Release archive
3.2.27

Trust Score v2.0 — Canonical API Endpoint & agent_tools.py weights

Release v3.2.27 focuses on Trust Score v2.0 — Canonical API Endpoint, agent_tools.py weights, and TrustScore.jsx.

AddedFixedInternal
Trust Score v2.0 — Canonical API Endpoint & agent_tools.py weights

Release digest

Added

  • Trust Score v2.0 — Canonical API Endpoint — Sprint 8 Priority 1 complete
    • New: GET /api/corporate/trust-score-details/{company_name}
    • Returns: 6-factor breakdown, per-source data, confidence level (low/medium/high), data freshness timestamp, and source_weights transparency object
    • Source quality weighting: Kununu 50% / Google Reviews 30% / Glassdoor 20%

Fixed

  • agent_tools.py weights — Corrected 6-factor breakdown to match official formula: 25% Rating Quality + 25% Sentiment + 15% Volume + 15% Source Diversity + 10% Consistency + 10% Recency (was incorrectly: 30/25/20 — missing Source Diversity pillar)

Internal

  • TrustScore.jsx — Added sourceWeights prop; each data source card now displays its quality weight as a cyan badge (Kununu 50% / Google 30% / Glassdoor 20%) for full algorithmic transparency
  • knowledge-base/docs/agents/ai-agents-overview.md — New multi-agent system overview (all 4 ADK agents: orchestrator, corporate, crypto, osint)
  • docs/infrastructure/GENAI_CREDITS_PLAN.md — Root cause analysis + OSINT migration plan
  • PRODUCT_ROADMAP.md — Sentinel Swarm table updated with live agent status + Sprint 9-10 features
  • .agent/SPRINT_PLAN.md — New Icebox items: KYB Agent, Self-Healing RAG, Rescue Payload

Technical notes