Eve library

Eve configuration and secrets assignment · 2026-09-08

Guide

Revision 1

On this page

This receipt records a complete editorial read of the pinned Configuration and secrets unit in V1/docs/eve/operations-observability-and-maintenance.md and its reviewed assignment to the new eve-detail-configuration-boundary slide. The slide separates the committed configuration-name map, resolved runtime bindings, secret values and the evidence needed for two distinct classes of release-affecting change.

Reviewed boundary#

The assignment binds source revision 1ceae52c7896a59ac74a4429d410f3218f0b91c6, exact lines 44–56, the 87-word unit digest, assembled slide digest, citation heading and eve-operations-detail guide membership. The review record retains the exact unit, the four earlier operations assignments and all five meaningful units still open. The learning map explains the missing detail in each open unit.

The unit is integrated on top of published field-guide revision c8b82d6fd87184503a1b0d3dcc40309f632f59cc. Its explicit designEdition: field-guide marker, masthead treatment and preservation tests match that approved Eve presentation system. The new slide is the sole addition to the 578-slide published baseline.

The four-row comparison preserves every source requirement:

  • Configuration names: provider endpoints, model bindings, limits, feature flags, release scopes, tool availability and kill switches belong to the committed canonical name map.
  • Runtime binding: Eve resolves and validates values at startup and redacts them in logs. Those facts do not turn the committed map into a value store or prove a live deployment.
  • Secret values: the platform secret store holds values; prompts, tool descriptions, client bundles and evidence attachments must not.
  • Change control: model and provider settings follow the release-affecting registry path. Prompts, skills, tool descriptions and evaluation dependencies follow prompt and quality change control.

The primary citation is the exact operations section. A secondary citation links the documented Prompt and skill change control procedure so the slide can teach the second change path without inventing it.

This assignment establishes complete teaching of the documented policy. It does not inspect credential values or certify a live environment, secret-store enforcement, startup success, log redaction, rollback, measured quality, reader comprehension or product acceptance.

Coverage result#

The global report moves from 62 reviewed / 103,264 unassigned to 63 reviewed / 103,263 unassigned, with 10,018 structural units unchanged. The source-scoped result is five reviewed / five unassigned with no errors. The strict global completion gate continues to return 1.

Delivery and evidence#

  • The generated portal contains 60 guides, 579 unique slides, 602 guide-slide instances and 348 cited paths.
  • The affected guide contains eight slides and remains completely narrated at 738.978 seconds. The new CPU-rendered track is 97.760 seconds, 782,829 bytes and has SHA-256 964868884f33c94e8594ece0ad1566ec256d86eb48fe6a2624298d256d275db7. The unrestricted integrity audit decodes and verifies all 579 tracks, totaling 14.39 hours.
  • The complete pinned source recollection matches 9,262 authored sources and 96,443 authored meaningful units.
  • The full Python and Node presentation suites pass 300 and 281 tests, respectively. Focused assignment, source, design-preservation, rollback, PDF, builder, coverage, formatting and syntax checks also pass.
  • The standalone guide suite passes 12 checks across all eight slides at desktop, compact and mobile sizes. It covers source/dialog behavior, accessibility, the eight-page PDF and actual narration decode, pause, resume, seek, speed and end transitions.
  • The center smoke suite passes 12 integrated checks across all 60 guides, including search, history, desktop and mobile navigation, drawer focus, audio interruption and runtime stability.
  • The task-specific browser proof opens the exact configuration slide at 1280×720 and 390×844. It verifies all four rows, the complete Explain text, both source excerpts, accessibility, horizontal fit and actual MP3 playback at ready state 4.
  • The PDF review and visual review retain the inspected page and desktop/mobile captures. The delivered tagged PDF contains eight 960×540-point pages and passes the repository checker.

Findings and retests#

The initial compact static layout measured 762.6875 pixels within a 720-pixel viewport. Removing the optional masthead brought it within bounds. Active playback then measured 723.9375 pixels; shortening the composition caption without removing source meaning cleared that final overflow. The complete standalone suite passed after both changes.

