2 Commits

Author	SHA1	Message	Date
Elie Habib	60e727679c	feat(supply-chain): Sprint E — scenario visual completion + service parity (#2910) ... * feat(supply-chain): Sprint E — scenario visual completion + service parity - E1: fetchSectorDependency exported from supply-chain service index - E2: PRO gate + all-renderer dispatch in MapContainer.activateScenario - E3: scenario summary banner in SupplyChainPanel (dismiss wired) - E4: "Simulate Closure" trigger button in expanded chokepoint cards - E5: affectedIso2s heat layer in DeckGLMap (GeoJsonLayer, red tint) - E6: SVG renderer setScenarioState (best-effort iso2 fill) - E7: Globe renderer scenario polygons via flushPolygons - E8: integration tests for scenario run/status endpoints * fix(supply-chain): address PR #2910 review findings (P1 + P2 + P3) - Wire setOnScenarioActivate + setOnDismissScenario in panel-layout.ts (todo #155) - Rename shadow variable t→tmpl in SCENARIO_TEMPLATES.find (todo #152) - Add statusResp.ok guard in scenario polling loop (todo #153) - Replace status.result! non-null assertion with shape guard (todo #154) - Add AbortController to prevent concurrent polling races (todo #162) - Add polygonStrokeColor scenario branch (transparent) in GlobeMap (todo #156) - Re-export SCENARIO_TEMPLATES via src/config/scenario-templates.ts (todo #157) - Cache affectedIso2Set in DeckGLMap.setScenarioState (todo #158) - Add scenario paths to PREMIUM_RPC_PATHS for auth injection (todo #160) - Show template name in scenario banner instead of raw ID (todo #163) * fix(supply-chain): address PR #2910 review findings - Add auth headers to scenario fetch calls in SupplyChainPanel - Reset button state on scenario dismiss - Poll status immediately on first iteration (no 2s delay) - Pre-compute scenario polygons in GlobeMap.setScenarioState - Use scenarioId for DeckGL updateTriggers precision * fix(supply-chain): wire panel instance to MapContainer, stop button click propagation - Call setSupplyChainPanel() in panel-layout.ts so scenario banner renders - Add stopPropagation() to Simulate Closure button to prevent card collapse	2026-04-10 21:31:26 +04:00
Elie Habib	12ea629a63	feat(supply-chain): Sprint C — scenario engine (templates, job API, Railway worker, map activation) (#2890) ... * feat(supply-chain): Sprint C — scenario engine templates, job API, Railway worker, map activation Adds the async scenario engine for supply chain disruption modelling: - src/config/scenario-templates.ts: 6 pre-built ScenarioTemplate definitions (Taiwan Strait closure, Suez+BaB simultaneous, Panama drought, Hormuz blockade, Russia Baltic grain suspension, US electronics tariff shock) with costShockMultiplier and optional HS2 sector scoping. Exports ScenarioVisualState + ScenarioResult types (no UI imports, avoids MapContainer <-> DeckGLMap circular dep). - api/scenario/v1/run.ts: PRO-gated edge function — validates scenarioId against template registry and iso2 format, enqueues job to Redis scenario-queue:pending via RPUSH. Returns {jobId, status:'pending'} HTTP 202. - api/scenario/v1/status.ts: Edge function — validates jobId via regex to prevent Redis key injection, reads scenario-result:{jobId}. Returns {status:'pending'} when unprocessed, or full worker result when done. - scripts/scenario-worker.mjs: Always-on Railway worker using BLMOVE LEFT RIGHT for atomic FIFO dequeue+claim. Idempotency check before compute. Writes result with 24h TTL; writes {status:'failed'} on error; always cleans processing list in finally. - DeckGLMap.ts: scenarioState field + setScenarioState(). createTradeRoutesLayer() overrides arc color to orange for segments whose route waypoints intersect scenario disruptedChokepointIds. Null state restores normal colors. - MapContainer.ts: activateScenario(id, result) and deactivateScenario() broadcast ScenarioVisualState to DeckGLMap. Globe/SVG deferred to Sprint D (best-effort). 