fix(yearly): preserve actual latest weekly summary under word budgets

[jmservera]

Summary

Fix the newly exposed chronological-evidence bug reported in #831 (discussion_r4250685368). This is distinct from merged #833.

• Preserve weekly summaries sorted by source YYYY-WNN headings independently of deduplicated monthly context, append-only backfills, and synthesis fallbacks.
• Reserve the actual final weekly source before selecting complete-sentence context within unchanged budgets; never label an earlier retained source latest.
• Keep reserved endpoint prose independent of optional risk/evidence paragraphs during final budget acceptance.
• Remove unsupported temporal opening/shift/month-closure assertions.
• Cover budget-dropped final summaries across all prose styles and 45/100-word limits, repeated endpoints, empty/default/short/oversized/incomplete sources, synthesis preference, optional prose overhead, and backfilled older reports in both monthly formats.

Validation

• Targeted rollups: 49 passed.
• Full Python suite: 2035 passed, 1 skipped (existing), 2 expected warnings.
• Repository-wide Ruff lint and formatting passed.
• Hugo minified build passed with project-local output.
• In-memory production-source regeneration: 1623 words; June correctly quotes W27 measurable inference/evaluation/security rather than W26.

Source-only PR: no generated pages edited. Owning sync regenerates #831 after independent review/merge. No publishing workflow changes and no merge requested here.

[Copilot]

🟡 Changes recommended

Style-specific paragraph merging can still budget out the reserved latest weekly summary.

1 open finding

• Preserve reserved endpoint despite final paragraph budget · New

What changed in this PR

Fixes yearly narratives so the true latest weekly summary remains distinct from monthly synthesis context.

Changes:

• Preserves ordered weekly summaries separately.
• Reserves the latest weekly summary before context budgeting.
• Removes unsupported temporal assertions and adds regression tests.

File	Description
scripts/​generate_yearly_narrative.py	Separates weekly endpoints and revises chapter prose.
tests/​test_generate_rollups.py	Tests endpoint ordering, budgets, and synthesis fallback.

🧠 Review effort: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

🔵 Needs a closer look

Backfilled older weeks can still be incorrectly labeled as the latest summary.

0 open findings

1 resolved since last review

• Preserve reserved endpoint despite final paragraph budget

Previously missed (1)

In code that hasn't changed since last review

Select latest summary by week, not block order

scripts/​generate_yearly_narrative.py:947

weekly_summaries[-1] assumes the Month Overview blocks are chronological, but merge_sections() preserves existing blocks and only appends missing entries (scripts/generate_rollups.py:496-500). If an older recovered week is added after a newer one, that older week becomes the last block and is again labeled as the “latest” summary. Bind summaries to their week headings and select the maximum week/date (or sort the blocks) before choosing the endpoint; a backfill-order regression would cover this case.

🧠 Review effort: Balanced

[jmservera]

Addressed the valid backfill-order observation from Copilot review 5494339238: weekly endpoints are now bound to their YYYY-WNN headings and sorted by source year/week, independently of append-only block order. Non-weekly context cannot become the endpoint. Added regression coverage for W27 preceding a backfilled W23, in both monthly formats. All 49 targeted tests and 2035 full-suite tests pass (one existing skip). No chronological budget was weakened.

[Copilot]

🟢 Approval recommended

The implementation addresses the reported chronology bug with focused logic and comprehensive regression coverage.

0 open findings

🧠 Review effort: Balanced

[jmservera]

READYFORINDEPENDENTREVIEW — source head 95e54fc.

Latest-head Copilot review 5494402579 completed 2026-10-12T18:05:21Z: approval recommended, zero open findings. The legitimate paragraph-budget thread is fixed and resolved; the additional backfill-order observation is fixed with regression coverage. All 18 checks succeeded; aggregate CI https://github.com/jmservera/SquadScope/actions/runs/38288340961 succeeded, merge state CLEAN, and source branch contains current main.

Validation: 49 targeted rollup tests; 2035 full-suite passes with one existing skip; repository-wide Ruff lint/format; minified Hugo build. In-memory 2026 generation is 1623 words and cites June W27 rather than W26. Worktree clean; scratch build/test outputs removed.

Not merged. No generated content or workflow edits. Owning sync should regenerate #831 after independent review and merge, then resolve its original generated-page finding when the corrected artifact is visible.

[jmservera]

Independent revision by Basher at 9ec3114 addresses Rusty’s rejected-head boundary: chronological extraction now retains an empty slot for a dated week with an empty, whitespace-only, or missing Summary label. A later unusable W27 no longer promotes W26 as latest. Loader-to-chapter regressions cover both monthly formats, chronological and backfill block order, and 45/100-word budgets; direct tuple regressions also cover empty endpoints. The new integration test failed before the fix. Validation: 50 rollup tests passed; full suite 2036 passed/1 pre-existing skipped; repository Ruff lint/format passed. No generated-content changes, merge, or operational-stage work. Awaiting latest-head CI/Copilot before Rusty re-review.

[Copilot]

🟢 Approval recommended

The implementation addresses the chronology bug, preserves budget limits, and includes comprehensive regression coverage.

0 open findings

🧠 Review effort: Balanced

[jmservera]

Ready for Rusty’s independent re-review: latest head 9ec3114. All 18 latest-head checks completed successfully. Copilot finished review of this exact SHA at 2026-10-12T18:37:28Z with approval recommended and 0 open findings; all review threads are resolved. Local rollup suite: 50 passed; full suite: 2036 passed, 1 existing skipped; repository Ruff lint/format clean. Working tree clean. Basher did not consult Frank, spawn agents, merge, regenerate data, or undertake W42 operational work.

[jmservera]

Rusty independent AI re-review verdict: APPROVE exact head 9ec3114. This is an AI-authored approval recommendation, not a human approval or a GitHub APPROVED-state review. GitHub rejected APPROVE because the authenticated account owns this PR; recording the verdict transparently as COMMENTED instead. Read repository instructions, Rusty charter, full replacement diff and loader/chapter context. Independently reproduced the prior 95e54fc W26 promotion when W27 is blank, whitespace-only or missing; 32 loader-to-chapter probes confirm the replacement preserves empty final-week slots and labels valid W27 correctly across both monthly formats, backfill orders and 45/100-word budgets. All 50 rollup tests and targeted Ruff lint/format checks passed locally. Observed all 18 latest-head hosted checks successful, completed exact-head Copilot approval recommendation and zero unresolved review threads. No high-confidence blocking findings. No code edits, delegation or merge performed.