\\n
\\n {% if a.modified_iso != a.published_iso %}
{% endif %}\\n-
{{ t.sourceTimeLabel }}{{ t.metaColon }}{{ a.meta_ts }}\\n+ {% if a.source_meta_ts %}
{{ t.sourceTimeLabel }}{{ t.metaColon }}{{ a.source_meta_ts }}{% endif %}\\n
{{ t.articleBylineLabel }}{{ t.metaColon }}{{ t.articleByline }}\\n
{{ t.autoCompiled }} · {{ t.mainSource }}: {{ a.source }}\\n
{{ t.refreshNote }}\\n@@ -284,6 +284,18 @@ def test_group_cap_and_recency_order() -> None:\\n # Single-source groups rank purely by newest publishedUtc, so the\\n # 12 newest stories are emitted in recency order.\\n assert [g[\"items\"][0][\"title\"] for g in groups] == titles[:12]\\n+ # Suppression happens before MAX_GROUPS: blocked stories cannot\\n+ # monopolize discovery or be reset by regenerated group IDs.\\n+ ledger = {\"schemaVersion\": 1, \"stories\": {\\n+ str(i): {\"sourceUrls\": [items[i][\"url\"]], \"contentFailures\": 4,\\n+ \"serviceFailures\": 0, \"status\": \"exhausted\",\\n+ \"updatedUtc\": _stamp(hours=0), \"nextRetryUtc\": None}\\n+ for i in range(12)\\n+ }}\\n+ (data_dir / \"news-retry.json\").write_text(json.dumps(ledger), encoding=\"utf-8\")\\n+ _run_dedupe(data_dir)\\n+ remaining = _load_pending(data_dir)[\"groups\"]\\n+ assert [g[\"items\"][0][\"title\"] for g in remaining] == titles[12:]\\n \\n \\n def test_taiwan_airline_story_is_reserved_ahead_of_general_cap() -> None:\\n@@ -38,15 +38,18 @@ def article(id, published, source=None, sort=None):\\n with patch.object(build, \"ARTICLES_DIR\", folder):\\n collected = build.collect_articles()\\n assert [a[\"id\"] for a in collected] == [\\n- \"newer-manual\", \"no-source-time\", \"older-late-generated\",\\n- \"invalid-source-time\", \"date-only-source\",\\n+ \"date-only-source\", \"older-late-generated\", \"newer-manual\",\\n+ \"no-source-time\", \"invalid-source-time\",\\n ]\\n- assert collected[0][\"meta_ts\"] == \"2026-09-30 4:50 PM UTC+8\"\\n- assert collected[-1][\"meta_ts\"] == \"2026-09-26\"\\n+ assert collected[0][\"meta_ts\"] == \"2026-10-01 11:00 AM UTC+8\"\\n+ assert collected[0][\"source_meta_ts\"] == \"2026-09-26\"\\n+ assert collected[-1][\"source_meta_ts\"] is None\\n for prepared in collected:\\n indexed = build.search_index_item(prepared, [])\\n assert build.parse_iso(indexed[\"published\"]) == prepared[\"news_dt\"]\\n- assert collected[0][\"published_iso\"] == \"2026-09-30T08:50:00Z\"\\n+ assert collected[2][\"published_iso\"] == \"2026-09-30T08:50:00Z\"\\n+ revised = rows[0] | {\"updatedUtc\": \"2026-10-02T08:00:00Z\"}\\n+ assert build.prep_article(revised)[\"news_dt\"] == collected[1][\"news_dt\"]\\n \\n \\n if __name__ == \"__main__\":\\n@@ -41,13 +41,14 @@ def main() -> None:\\n assert source[\"type\"] == \"rss\"\\n assert \"news.google.com/rss/search\" in source[\"endpoint\"]\\n \\n- # Source publication time drives reader-facing chronology.\\n+ # First site publication drives reader-facing chronology.\\n late = build.prep_article(_article(\\n published=\"2026-10-01T12:00Z\",\\n source_published=\"2026-09-29T10:00Z\",\\n ))\\n assert late is not None\\n- assert late[\"news_dt\"].isoformat().startswith(\"2026-09-29T10:00\")\\n+ assert late[\"news_dt\"].isoformat().startswith(\"2026-10-01T12:00\")\\n+ assert late[\"source_meta_ts\"].startswith(\"2026-09-29\")\\n assert late[\"published_dt\"].isoformat().startswith(\"2026-10-01T12:00\")\\n assert late[\"late_ingest\"] is True\\n \\n@@ -0,0 +1,66 @@\\n+\"\"\"Persistent story retry budgets without API calls or production writes.