🤖 Generated with Claude Sonnet 4.6 via Claude Code (https://claude.com/claude-code) + Compound Engineering v2.49.0 Co-Authored-By: Claude Sonnet 4.6 (200K context) <noreply@anthropic.com> * fix(supply-chain): move scenario-templates to server/ to satisfy arch boundary api/ edge functions may not import from src/ app code. Move the authoritative scenario-templates.ts to server/worldmonitor/supply-chain/v1/ and replace src/config/scenario-templates.ts with a type-only re-export for src/ consumers. * fix(supply-chain): guard scenario-worker runWorker() behind isMain check Without the isMain guard, the pre-push hook picks up scenario-worker.mjs as a seed test candidate (non-matching lines pass through sed unchanged) and starts the long-running worker process, causing push failures. * fix(pre-push): filter non-matching lines from seed test selector The sed transform passes non-matching lines (e.g. scenario-worker.mjs) through unchanged. Adding grep "^tests/" ensures only successfully transformed test paths are passed to the test runner. * fix(supply-chain): address PR #2890 review findings — worker data shapes + status PRO gate Three bugs found in PR #2890 code review: 1. [High] scenario-worker.mjs read wrong cache shape for exposure data. supply-chain:exposure:{iso2}:{hs2}:v1 caches GetCountryChokepointIndexResponse ({ iso2, hs2, exposures: [{chokepointId, exposureScore}], ... }), not a chokepointId-keyed object. Worker now iterates data.exposures[], filters by template.affectedChokepointIds, and ranks by exposureScore (importValue does not exist on ChokepointExposureEntry). adjustedImpact = exposureScore x (disruptionPct/100) x costShockMultiplier. 2. [Medium] api/scenario/v1/status.ts was not PRO-gated, allowing anyone with a valid jobId to retrieve full premium scenario results. Added isCallerPremium() check; returns HTTP 403 for non-PRO callers, matching run.ts behavior. 3. [Low] Worker parsed chokepoint status cache as Array but actual shape is { chokepoints: [], fetchedAt, upstreamUnavailable }. Fixed to access cpData.chokepoints array. * fix(scenario): per-country impactPct + O(1) route lookup in arc layer - impactPct now reflects each country's relative share of the worst-hit country (0-100) instead of the flat template.disruptionPct for all - Pre-build routeId→waypoints Map in createTradeRoutesLayer() so getColor() is O(1) per segment instead of O(n) per frame * fix(scenario): rate limit, pipeline GETs, error sanitization, processing state, orphan drain - Add per-user rate limit (10 jobs/min) + queue depth cap to run.ts - Replace 594 sequential Redis GETs with single Upstash pipeline call - Sanitize worker err.message to 'computation_error' in failed results - Remove dead validateApiKey() calls (isCallerPremium covers this) - Write processing state before computeScenario() starts - Add SIGTERM handler + startup orphan drain to worker loop - Validate dequeued job payload fields before use as Redis key fragments - Fix maxImpact divide-by-zero with Math.max(..., 1) - Hoist routeWaypoints Map to module level in DeckGLMap - Add GET /api/scenario/v1/templates discovery endpoint - Fix template sync comment to reference correct authoritative file * docs(plan): mark Sprint C complete, record deferrals to Sprint D - Sprint status table added: Sprints 0-2 merged, C ready to merge (#2890), A/B/D not started - Sprint C checklist: 4 ACs checked off, panel UI + tariff-shock visual deferred - Sprint D section updated to carry over Sprint C visual deferrals - PR #2890 added to Related PRs --------- Co-authored-by: Claude Sonnet 4.6 (200K context) <noreply@anthropic.com>	2026-04-10 14:44:14 +04:00