Separate source PRs from releases; add render_sample.py (#42)

- Separate source PRs from releases; add render_sample.py (1a89e023850ce2b269ebcd50310d4870eca85972)

.gitattributes

@@ -59,6 +59,6 @@ data/samorzad_gov_pl/*.jsonl filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
-data/sejm_api/sejm_api.attribution.jsonl filter=lfs diff=lfs merge=lfs -text
-data/uzp_orzeczenia_pl/uzp_orzeczenia_pl.attribution.jsonl filter=lfs diff=lfs merge=lfs -text
-data/uzp_orzeczenia_pl/uzp_orzeczenia_pl.decisions.jsonl filter=lfs diff=lfs merge=lfs -text
+# Per-source audit trails: one rule for all sources, so source PRs don't edit this file
+data/*/*.attribution.jsonl filter=lfs diff=lfs merge=lfs -text
+data/*/*.decisions.jsonl filter=lfs diff=lfs merge=lfs -text

AGENTS.md

@@ -1,8 +1,9 @@
 # AGENTS.md
 
 Guidance for agents and contributors adding a source and cutting a release.
-Read this before opening a PR — several recent contributions added data but
-skipped the doc/version bump, forcing a maintainer cleanup release.
+Read this before opening a PR. Source PRs and releases are separate: a source PR
+adds one source with `release: None`; the maintainer later cuts one release that
+bumps the version and docs for every source merged since the last one.
 
 ## How the docs are actually maintained
 
@@ -54,6 +55,24 @@ answer was yes or no.
    text, positive token counts, uniform source/license, stats-file consistency).
    Run `python3 -m pytest src/` before committing.
 
+### Keep the PR to your source
+
+Many source PRs are open at once, and each one that touches shared files
+conflicts with the others. A source PR changes only:
+
+- `data/<key>/*`, `src/fetch_<key>.py` / `src/clean_<key>.py`,
+  `src/test_<key>_contract.py`;
+- **one** new entry in `src/sources.py`, with `release: None`;
+- your entry appended to `artifacts/source_findings.md`.
+
+It does **not** touch `README.md`, `CHANGELOG.md`, `VERSION`/`RELEASE_DATE` in
+`src/make_docs.py`, the phrase-frequency report, `AGENTS.md`, or other sources'
+fetchers — those change in the release commit (below) or in their own PR.
+`.gitattributes` already LFS-tracks `data/*/*.attribution.jsonl` and
+`*.decisions.jsonl`; do not add per-file lines.
+Commit with LF line endings — a CRLF `sources.py` shows as a whole-file rewrite
+and conflicts with every other PR. Rebase onto `main` before asking for review.
+
 ## What the PR must show
 
 These two are not optional — they are the difference between a source a reviewer
@@ -64,7 +83,31 @@ example other contributors will copy.
    (or truncated ones), verbatim, so a reviewer can see at a glance what kind of
    text this actually is — prose, transcripts, boilerplate, OCR noise, HTML
    leftovers. Include the metadata columns too, not just `text`. Don't make the
-   reviewer download a parquet to find out.
+   reviewer download a parquet to find out — and a linked `.sample.jsonl` is not
+   a substitute: on the Hub it renders as one escaped line per document.
+
+   Generate it from the built shard and paste the output into the PR description:
+   ```bash
+   python3 src/render_sample.py data/<key>/<key>.parquet -n 3
+   ```
+   Each document is a fenced block — metadata as `key: value` lines, a blank
+   line, then the text with its real line breaks, truncated with `[…]`:
+   ````
+   ```
+   id: sejm_interpellations_2388
+   source: sejm_interpellations
+   added: 2026-07-19
+   token_count: 415
+   license: public-domain (official documents)
+   author: Daniel Milewski
+
+   Szanowny Panie Ministrze,
+
+   latem 2022 roku na podlaskiej części granicy z Białorusią ukończono […]
+   ```
+   ````
+   Pick rows that show the typical document; if the source has distinct
+   document types (question vs reply, article vs talk page), show one of each.
 
