Cleanup untracked scratch on kg-microbe main tree

[realmarcin]

Context

During the 2026-07-02 branch triage on `chore/refresh-unified-chemical-mappings`, `git status` surfaced 126 untracked entries in the working tree. Most were captured by the new `.gitignore` rules in #578 — but ~30 entries need per-file decisions before they can be either committed, deleted, or gitignored.

What still needs a call

New utility modules (no in-repo references — either real work on the wrong branch, or dead scratch)

• `kg_microbe/utils/biohub_converter.py`
• `kg_microbe/utils/diagnose_duplicates.py`
• `kg_microbe/utils/extract_taxon_strain_nodes.py`
• `kg_microbe/utils/nlp_utils.py`
• `kg_microbe/utils/parse_taxon_rank.py` (plus `.py_bak` sibling — the `_bak` is now gitignored by Gitignore local build artifacts and dated snapshot dirs #578)
• `kg_microbe/utils/transform_utils.py` (name collides with the top-level `transform_utils/` package — check for shadowing bugs)

Scripts at repo root (root-level Python is almost always a one-off)

• `run.py` — likely wrong (canonical entrypoint is `poetry run kg`)
• `analyze_categories.py`
• `download_mediadive_bulk.py` — probably belongs alongside the download.yaml MediaDive work; move under `kg_microbe/` or `scripts/` if permanent
• `convert_merged_to_nt.yaml`

Unknown directories (peek and decide)

• `kg_microbe/transform_utils/data/`
• `kg_microbe/transform_utils/ontology/` — possible typo of the existing `ontologies/` package

Data / mapping / xref files

• `kg_microbe/transform_utils/bacdive/metabolite_mapping.json`
• `kg_microbe/transform_utils/bactotraits/BactoTraits.tsv` — 6.2 MB. Per the download.yaml disable PR, BactoTraits V2 is now vendored because upstream TLS is broken. Probably needs to be committed alongside the download.yaml change.
• `kg_microbe/transform_utils/ontologies/xrefs/mondo_xrefs.tsv`
• `kg_microbe/transform_utils/ontologies/xrefs/unipathways_xrefs.tsv`

Docs

• `docs/MERGE_CLEANUP.md`
• `docs/metatraits/unmapped_compounds.tsv`
• `docs/metpo/` — likely belongs with the METPO commits in Fix pre-existing tox failures: METPO chemical-tolerance predicates + proposal collisions + docstr gaps #577

Mappings

• `mappings/mediadive_unmapped_ingredients_to_curate.tsv` — generated by `scripts/dump_unmapped_mediadive_ingredients.py`; either commit as a curation-state snapshot, or gitignore

Notebooks

Rule of thumb: notebooks with a clear analytical purpose can be committed; `.ipynb_bak`, `_withoutput.ipynb`, `catboost_info/`, and rendered `.html` are now gitignored by #578.

• `notebooks/bacdive-api-summary-table.ipynb`
• `notebooks/bacdive-process-errors-with-genus.ipynb`
• `notebooks/bacdive-summary-counts.ipynb`
• `notebooks/bacdive_reformat.ipynb`
• `notebooks/evaluate_link_pred.ipynb`
• `notebooks/kg_bacdive.ipynb`
• `notebooks/kg_microbe_embedding.ipynb`
• `notebooks/kg_microbe_train_taxa_to_media.ipynb`
• `notebooks/load_feba.ipynb` — plus `load_feba.ipynb` at repo root (duplicate — pick one location)
• `notebooks/train_taxa_to_media.ipynb`

Decision checklist per file

1. Is anything in the repo already importing / referencing it? `grep -r 'basename' .` — no hits = not wired in.
2. Was it ever tracked? `git log --all --follow -- ` — empty = fresh scratch.
3. Would losing it be a problem tomorrow? If not, delete.

Related

• Branch cleanup PR: Gitignore local build artifacts and dated snapshot dirs #578 (adds .gitignore rules that hide ~90 of the 126 entries)
• Sibling issue: "Untrack 7 regenerable pipeline outputs in `transform_utils/*/tmp/`"