Epic: Cross-KB investigation search and entity correlationbacklog_item

cross-kbjournalismsearchinvestigationepic
1 min read · Edit on Pyrite

Overview

Investigative journalists work across multiple KBs simultaneously — their current investigation, prior investigations, a shared reference KB of known entities, and external MCP data sources. The research phase involves searching everywhere at once and correlating results: "Does this person appear in any of our other investigations? Is this company in the Panama Papers?"

This epic makes cross-KB search a first-class operation, with entity deduplication and a "known entities" reference KB pattern.

User Workflow

``` Journalist asks: "What do we know about Company X?" ↓ ┌─────────────────┼──────────────────┐ ↓ ↓ ↓ Current KB Prior KBs External MCP (investigation) (old investigations) (Panama Papers, documents, web) ↓ ↓ ↓ └─────────────────┼──────────────────┘ ↓ Correlated results: - Entity profile from current KB - 3 mentions in prior investigations - 2 Panama Papers matches - 5 corporate registry hits ```

Subtasks

1. Unified cross-KB search — single query across all configured KBs with result correlation 2. Cross-KB entity deduplication — detect same entity across KBs, suggest merges 3. Known entities KB pattern — shared reference KB of established entities reusable across investigations

Success Criteria

  • Single search query returns correlated results from all KBs
  • Same entity in multiple KBs is identified and linked
  • Known entities KB provides shared baseline for all investigations
  • External MCP sources integrated into search results