docs/decisions/issue-6507-resilience-indicators.tsv
Rama main · réplica de lectura del repositorio original.
ts phase decision why evidence result
2026-08-30T03:24:41Z frame stopped at the first blocked preflight and repaired the checkout before implementation the repository requires a current branch, live duplicate check, Node 24, and complete dependencies before expensive work agent:preflight issue 6507 ready at 2243035c516742b93eea154c568cd2103a2438ff
2026-08-30T03:28:54Z frame created codex/issue-6507-resilience-indicators from refreshed origin/main no open pull request implements issue 6507 and the delegated task requires an independent branch live issue 6507 and duplicate PR query clean branch at current origin/main
2026-08-30T03:33:42Z design kept indicator contribution arithmetic inside the scorer and the full-imputation factor at the dimension-to-domain boundary weightedBlend determines indicator contributions while coverageWeightedMean applies the 0.5 factor only after the dimension score exists server/worldmonitor/resilience/v1/_dimension-scorers.ts:883 and server/worldmonitor/resilience/v1/_shared.ts:568 design constraint verified
2026-08-30T04:01:00Z design selected request-local scorer tracing with provider-specific raw observation policy literal scorer arithmetic is reproducible while post-hoc projection can reconcile totals without matching the executed formula server/worldmonitor/resilience/v1/_indicator-trace.ts and tests/resilience-indicator-trace.test.mts trace owns normalized scores and contributions; policy owns raw release and provenance
2026-08-30T04:05:00Z plan split implementation into disjoint scorer-trace, source-policy, and API-contract units the seams can be proved independently and avoid concurrent edits to generated artifacts server/worldmonitor/resilience/v1/_indicator-trace.ts server/worldmonitor/resilience/v1/_indicator-source-policy.ts proto/worldmonitor/resilience/v1/get_resilience_indicators.proto parent retained handler integration, generation, commits, PR delivery, review, and CI
2026-08-30T05:28:00Z verification exposed the retired reserve-adequacy dimension as inactive and reconciliation-unavailable instead of inventing a contribution the scorer intentionally returns a neutral display score with zero coverage and zero executed indicator contribution focused RPC test first failed on the retired placeholder contract now reports active=false with a reason and keeps all contributions at zero
2026-08-30T05:36:00Z verification regenerated sebuf clients servers request validation OpenAPI and public MCP facts after the final contract change generated artifacts must match the source proto and registered MCP tool make generate; npm run product:facts:check; tests/resilience-indicators-contract.test.mts focused contract suite and 499-test parity suite passed
2026-08-30T06:14:00Z simplification cached derived trace allocations per dimension and removed avoidable trace-path object copies the private API always emits 72 rows while the public score path must not pay trace-only allocation costs server/worldmonitor/resilience/v1/_indicator-trace.ts:materializeIndicatorTrace and _dimension-scorers.ts:tracedBlend 84 focused tests plus npm run typecheck and npm run typecheck:api passed
2026-08-30T06:29:00Z licensing required a reviewed API or dataset path and marked exact observation provenance separately from registry attribution a provider host or descriptive source link does not prove that the reviewed dataset produced the selected observation tests/resilience-indicator-source-policy.test.mts and ResilienceIndicatorSource.observation_provenance World Bank raw release now requires api.worldbank.org/v2 provenance; documentation and terms URLs fail closed
2026-08-30T06:29:00Z provenance left retrievedAt unavailable on every row the scorer preserves provider observation time or source year but not a per-indicator seed retrieval time server/worldmonitor/resilience/v1/get-resilience-indicators.ts and docs/methodology/resilience-indicator-licensing.mdx response dataVersion is not misreported as per-indicator retrieval time
2026-08-30T08:10:00Z review kept fallback contributions included when a dimension is marked source-failure the scorer publishes the substituted score and the drill-down must reconcile to that exact score tests/resilience-indicator-trace.test.mts source-failure reconciliation case source-failure remains a provenance state without erasing executed contributions
2026-08-30T08:14:00Z review resolved source attribution and observation provenance per exact contributing source aggregate indicator policy metadata can mislabel Eurostat or an unused composite dependency tests/resilience-indicators-rpc.test.mts composite and Eurostat cases missing or unused sources are registry attribution only and Eurostat stays audit-incomplete
2026-08-30T08:18:00Z review used the oldest constituent year for composite observation age a recent source must not hide an older input that contributed to the score tests/resilience-energy-v2.test.mts and tests/resilience-dimension-scorers.test.mts energy and liquid-reserve composite ages are conservative
2026-08-30T08:24:00Z review added a five-minute versioned response cache strict diagnostic seed reads and a fail-closed 60-per-minute route budget a cold build fans out across 62 source keys and Redis errors must not look like valid missing observations tests/resilience-indicators-rpc.test.mts and tests/rate-limit.test.mts repeated builds coalesce while source-read and limiter failures fail closed
2026-08-30T08:31:00Z review corrected energy-v2 provenance and failed closed on low-carbon raw redistribution the fossil-electricity seed is World Bank WDI while the low-carbon OWID chart contains Ember and Energy Institute upstream data scripts/seed-fossil-electricity-share.mjs scripts/seed-low-carbon-generation.mjs and tests/resilience-indicator-source-policy.test.mts imported-fossil provenance now names World Bank and low-carbon raw values remain audit-incomplete
2026-08-30T09:20:00Z review narrowed World Bank raw release to exact indicator paths and required an exact observation-level seed timestamp a provider-wide API root can match unrelated data and an extraction date from another seed is false provenance server/worldmonitor/resilience/v1/_indicator-source-policy.ts and get-resilience-indicators.ts static rows trace country seededAt or _recovered.seededAt; other otherwise licensable rows fail closed until their own seed time is traced
2026-08-30T09:45:00Z review disabled JMESPath for the attribution-bound response and traced only actual WGI contributors with the oldest source year projection can separate a raw value from required attribution and a composite must not hide an older input api/mcp/registry/index.ts scripts/openapi-inject-jmespath.mjs server/gateway.ts and tests/resilience-indicators-rpc.test.mts MCP and OpenAPI discovery omit projection; both runtimes reject it; WGI continuity exposes raw data only with complete exact provenance and conservative age