 2. **Provenance and licensing, written out in the datasheet (`data/<key>/<key>.md`)
    and summarized in the PR description:**
@@ -119,6 +162,10 @@ What a fetcher should do to `text` before writing the `.jsonl.zst`:
   parquet is written (frozen regex ∪ model; placeholders `[PII]` / `[Telefon]`).
   Names of public officials acting in an official capacity are not PII and
   should stay — removing them would gut parliamentary and judicial sources.
+  A source scrubbed some other way may still merge with `release: None`, but it
+  is not admitted to a release until its shard has been re-run through NERGAL.
+  Say in the datasheet which scrubber ran. Note that `main` is public on the Hub
+  whether or not a source is released.
 
 Two rules about all of the above:
 
@@ -129,11 +176,13 @@ Two rules about all of the above:
    Include a handful of before/after excerpts in the PR — this is the fastest way
    for a reviewer to see whether normalization ate real content.
 
-## Cut the release (the part that gets skipped)
+## Cut the release (maintainer, one commit after the source PRs merge)
 
+0. In `src/sources.py`, set `release` on each merged source this release ships
+   (check its shard went through NERGAL first).
 1. In `src/make_docs.py`: bump `VERSION` and `RELEASE_DATE`. Add a row for the new
-   version to the version table and a row for yourself to the Contributors table
-   (both are literal rows in the template).
+   version to the version table and a row for each new contributor to the
+   Contributors table (both are literal rows in the template).
 2. Regenerate the phrase-frequency report and charts (registry-independent — it
    globs `data/*/*.parquet`):
    ```bash

