Datasets:

wirthal1990-tech
/

USDA-Phytochemical-Database-JSON

Timestamp (UTC): 2026-04-21T12:44:45Z
Scope:
- /opt/ethno-api/ethno-enrichment/USDA-Phytochemical-Database-JSON v2
- /opt/ethno-api/nexus_api/pseo
Live URLs:
- https://ethno-api.com/
- https://ethno-api.com/enterprise.html
Constraints honored:
- no force push
- no reset --hard
- additive/corrective commits only

Checkpoint 0 - Soll/Ist before correction

Topic	Soll	Ist before correction	Evidence
Remote heads	origin/main == hf/main (or explicit divergence)	origin/main=dea08aebe0e7381d108ebd452c456f102693770c, hf/main=ab3a21ced9f43637054b7d71fbe4f97d22e319ec	`git ls-remote origin refs/heads/main`, `git ls-remote hf refs/heads/main`
Merge relation	deterministic relation known	merge-base=dea08aebe0e7381d108ebd452c456f102693770c, origin ancestor of hf	`git merge-base`, `git merge-base --is-ancestor`
Canonical local landing source	should match live markers	/opt/ethno-api/nexus_api/pseo/enterprise.html had DOI_NEW=0, DOI_OLD=4, V240=0, V231=11, field markers=0	`grep -o ...
Version authority	current version should reflect active release	`Current version: v2.3.1 (April 2026)` while v2.4.0 changelog present	`grep -n 'Current version:' UPDATE_POLICY.md`

Deterministic baseline measurement (git + live + local markers).
Remote divergence handling without rewrite:
- fast-forward local main to hf/main: git merge --ff-only hf/main
- result local head: ab3a21ced9f43637054b7d71fbe4f97d22e319ec
Version authority correction:
- file: UPDATE_POLICY.md
- change: Current version: v2.3.1 -> Current version: v2.4.0
- commit: dba76475dd07e3c4703b142000a757d5f991d340
Source-of-truth correction for landing page:
- canonical file set to: /opt/ethno-api/nexus_api/pseo/enterprise.html
- updated from live enterprise response and re-verified with marker counts.
Pushes (non-destructive):
- git push origin main
- git push hf main
SC1 incident recovery (post-push):

observed temporary regression: viewer:false and first-rows parse error
root cause: README frontmatter dataset_info.features had dtype mismatches vs actual parquet schema
fix: aligned dtypes in README for patent_count_since_2020, inchi_key, iupac_verified
commit: 912fd328ff7d6852ca60b258cfc73a666ef37d97
re-verification: viewer:true, first-rows returns feature payload

Commands:

git rev-parse HEAD
GIT_TERMINAL_PROMPT=0 git ls-remote origin refs/heads/main
GIT_TERMINAL_PROMPT=0 git ls-remote hf refs/heads/main

Measured:

Result:

Canonical local file:

Measured local canonical markers:

Measured live markers:

https://ethno-api.com/
- BYTES=89560
- DOI_NEW=4, DOI_OLD=0, V240=11, V231=0
- partner_cid=2, inchi_key=2, iupac_verified=2, partner_match_method=2
https://ethno-api.com/enterprise.html
- BYTES=89560
- DOI_NEW=4, DOI_OLD=0, V240=11, V231=0
- partner_cid=2, inchi_key=2, iupac_verified=2, partner_match_method=2

Note on SHA drift:

Marker and byte-size alignment is exact.
Full-file SHA differs between requests due Cloudflare dynamic challenge token injection in response body tail.
This was validated by cmp -l with only 14 differing bytes near the CF token segment.

Command:

grep -n 'Current version:' UPDATE_POLICY.md

Measured:

Result:

SC	Soll	Measurement command(s)	Ist	Result
SC1	HF Dataset Viewer available	`curl -4 -s "https://datasets-server.huggingface.co/is-valid?dataset=wirthal1990-tech/USDA-Phytochemical-Database-JSON"`	`{... "viewer":true, "search":true, "filter":true ...}`	PASS
SC2	16-field marker presence on live (new 4 fields each x2)	`grep -o 'partner_cid\|inchi_key\|iupac_verified\|partner_match_method'` on live body	each marker count = 2	PASS
SC3	New DOI present, old DOI absent	`grep -o '19660107'` and `grep -o '19265853'` on live + repo files	live: 4/0, README: 6/0, llms: 3/0, UPDATE_POLICY: 3/0, noise list: 1/0	PASS
SC4	Build vs Buy EUR values present	`grep -o '16,900'`, `grep -o '11,300'` on live; `grep -n '16,900\|11,300' README.md`	live: 1 and 1, README line 168 shows `~€16,900 / ~€11,300`	PASS
SC5	Repo update consistency	marker counts in README/llms/UPDATE_POLICY/noise_exclusion/METHODOLOGY + unified remote head	counts consistent and both remotes on same head	PASS
SC6	No v2.3.1 on live landing	`grep -o 'v2\.3\.1'` on live URLs	0	PASS

Required files present in dataset repo.
Canonical landing source named and present: /opt/ethno-api/nexus_api/pseo/enterprise.html.
Persistent report created: SC_VERIFICATION_v2.4.0_2026-04-21.md.
Status: PASS.

Initial failing evidence:

curl -4 -s "https://datasets-server.huggingface.co/is-valid?dataset=wirthal1990-tech/USDA-Phytochemical-Database-JSON" returned {"preview":false,"viewer":false,"search":false,"filter":false,"statistics":false}
curl -4 -s "https://datasets-server.huggingface.co/first-rows?dataset=wirthal1990-tech/USDA-Phytochemical-Database-JSON&config=default&split=train" returned parse error for string-to-double conversion.

Resolved evidence:

is-valid now returns {"preview":true,"viewer":true,"search":true,"filter":true,"statistics":false}
first-rows now returns row/feature payload.

LOW: Full-file SHA comparison against live is unstable because Cloudflare injects per-request challenge token bytes into HTML tail. Marker-level and byte-size checks remain reproducible.

GitHub main and HF main synchronized with identical head (reproducible via Checkpoint 1 commands).
Canonical local landing source aligned to live markers and byte size.
SC1-SC6 all PASS with numeric evidence.