The first PDF check exposed a stale seven-page export. The dedicated browser exporter regenerated the final eight-page artifact, after which metadata, raster, text and repository PDF checks passed.

Two full narration attempts returned 143 after their outer supervisor received SIGTERM. The first retains resource polls without a completion result; the second logged successful verification of all 579 tracks before termination. Both attempt logs and exit records are retained. The same checker was rerun directly with manual resource polling every 30 seconds; it completed with exit 0. Available memory remained well above the repository's stop threshold.

The first final design-regression command omitted the repository test-module path, and the first focused narration command passed two IDs as separate CLI arguments instead of one comma-delimited value. Their exit-1 and exit-2 records are retained as command-correction evidence. The corrected design and focused narration commands both pass.

After origin/main advanced, the first merged Python run exposed 111 stale rollback-baseline comparisons against the pre-redesign slides. The guard now uses the published field-guide revision and permits only this new slide. The next broad run found the old 204-slide count; after it moved to 205, the focused completion test exposed the missing explicit edition marker. Adding that marker changed the assembled slide fingerprint, so the assignment, generated outputs, PDF and browser evidence were all rebound and rerun. The final 300-test suite passes. The post-rebase coverage check also rejected its stale generated teaching report; regenerating that report restored the exact 63 / 10,018 / 103,263 freshness result.

Two further post-rebase commands used superseded option combinations: the narration renderer paired --batch-size with --ids, and the PDF checker omitted the guide coverage file. Both commands failed before accepting evidence, their logs remain in this receipt, and the supported invocations pass.

Normal-speed listening was attempted with three review segments, but this environment's audio tool explicitly reported that audio input is unsupported. The MP3 is covered by full decode/integrity checks and real browser playback, but normal-speed human listening remains open and is not claimed by this receipt.

Reproduce#

Run broad and browser checks serially after the host resource checks required by AGENTS.md:

sh
CHROME_CHANNEL=chromium \
  OSHUN_DECK_QA_DIR=/tmp/eve-operations-configuration-guide \
  python3 tools/presentations/run-supervised.py \
  node tools/presentations/tests/eve-oshun-deck.mjs \
  --guide eve-operations-detail

CHROME_CHANNEL=chromium \
  OSHUN_CENTER_QA_DIR=/tmp/eve-operations-configuration-center \
  python3 tools/presentations/run-supervised.py \
  node tools/presentations/tests/presentation-center.mjs \
  --smoke

CHROME_CHANNEL=chromium \
  node docs/presentations/presentation-center/verification/eve-operations-configuration-assignment-2026-09-08/export-configuration-pdf.mjs

CHROME_CHANNEL=chromium \
  python3 tools/presentations/run-supervised.py \
  node docs/presentations/presentation-center/verification/eve-operations-configuration-assignment-2026-09-08/verify-configuration-browser.mjs

python3 tools/presentations/run-supervised.py \
  /tmp/oshun-presentations-venv.pot9EA/bin/python -m unittest discover \
  -s tools/presentations/tests -p 'test_*.py'
python3 tools/presentations/run-supervised.py \
  node --test tools/presentations/tests/*.test.mjs
/tmp/oshun-presentations-venv.pot9EA/bin/python \
  tools/presentations/inventory-center.py --check
/tmp/oshun-presentations-venv.pot9EA/bin/python \
  tools/presentations/build-eve-oshun.py --check
python3 tools/presentations/check-center-coverage.py --check
python3 tools/presentations/check-center-coverage.py --require-complete  # exits 1
/tmp/oshun-presentations-venv.pot9EA/bin/python \
  tools/presentations/render-eve-oshun-narration.py --check

The receipt self-verifier rechecks the source and unit pins, assignment, slide, citations, guide sequence, generated artifacts, narration, PDF/browser reports, retained evidence hashes, task path boundary, scoped/global counts and live focused gates:

sh
/tmp/oshun-presentations-venv.pot9EA/bin/python \
  docs/presentations/presentation-center/verification/eve-operations-configuration-assignment-2026-09-08/verify-receipt.py

Five meaningful units in this source and 103,263 units globally remain open. This receipt completes one reviewed unit; it does not close the presentation center's exhaustive teaching backlog.