data/rock_pollub_pl/rock_pollub_pl.attribution.jsonl

@@ -1,8 +1,3 @@
-{"author": "Gułkowski, Sławomir", "authors": ["Gułkowski, Sławomir"], "editors": [], "eisbn": ["978-83-7947-674-9"], "extracted_text": {"bytes": 312627, "extractor": "pdfplumber", "page_count": 171, "page_failures": [], "sha256": "a0e3089beb113db20ca75e48d146354245d20cf13d2ac5835e25b98c0cf374f2"}, "handle": "20.500.14629/24876", "id": "rock_pollub_pl_513c690d-72b3-4604-b62e-e21e44b91645", "isbn": ["978-83-7947-673-2"], "issued": ["2026"], "landing_url": "https://rock.pollub.pl/entities/publication/513c690d-72b3-4604-b62e-e21e44b91645", "license": "CC-BY-SA-4.0", "license_evidence": {"item": ["CC-BY-SA-4.0"], "pdf_bitstream": ["CC-BY-SA 4.0"]}, "license_policy_url": "https://wpl.pollub.pl/pl/i/Polityka-publikacyjna/20", "license_url": "https://creativecommons.org/licenses/by-sa/4.0/", "pdf": {"bytes": 12633380, "checksum": {"checkSumAlgorithm": "MD5", "value": "f69d32dda85d12c91d9cd96a7bb3fe52"}, "content_url": "https://rock.pollub.pl/server/api/core/bitstreams/425c2883-2d3b-4260-b2ff-c07f6628f56d/content", "id": "425c2883-2d3b-4260-b2ff-c07f6628f56d", "name": "modelowanie_i_analiza_9788379476749.pdf", "rights": ["CC-BY-SA 4.0"], "sha256": "a98b86ec1cffd62336b4c708ed4cf18efea78088d55f35a0b2850fa326e577fc"}, "publisher": "Politechnika Lubelska", "subtype": ["Monograph"], "text_sha256": "4c8de49a4349d0cddc8110c7e8facf31dd3309a56d5029b62af92d65a75b2ae1", "title": "Modelowanie i analiza eksperymentalna parametrów pracy modułów fotowoltaicznych w warunkach rzeczywistych", "transformations": ["pdfplumber page extraction", "Unicode/whitespace normalization", "page-number-only removal", "line-wrap repair", "limited email/labelled-phone redaction"]}
-{"author": "Szuster, Janusz J.", "authors": ["Szuster, Janusz J."], "editors": [], "eisbn": ["978-83-7947-647-3"], "extracted_text": {"bytes": 304945, "extractor": "pdfplumber", "page_count": 213, "page_failures": [], "sha256": "f557c50de3c272fafce1beeb75ee3bca816e2b966b631352001ee4221868ee46"}, "handle": "20.500.14629/24871", "id": "rock_pollub_pl_e9e7a67e-c405-4968-b87a-0ff169c7b3d1", "isbn": ["978-83-7947-646-6"], "issued": ["2026"], "landing_url": "https://rock.pollub.pl/entities/publication/e9e7a67e-c405-4968-b87a-0ff169c7b3d1", "license": "CC-BY-SA-4.0", "license_evidence": {"item": ["CC-BY-SA-4.0"], "pdf_bitstream": ["CC-BY-SA 4.0"]}, "license_policy_url": "https://wpl.pollub.pl/pl/i/Polityka-publikacyjna/20", "license_url": "https://creativecommons.org/licenses/by-sa/4.0/", "pdf": {"bytes": 5813455, "checksum": {"checkSumAlgorithm": "MD5", "value": "20e8f6af489fbea09a84c4bb83eea9e0"}, "content_url": "https://rock.pollub.pl/server/api/core/bitstreams/b1a29ec6-ad9b-46ea-acc4-59d1c3fe2b0f/content", "id": "b1a29ec6-ad9b-46ea-acc4-59d1c3fe2b0f", "name": "MatematykaFinansowa_9788379476473.pdf", "rights": ["CC-BY-SA 4.0"], "sha256": "2d4ff03c21d753c44e3e80cec8b1af4ce3358591dd88b322c7756b7d022db957"}, "publisher": "Politechnika Lubelska", "subtype": ["Handbook"], "text_sha256": "9b67845bb49aa6b20c8a41f2566e09411d9b85e1542d71f3f8ba9daba4f9d42c", "title": "Ćwiczenia i zadania ze wstępu do matematyki finansowej", "transformations": ["pdfplumber page extraction", "Unicode/whitespace normalization", "page-number-only removal", "line-wrap repair", "limited email/labelled-phone redaction"]}
