Dynamic Algorithms

src/lib/cases/dynamic-algorithms.ts is the deterministic case-intelligence layer for launch. It turns normalized case signals into explainable scores, ranked risks, optimization metrics, and next-best actions without making external model calls.

Covered Algorithms

• Case health: composes the existing weighted risk-scoring engine into a health/risk outcome.
• Statute risk: uses the sourced SOL catalog to compare stored and derived deadline signals.
• Settlement optimization: models liability-adjusted value, settlement range, target, and demand anchor.
• Lien and disbursement optimization: models fee/cost/lien impact, target reductions, lien savings, and net-to-client.
• Treatment gap detection: detects delayed initial treatment, chronology gaps, and current active-case gaps.
• Deposition contradictions: scores transcript coverage, admissions, contradictions, and follow-up gaps.
• Workload optimization: detects over-capacity staff, under-capacity staff, overdue work, and imbalance.
• Revenue leakage: estimates stale-case, undervalued-settlement, medical-cost, unbilled-work, and receivables leakage.
• Integration reliability: rolls up blocked connectors, attention connectors, backlog, and expiring credentials.

Output Contract

buildDynamicAlgorithmReport(input) returns:

• algorithms: one named outcome per algorithm, each with score, riskScore, level, confidence, summary, and ranked signals.
• metrics: domain-specific numeric output for settlement, liens, treatment, deposition, workload, revenue, and integrations.
• nextBestActions: ranked action queue with category, owner role, due window, evidence, confidence, and optional dollar impact.

Scores are normalized so higher score means healthier/safer and higher riskScore means more urgent.

Product Guidance

• Use this engine as the source of truth for dashboard cards, alerts, command-center panels, and AI review context.
• Keep AI generation separate from this layer. AI may explain or draft from these outputs, but these calculations should remain deterministic and testable.
• Add citations/source spans at the route or AI-envelope layer when actions are shown to attorneys or clients.
• Keep route-specific data fetching outside this file. Pass normalized inputs into the engine so unit tests stay fast.

Verification

Focused coverage lives in src/__tests__/lib/dynamic-algorithms.test.ts.

Run:

npm run test:run -- dynamic-algorithms
npx eslint src/lib/cases/dynamic-algorithms.ts src/__tests__/lib/dynamic-algorithms.test.ts
npm run typecheck