-imported numpy

The tests rely on the helper variable fake_np, but it is only created when 'numpy' is not yet present in sys.modules. If any plugin or dependency imports numpy before this module, the if block is skipped and fake_np is never defined, so FakeSession.run will raise NameError when it accesses fake_np.zeros. Consider always defining a local alias (e.g. import numpy as np and fall back to a stub when the import fails) so the tests work regardless of whether numpy is preloaded.

Useful? React with 👍 / 👎.