-{"author": "Kulisz, Monika", "authors": ["Kulisz, Monika"], "editors": [], "eisbn": ["978-83-7947-636-7"], "extracted_text": {"bytes": 482449, "extractor": "pdfplumber", "page_count": 257, "page_failures": [], "sha256": "cb4cdd675df2221667c5cd3e458d60b55fec8da89a5e6f44c48a6ac44f644d28"}, "handle": "20.500.14629/18535", "id": "rock_pollub_pl_a6c38c5d-ba8f-47ac-a3e6-b50df14a595b", "isbn": ["978-83-7947-635-0"], "issued": ["2025"], "landing_url": "https://rock.pollub.pl/entities/publication/a6c38c5d-ba8f-47ac-a3e6-b50df14a595b", "license": "CC-BY-SA-4.0", "license_evidence": {"item": ["CC-BY-SA-4.0"], "pdf_bitstream": ["CC-BY-SA 4.0"]}, "license_policy_url": "https://wpl.pollub.pl/pl/i/Polityka-publikacyjna/20", "license_url": "https://creativecommons.org/licenses/by-sa/4.0/", "pdf": {"bytes": 24706728, "checksum": {"checkSumAlgorithm": "MD5", "value": "553e78c2bdb6dae7c5967edf1dc80626"}, "content_url": "https://rock.pollub.pl/server/api/core/bitstreams/5710059d-e91e-48cd-b9f7-35a335e6f38e/content", "id": "5710059d-e91e-48cd-b9f7-35a335e6f38e", "name": "Dobor_architektury_9788379476367.pdf", "rights": ["CC-BY-SA 4.0"], "sha256": "845881788fccb341b0c77a5c51afc64da148bccfb8a3f5da0bc42494c799d0e2"}, "publisher": "Politechnika Lubelska", "subtype": ["Monograph"], "text_sha256": "eac5c74faf081d523cd7c2f3abcf3842629c770d6437aafffe099c8a52f92910", "title": "Dobór architektury i parametrów procesu uczenia sieci neuronowych w doskonaleniu procesów produkcyjnych", "transformations": ["pdfplumber page extraction", "Unicode/whitespace normalization", "page-number-only removal", "line-wrap repair", "limited email/labelled-phone redaction"]}
-{"author": "Weinper, Katarzyna", "authors": [], "editors": ["Weinper, Katarzyna"], "eisbn": ["978-83-7947-652-7"], "extracted_text": {"bytes": 749850, "extractor": "pdfplumber", "page_count": 324, "page_failures": [], "sha256": "ea82e63ee099d5c4f3fee69434d835da9b3ee5cb0cb50eda15b06309dfa9131a"}, "handle": "20.500.14629/18743", "id": "rock_pollub_pl_903f22e1-a2ab-4231-a402-2bc9535c12f2", "isbn": ["978-83-7947-651-0"], "issued": ["2025"], "landing_url": "https://rock.pollub.pl/entities/publication/903f22e1-a2ab-4231-a402-2bc9535c12f2", "license": "CC-BY-SA-4.0", "license_evidence": {"item": ["CC-BY-SA-4.0"], "pdf_bitstream": ["CC-BY-SA 4.0"]}, "license_policy_url": "https://wpl.pollub.pl/pl/i/Polityka-publikacyjna/20", "license_url": "https://creativecommons.org/licenses/by-sa/4.0/", "pdf": {"bytes": 11242006, "checksum": {"checkSumAlgorithm": "MD5", "value": "df8dc6047586f20b0a7e5aef3f2ef09a"}, "content_url": "https://rock.pollub.pl/server/api/core/bitstreams/028ebc18-0b71-4cfd-8799-58ca0c40b941/content", "id": "028ebc18-0b71-4cfd-8799-58ca0c40b941", "name": "Wyzwania_Wspolczesnego_9788379476527.pdf", "rights": ["CC-BY-SA 4.0"], "sha256": "899458fc20d68bb3e146cc9d6eef32229c17b1cde5e4629e90a1760bf949e412"}, "publisher": "Politechnika Lubelska", "subtype": ["Monograph"], "text_sha256": "28a7eb8b790116c641427e8ad681942cc058b9f0fac83658e720b875eca33983", "title": "Wyzwania współczesnego publikowania", "transformations": ["pdfplumber page extraction", "Unicode/whitespace normalization", "page-number-only removal", "line-wrap repair", "limited email/labelled-phone redaction"]}
