Fused risk assessment: INSUFFICIENT EVIDENCENot submitted
Live case data
CASE WORKSPACE

System Architecture

Browser → application API → investigation orchestrator → media processing → analysis adapters → evidence normalizer → fusion → reasoner → assessment → mitigation → reporting. Select any component to inspect it.

Controlled demonstration

Processing fabric

Component detail

EVIDENCE FUSION

Weighted, explainable combination — not an average. Independent corroboration raises weight; unavailable modules lower confidence rather than the score.

Backend functions
runEvidenceFusion()

Deployment model

  • frontend — React + TypeScript, server-rendered
  • api — typed server functions, validated inputs, structured errors
  • database — PostgreSQL with row-level security
  • storage — private object storage, signed short-lived reads
  • workers — containerised Python / FFmpeg media services
  • reasoning — server-side LLM adapter, keys held server-side only

This build runs the deterministic demonstration path in-process so the full workflow can be evaluated offline. The adapter boundaries above are where real services attach.

Data model

users · investigations · media_assets · media_segments · analysis_runs · analysis_modules · analysis_results · evidence · evidence_links · reasoning_results · assessments · review_actions · reports · audit_events · model_registry

UUID primary keys, created_at/updated_at on every table, foreign keys throughout, RLS enabled and private storage policies.