P0-1: Evidence Agent — Document Intelligence

[haJ1t]

P0-1: Evidence Agent — Document Intelligence

AI agent that transforms the Evidence Hub from passive file storage into an active compliance intelligence system. Analyzes uploaded documents, scores evidence quality, suggests control links, and detects evidence gaps.

Dependencies

• Phase 0 — Shared Infrastructure (DB Tables, Types, Agent Base) #3596 (Shared Infrastructure) — DB tables, interfaces, parsers

Tasks

Backend — Agent and AI Logic

• Create Servers/advisor/agents/evidence.agent.ts — Evidence Agent definition (Vercel AI SDK)
• Create Servers/advisor/tools/evidenceAiTools.ts — Tool schemas (analyzeDocument, scoreQuality, matchControls, detectGaps)
• Create Servers/advisor/functions/evidenceAiFunctions.ts — Tool implementations
• Implement content analyzer — generateObject() for summary, key_findings, compliance_areas
• Implement quality scorer — 5-dimension scoring (relevance, completeness, recency, reliability, specificity)
• Implement control matcher — compare extracted content against control requirements, suggest file_entity_links
• Implement gap analyzer — query controls without evidence or with low-quality evidence

Backend — Route / Controller / Utils

• Create Servers/utils/evidenceAi.utils.ts — DB queries (insert/get analysis, quality scores, gaps)
• Create Servers/controllers/evidenceAi.ctrl.ts — Request handling with logProcessing/logSuccess/logFailure
• Create Servers/routes/evidenceAi.route.ts — 6 endpoints:
  • POST /api/evidence-ai/analyze/:fileId — Trigger AI analysis
  • GET /api/evidence-ai/analysis/:fileId — Get analysis results
  • GET /api/evidence-ai/quality-scores — Quality scores dashboard
  • GET /api/evidence-ai/gaps — Evidence gap analysis
  • GET /api/evidence-ai/suggestions/:fileId — Suggested control links
  • POST /api/evidence-ai/suggestions/:fileId/apply — Apply suggestions
• Register routes in Servers/index.ts

Frontend — Components

• Create EvidenceQualityBadge/ — Quality score badge (0-100, color coded)
• Create EvidenceAnalysisPanel/ — Full analysis view (summary, findings, quality breakdown, suggested links)
• Create EvidenceGapChart/ — Gap visualization (controls without/with weak evidence)

Frontend — Hooks and Integration

• Create Clients/src/application/hooks/useEvidenceAi.ts — React Query hooks (useAnalysis, useQualityScores, useGaps, useTriggerAnalysis)
• Create Clients/src/application/repository/evidenceAi.repository.ts — Axios calls
• Integrate quality badge into Evidence Hub file list
• Integrate AI Analyze button into Evidence Hub actions column
• Add evidence coverage indicator to control pages

API Endpoints

Route	Method	Purpose
/api/evidence-ai/analyze/:fileId	POST	Trigger AI analysis
/api/evidence-ai/analysis/:fileId	GET	Get analysis results
/api/evidence-ai/quality-scores	GET	Dashboard quality scores
/api/evidence-ai/gaps	GET	Evidence gap analysis
/api/evidence-ai/suggestions/:fileId	GET	Suggested control links
/api/evidence-ai/suggestions/:fileId/apply	POST	Apply suggestions

Branch

claude/practical-euler

Assignees

@haJ1t @HarshP4585

---

Part of Phase 0 — Immediate Priority AI Features