-{"author": "Wlaź, Paweł", "authors": ["Wlaź, Paweł"], "editors": [], "eisbn": ["978-83-7947-640-4"], "extracted_text": {"bytes": 179984, "extractor": "pdfplumber", "page_count": 124, "page_failures": [], "sha256": "bf16f790920e72199c9eddbe70922d941e89b8defd35e6c35e5401c304e5b9db"}, "handle": "20.500.14629/18557", "id": "rock_pollub_pl_0e26bff5-e51a-4d36-9524-fdef6e3e6ab7", "isbn": ["978-83-7947-639-8"], "issued": ["2025"], "landing_url": "https://rock.pollub.pl/entities/publication/0e26bff5-e51a-4d36-9524-fdef6e3e6ab7", "license": "CC-BY-SA-4.0", "license_evidence": {"item": ["CC-BY-SA-4.0"], "pdf_bitstream": ["CC-BY-SA 4.0"]}, "license_policy_url": "https://wpl.pollub.pl/pl/i/Polityka-publikacyjna/20", "license_url": "https://creativecommons.org/licenses/by-sa/4.0/", "pdf": {"bytes": 3165127, "checksum": {"checkSumAlgorithm": "MD5", "value": "afb17dc3ebbcbed87c25aebc7c007ff7"}, "content_url": "https://rock.pollub.pl/server/api/core/bitstreams/a8f29876-c98e-4feb-be00-cd21541930ca/content", "id": "a8f29876-c98e-4feb-be00-cd21541930ca", "name": "Symulacje_MC_9788379476404.pdf", "rights": ["CC-BY-SA 4.0"], "sha256": "64d4df1dc99517c83a4a6d99ec1d613d8debc9a17e08e944fb564c26e172e454"}, "publisher": "Politechnika Lubelska", "subtype": ["Handbook"], "text_sha256": "aee9a79540fd25dd17e81c4f9eda021b45630fb722c3c1a293582384d8c72db5", "title": "Symulacje Monte Carlo. Teoria i zastosowania", "transformations": ["pdfplumber page extraction", "Unicode/whitespace normalization", "page-number-only removal", "line-wrap repair", "limited email/labelled-phone redaction"]}
-{"author": "Kłosowski, Grzegorz", "authors": ["Kłosowski, Grzegorz"], "editors": [], "eisbn": ["978-83-7947-628-2"], "extracted_text": {"bytes": 420051, "extractor": "pdfplumber", "page_count": 223, "page_failures": [], "sha256": "6c036af4551875a4db4a2b97d765ca2279e6cf55654f5822d79d13656fa4b921"}, "handle": "20.500.14629/18428", "id": "rock_pollub_pl_032a14e5-5c8e-4153-9cf7-2dfd9e361e6d", "isbn": ["978-83-7947-627-5"], "issued": ["2025"], "landing_url": "https://rock.pollub.pl/entities/publication/032a14e5-5c8e-4153-9cf7-2dfd9e361e6d", "license": "CC-BY-SA-4.0", "license_evidence": {"item": ["CC-BY-SA-4.0"], "pdf_bitstream": ["CC-BY-SA 4.0"]}, "license_policy_url": "https://wpl.pollub.pl/pl/i/Polityka-publikacyjna/20", "license_url": "https://creativecommons.org/licenses/by-sa/4.0/", "pdf": {"bytes": 44067073, "checksum": {"checkSumAlgorithm": "MD5", "value": "549565d7f323947cc6630a8db3fc97a2"}, "content_url": "https://rock.pollub.pl/server/api/core/bitstreams/99e482b2-ba18-42f0-82aa-4c47a831d550/content", "id": "99e482b2-ba18-42f0-82aa-4c47a831d550", "name": "Zastosowanie_uczenia_9788379476282.pdf", "rights": ["CC-BY-SA 4.0"], "sha256": "7879b3aaf589868bea40ec5af397e0a459785de150bd66ba51b3b63283111bc5"}, "publisher": "Politechnika Lubelska", "subtype": ["Monograph"], "text_sha256": "ed64bc684dd91eb110bcfe32b242560a33310a127c4c6eca060cc53c6adc2060", "title": "Zastosowanie uczenia maszynowego do obrazowania stanu zawilgocenia murów z wykorzystaniem tomografii impedancyjnej", "transformations": ["pdfplumber page extraction", "Unicode/whitespace normalization", "page-number-only removal", "line-wrap repair", "limited email/labelled-phone redaction"]}
