example: regenerate the worked example from current Tartu sources - #28
Conversation
![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: 4ca7e02b7e
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".
| row_count: 79082 | ||
| column_count: 32 | ||
| request_spec: 'Direct county GPKG zip download from Maa- ja Ruumiamet S3 storage: https://s3.pilw.io/rp-kemit-kataster/ANDMED/Tartu_maakond_KATASTER_GPKG.zip | ||
| extracted to data/source/Tartu_maakond_KATASTER_GPKG.gpkg containing layer ''Tartu maakond'' (79,056 |
There was a problem hiding this comment.
This refresh changes the cadastral source from 79,056 to 79,082 rows and the run record changes its source SHA, but the surrounding provenance still identifies it as the August 25 snapshot, retains the August 25 retrieval time and previous ETag, and even describes the extracted file as containing 79,056 parcels. When someone attempts to reproduce the September run, these fields point to a different snapshot; update the retrieval/version metadata and remove the hard-coded August metadata in the pipeline.
Useful? React with 👍 / 👎.
| - Tier 2 — road plus either municipal education proxy: **85 parcels / 2,472.3 ha** | ||
| - Tier 3 — road access only: **367 parcels / 6,249.4 ha** | ||
| - Total road-accessible candidates: **518 parcels / 9,570.8 ha** | ||
| - Total road-accessible candidates: **517 parcels / 9,570.8 ha** |
There was a problem hiding this comment.
The refreshed embedded data has 84 Tier 2 parcels (2,468.3 ha), 517 total parcels (9,566.9 ha), 59 scenario-road candidates, and 27 active schools out of 62 source facilities, but this edit updates only the total count while retaining the old total area and all of those adjacent figures. Readers of “Current regenerated result” therefore see a mixture of two runs; regenerate the entire summary block rather than changing only this count.
AGENTS.md reference: AGENTS.md:L144-L144
Useful? React with 👍 / 👎.
The regeneration updated the candidate total but left the figures around it describing the previous run, so the section contradicted the artifacts shipped beside it: the three tiers summed to 518 against a stated 517. Recomputed from the dashboard data this commit ships: - Tier 2 lost the parcel and 4.0 ha (85 / 2,472.3 ha -> 84 / 2,468.3 ha); Tier 1 and Tier 3 are unchanged, so the total is 9,566.9 ha, not 9,570.8; - 59 candidates now reach their nearest qualifying road by the scenario geometry, not 60; - the authoritative education source published a school: 62 facilities, 27 of them schools, with the same single kindergarten switched off by OVERRIDE-001; - removing that override therefore moves 63 parcels rather than 64, giving a baseline Tier 1 of 129 parcels / 3,181.2 ha. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DqoLGyKY5opHGjNMSerpHg
jaakla
changed the title
1 participant
What
Regenerates
examples/tartu-development/from the live Estonian sources and commits the artifacts that changed, clearing the artifact-currency failure onmain's scheduled run.Upstream had moved since the last regeneration:
OVERRIDE-001);The
README.md"Current regenerated result" figures are recomputed from the dashboard data this PR ships, so the prose and the artifacts agree — the three tiers now sum to the stated total.Evidence
The
worked-examplejob passes on this branch, and this branch predates #27, so the artifact-currency step ran here as a hard failure rather than advisory. That is the check that was red onmain, so its passing is the proof the regeneration is what CI produces.validateandverifyreportwarning, which is this project's honest status — the education source publishes no reuse license and the walking criterion uses a straight-line proxy.Note
Two run records are included.
run-20260909-092852is a run that reused the local source cache;run-20260909-093408is the fresh one and is whatruns.latestpoints at. Keeping both matches howruns/has accumulated historically.🤖 Generated with Claude Code
https://claude.ai/code/session_01DqoLGyKY5opHGjNMSerpHg