\"\"\"\\n+import sys\\n+import tempfile\\n+from datetime import datetime, timedelta, timezone\\n+from pathlib import Path\\n+from unittest.mock import patch\\n+\\n+sys.path.insert(0, str(Path(__file__).resolve().parents[1] / \"pipeline\"))\\n+import news_retry\\n+\\n+\\n+def story(n, extra=False):\\n+ urls = [f\"https://example.com/news/{n}\"]\\n+ if extra:\\n+ urls.append(f\"https://official.example.com/news/{n}\")\\n+ return {\"id\": str(n), \"items\": [{\"url\": u, \"title\": \"Test news\"} for u in urls]}\\n+\\n+\\n+def main():\\n+ now = datetime(2026, 10, 1, tzinfo=timezone.utc)\\n+ with tempfile.TemporaryDirectory() as tmp, patch.object(news_retry, \"DATA_DIR\", Path(tmp)):\\n+ ledger = news_retry.RetryLedger(now)\\n+ group = story(1)\\n+ for failure, hours in enumerate((2, 6, 24), 1):\\n+ ledger.failure(group, content=True)\\n+ ledger.save()\\n+ ledger = news_retry.RetryLedger(now)\\n+ assert ledger.state(group)[1][\"contentFailures\"] == failure\\n+ chosen, deferred = ledger.plan([group | {\"id\": \"regenerated\"}], 10)\\n+ assert not chosen and len(deferred) == 1\\n+ now += timedelta(hours=hours)\\n+ ledger = news_retry.RetryLedger(now)\\n+ assert ledger.plan([group], 10)[0] == [group]\\n+ ledger.failure(group, content=True)\\n+ ledger.save()\\n+ ledger = news_retry.RetryLedger(now + timedelta(days=1))\\n+ assert ledger.exhausted(group)\\n+ assert not ledger.plan([group], 10)[0]\\n+ renewed = story(1, extra=True)\\n+ assert ledger.plan([renewed], 10)[0] == [renewed]\\n+ assert ledger.state(renewed)[1][\"contentFailures\"] == 0\\n+ assert not ledger.exhausted(group)\\n+ ledger.failure(renewed, content=True)\\n+ ledger.complete(renewed)\\n+ assert not ledger.rows\\n+\\n+ for n in range(5):\\n+ ledger.failure(story(n), content=False)\\n+ ledger.now += timedelta(hours=2)\\n+ fresh = [story(n) for n in range(10, 22)]\\n+ chosen, deferred = ledger.plan([story(n) for n in range(5)] + fresh, 10)\\n+ assert chosen[:8] == fresh[:8]\\n+ assert len(chosen) == 10 and len(deferred) == 7\\n+ assert all(r[\"contentFailures\"] == 0 for r in ledger.rows.values())\\n+ for _ in range(5):\\n+ ledger.failure(story(0), content=False)\\n+ assert not ledger.exhausted(story(0))\\n+ assert ledger.state(story(0))[1][\"nextRetryUtc\"] == news_retry.iso_minute(\\n+ ledger.now + timedelta(hours=24))\\n+ ledger.save()\\n+ assert not news_retry.RetryLedger(ledger.now + timedelta(days=31)).rows\\n+ print(\"PASS retry persistence, 2/6/24h backoff, cap, new sources and provider separation\")\\n+\\n+\\n+if __name__ == \"__main__\":\\n+ main()\\n@@ -1145,8 +1145,45 @@ def draft(self, *args):\\n \"timed-out story is not marked consumed\")\\n \\n \\n+def test_persistent_story_retry_budget():\\n+ reset_data_dir()\\n+ pending = load(DATA / \"pending.json\")\\n+ group = pending[\"groups\"][0]\\n+ pending[\"groups\"] = [group]\\n+ common.save_json(DATA / \"pending.json\", pending)\\n+ solo = FakeProvider(\"gemini\", [None])\\n+ original_build, original_now = write.build_providers, write.now_utc\\n+ clock = [NOW]\\n+ write.build_providers = lambda: [solo]\\n+ write.now_utc = lambda: clock[0]\\n+ try:\\n+ for hours in (2, 6, 24):\\n+ write.main()\\n+ attempted = solo.calls\\n+ write.main()\\n+ check(solo.calls == attempted, \"cooling story spends zero model calls\")\\n+ clock[0] += timedelta(hours=hours)\\n+ write.main()\\n+ check(solo.calls == 4, \"initial failure plus three retries exhausts the budget\")\\n+ check(not load(DATA / \"pending.json\")[\"groups\"],\\n+ \"exhausted story leaves pending and remains in failure register\")\\n+ check(next(iter(load(DATA / \"news-retry.json\")[\"stories\"].values()))[\"status\"]\\n+ == \"exhausted\", \"exhaustion persists across runs\")\\n+ pending[\"groups\"] = [group | {\"id\": \"rediscovered\"}]\\n+ common.save_json(DATA / \"pending.json\", pending)\\n+ write.main()\\n+ check(solo.calls == 4, \"rediscovery with a new group ID cannot bypass the cap\")\\n+ group[\"items\"].append(group[\"items\"][0] | {\"url\": \"https://official.example.com/new-source\"})\\n+ common.save_json(DATA / \"pending.json\", pending)\\n+ write.main()\\n+ check(solo.calls == 5, \"a new source reopens the story\")\\n+ finally:\\n+ write.build_providers, write.now_utc = original_build, original_now\\n+\\n+\\n def main():\\n tests = [test_flexible_publication_keeps_evidence_gates,\\n+ test_persistent_story_retry_budget,\\n test_story_budget_allows_later_groups,\\n test_budget_saves_completed_articles_and_retains_unattempted_groups,\\n test_publish_flow,"
}