-{"author": "Zienkiewicz, Tadeusz; Zienkiewicz, Konrad", "authors": ["Zienkiewicz, Tadeusz", "Zienkiewicz, Konrad"], "editors": ["Zienkiewicz, Tadeusz"], "eisbn": ["978-83-7947-662-6"], "extracted_text": {"bytes": 1022411, "extractor": "pdfplumber", "page_count": 393, "page_failures": [], "sha256": "a9df4a08b74e654091b368e7ba1b9d4732bcaa2448301a2a65105b3655e636ef"}, "handle": "20.500.14629/18744", "id": "rock_pollub_pl_da87b1d3-7018-46fa-81bd-380f0780a1b3", "isbn": [], "issued": ["2025"], "landing_url": "https://rock.pollub.pl/entities/publication/da87b1d3-7018-46fa-81bd-380f0780a1b3", "license": "CC-BY-SA-4.0", "license_evidence": {"item": ["CC-BY-SA-4.0"], "pdf_bitstream": ["CC-BY-SA 4.0"]}, "license_policy_url": "https://wpl.pollub.pl/pl/i/Polityka-publikacyjna/20", "license_url": "https://creativecommons.org/licenses/by-sa/4.0/", "pdf": {"bytes": 3569880, "checksum": {"checkSumAlgorithm": "MD5", "value": "bf6b100ecbc6a12ae37dbfe4be44d09d"}, "content_url": "https://rock.pollub.pl/server/api/core/bitstreams/8086c994-dd9a-449f-aeb9-e4edd06cbd95/content", "id": "8086c994-dd9a-449f-aeb9-e4edd06cbd95", "name": "Zarzadzanie_Wspolpraca_Instytucji__9788379476626.pdf", "rights": ["CC-BY-SA 4.0"], "sha256": "c75ae222fb9362de9ed22e9f45550c3443d0617d31bb0befbee0c891347aa803"}, "publisher": "Politechnika Lubelska", "subtype": ["Monograph"], "text_sha256": "30863bfbfde1222ee44de8c1a8815a72ff9febec8a8b62ef85a6075d78358529", "title": "Zarządzanie współpracą instytucji państwa z diasporą. Przegląd teorii, analiza porównawcza wybranych państw", "transformations": ["pdfplumber page extraction", "Unicode/whitespace normalization", "page-number-only removal", "line-wrap repair", "limited email/labelled-phone redaction"]}
-{"author": "Montusiewicz, Jerzy; Dziedzic, Krzysztof; Barszcz, Marcin; Urzędowski, Arkadiusz", "authors": ["Montusiewicz, Jerzy", "Dziedzic, Krzysztof", "Barszcz, Marcin", "Urzędowski, Arkadiusz"], "editors": [], "eisbn": [], "extracted_text": {"bytes": 182501, "extractor": "pdfplumber", "page_count": 153, "page_failures": [], "sha256": "09794f0d77aa5ec1fa00b027703d4e0d620778b5d53bcac2799da6963cb29b3b"}, "handle": "20.500.14629/211", "id": "rock_pollub_pl_a257b1ce-1af7-4dfc-ae38-ec67ef877689", "isbn": ["978–83–7947–568–1"], "issued": ["2023"], "landing_url": "https://rock.pollub.pl/entities/publication/a257b1ce-1af7-4dfc-ae38-ec67ef877689", "license": "CC-BY-SA-4.0", "license_evidence": {"item": ["CC-BY-SA-4.0"], "pdf_bitstream": ["CC-BY-SA 4.0"]}, "license_policy_url": "https://wpl.pollub.pl/pl/i/Polityka-publikacyjna/20", "license_url": "https://creativecommons.org/licenses/by-sa/4.0/", "pdf": {"bytes": 12577248, "checksum": {"checkSumAlgorithm": "MD5", "value": "55c99ba8c2962c763c847386deabd2ca"}, "content_url": "https://rock.pollub.pl/server/api/core/bitstreams/75143e05-fdbc-4af3-ad39-07ad2cfa0c15/content", "id": "75143e05-fdbc-4af3-ad39-07ad2cfa0c15", "name": "AutoCAD2023.pdf", "rights": ["CC-BY-SA 4.0"], "sha256": "e6831f3e9c05555815d61e97a146b878fc49bf992374411fda4ded824f31d8ee"}, "publisher": "Politechnika Lubelska", "subtype": ["Handbook"], "text_sha256": "bebc8b603695bdbfebc2a136c26f4f1aab5428cee5243fe6781102d2c3821514", "title": "Komputerowa grafika inżynierska. Ćwiczenia do programu AutoCAD 2023", "transformations": ["pdfplumber page extraction", "Unicode/whitespace normalization", "page-number-only removal", "line-wrap repair", "limited email/labelled-phone redaction"]}
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f402aa851f338dd7661310e886a140f551ce58d8cd42a04034104204c7b9ea9
+size 13279

data/rock_pollub_pl/rock_pollub_pl.decisions.jsonl

@@ -1,8 +1,3 @@
-{"characters": 291633, "id": "rock_pollub_pl_513c690d-72b3-4604-b62e-e21e44b91645", "language": "pl", "language_confidence": 1.0, "letter_ratio": 0.7548699907074988, "reason": "include", "replacement_characters": 0, "selected": true}
-{"characters": 247687, "id": "rock_pollub_pl_e9e7a67e-c405-4968-b87a-0ff169c7b3d1", "language": "pl", "language_confidence": 1.0, "letter_ratio": 0.6617666651863037, "reason": "include", "replacement_characters": 0, "selected": true}
-{"characters": 452020, "id": "rock_pollub_pl_a6c38c5d-ba8f-47ac-a3e6-b50df14a595b", "language": "pl", "language_confidence": 1.0, "letter_ratio": 0.7953696739082341, "reason": "include", "replacement_characters": 1, "selected": true}
-{"characters": 716909, "id": "rock_pollub_pl_903f22e1-a2ab-4231-a402-2bc9535c12f2", "language": "pl", "language_confidence": 1.0, "letter_ratio": 0.8020348468215631, "reason": "include", "replacement_characters": 0, "selected": true}
-{"characters": 152666, "id": "rock_pollub_pl_0e26bff5-e51a-4d36-9524-fdef6e3e6ab7", "language": "pl", "language_confidence": 1.0, "letter_ratio": 0.7383372853156565, "reason": "include", "replacement_characters": 0, "selected": true}
-{"characters": 391855, "id": "rock_pollub_pl_032a14e5-5c8e-4153-9cf7-2dfd9e361e6d", "language": "pl", "language_confidence": 1.0, "letter_ratio": 0.8003240994755714, "reason": "include", "replacement_characters": 1, "selected": true}
-{"characters": 959348, "id": "rock_pollub_pl_da87b1d3-7018-46fa-81bd-380f0780a1b3", "language": "pl", "language_confidence": 1.0, "letter_ratio": 0.8103201340910702, "reason": "include", "replacement_characters": 0, "selected": true}
-{"characters": 168819, "id": "rock_pollub_pl_a257b1ce-1af7-4dfc-ae38-ec67ef877689", "language": "pl", "language_confidence": 1.0, "letter_ratio": 0.7856935534507372, "reason": "include", "replacement_characters": 0, "selected": true}
+version https://git-lfs.github.com/spec/v1
+oid sha256:e60fe6b5610cf9453b432ffb4e5cb3630bf3b313a15e5042ea0a176feb1ea6d7
+size 1880

src/render_sample.py

@@ -0,0 +1,66 @@
+"""Render a few parquet rows as readable markdown for a PR description.
+
+Usage: python3 src/render_sample.py data/<key>/<key>.parquet [-n 3] [--max-chars 800]
+
+Each document becomes a fenced block: metadata as `key: value` lines, a blank
+line, then the text with real newlines, truncated with `[…]`.
+"""
+import argparse
+import random
+import re
+
+import pyarrow.parquet as pq
+
+
+def sample_rows(path, n, seed=0):
+    """n rows from distinct random row groups (seeded), so a 2B-token shard isn't read whole."""
+    pf = pq.ParquetFile(path)
+    if pf.num_row_groups == 0 or n <= 0:
+        return []
+    rng = random.Random(seed)
+    groups = rng.sample(range(pf.num_row_groups), min(n, pf.num_row_groups))
+    per_group = -(-n // len(groups))  # ceil: few row groups -> several rows each
+    rows = []
+    for g in sorted(groups):
+        batch = pf.read_row_group(g).to_pylist()
+        rows += rng.sample(batch, min(per_group, len(batch)))
+    return rows[:n]
+
+
+def truncate(text, max_chars):
+    text = text.strip()
+    if len(text) <= max_chars:
+        return text
+    cut = text[:max_chars]
+    space = max(cut.rfind(" "), cut.rfind("\n"))
+    return (cut[:space] if space > 0 else cut).rstrip() + " […]"
+
+
+def fmt_value(v):
+    if isinstance(v, (list, tuple)):
+        return ", ".join(fmt_value(x) for x in v)
+    if hasattr(v, "isoformat"):
+        return v.isoformat()
+    return str(v)
+
+
+def render_doc(row, max_chars):
+    meta = [f"{k}: {fmt_value(v)}" for k, v in row.items() if k != "text"]
+    body = "\n".join(meta) + "\n\n" + truncate(row.get("text") or "", max_chars)
+    longest = max((len(m) for m in re.findall(r"`+", body)), default=0)
+    fence = "`" * max(3, longest + 1)
+    return f"{fence}\n{body}\n{fence}"
+
+
+def render(path, n=3, max_chars=800, seed=0):
+    return "\n\n".join(render_doc(r, max_chars) for r in sample_rows(path, n, seed))
+
+
+if __name__ == "__main__":
+    ap = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
+    ap.add_argument("parquet")
+    ap.add_argument("-n", type=int, default=3)
+    ap.add_argument("--max-chars", type=int, default=800)
+    ap.add_argument("--seed", type=int, default=0)
+    a = ap.parse_args()
+    print(render(a.parquet, a.n, a.max_chars, a.seed))

src/test_render_sample.py

@@ -0,0 +1,81 @@
+"""Contract for render_sample: readable, deterministic PR-description samples."""
+import datetime
+import sys
+import tempfile
+import unittest
+from pathlib import Path
+
+import pyarrow as pa
+import pyarrow.parquet as pq
+
+sys.path.insert(0, str(Path(__file__).resolve().parent))
+
+from render_sample import render, render_doc, sample_rows, truncate
+
+
+def write_parquet(path, n, row_group_size):
+    table = pa.table({
+        "id": [f"src_{i}" for i in range(n)],
+        "text": [f"Akapit pierwszy {i}.\n\nAkapit drugi." for i in range(n)],
+        "source": ["src"] * n,
+        "added": [datetime.date(2026, 7, 19)] * n,
+        "created": [["2023-11-15", "2026-09-03"]] * n,
+        "token_count": list(range(n)),
+    })
+    pq.write_table(table, path, row_group_size=row_group_size)
+
+
+class RenderSampleTest(unittest.TestCase):
+    def setUp(self):
+        self.tmp = tempfile.TemporaryDirectory()
+        self.path = Path(self.tmp.name) / "src.parquet"
+
+    def tearDown(self):
+        self.tmp.cleanup()
+
+    def test_block_layout_matches_pr24_format(self):
+        row = {"id": "x", "text": "Line one.\n\nLine two.", "added": datetime.date(2026, 7, 19),
+               "created": ["2023-11-15", "2026-09-03"], "token_count": 5}
+        self.assertEqual(
+            render_doc(row, 800),
+            "```\nid: x\nadded: 2026-07-19\ncreated: 2023-11-15, 2026-09-03\ntoken_count: 5\n\n"
+            "Line one.\n\nLine two.\n```",
+        )
+
+    def test_real_newlines_not_escaped(self):
+        write_parquet(self.path, 5, 2)
+        out = render(self.path, n=1)
+        self.assertIn("\n\nAkapit drugi.", out)
+        self.assertNotIn("\\n", out)
+
+    def test_truncates_on_word_boundary(self):
+        self.assertEqual(truncate("alpha beta gamma", 12), "alpha beta […]")
+        self.assertEqual(truncate("short", 12), "short")
+        self.assertEqual(truncate("x" * 20, 5), "xxxxx […]")
+
+    def test_fence_outgrows_backticks_in_text(self):
+        out = render_doc({"text": "code: ```py\nx```"}, 800)
+        self.assertTrue(out.startswith("````\n") and out.endswith("\n````"))
+
+    def test_deterministic_and_bounded(self):
+        write_parquet(self.path, 50, 10)
+        self.assertEqual(render(self.path, n=3, seed=1), render(self.path, n=3, seed=1))
+        self.assertEqual(len(sample_rows(self.path, 3)), 3)
+        ids = [r["id"] for r in sample_rows(self.path, 3)]
+        self.assertEqual(len(set(ids)), 3)
+
+    def test_fewer_row_groups_than_n(self):
+        write_parquet(self.path, 6, 6)  # one row group
+        self.assertEqual(len(sample_rows(self.path, 4)), 4)
+
+    def test_n_larger_than_table(self):
+        write_parquet(self.path, 2, 1)
+        self.assertEqual(len(sample_rows(self.path, 10)), 2)
+
+    def test_empty_parquet(self):
+        pq.write_table(pa.table({"id": pa.array([], pa.string()), "text": pa.array([], pa.string())}), self.path)
+        self.assertEqual(render(self.path), "")
+
+
+if __name__ == "__main__":
+    unittest.main()