Build the application

## Build
Analyze historically whitelisted screening matches against fresh yente FollowTheMoney entity updates. Flag previously cleared entities that have accumulated new risk identifiers requiring re-adjudication.

## User
Senior Sanctions Compliance Officer

## Workflow
1. Ingest historical screening whitelists containing cleared FollowTheMoney entities and original analyst rationales.
2. Query a simulated yente service for the latest entity properties and sanction listing history for each whitelisted target.
3. Compare historical entity profiles against current properties to detect newly added aliases, passports, or risk tags.
4. Generate a re-adjudication queue prioritizing whitelisted matches with significant material changes in their underlying risk profile.

## Interface
Browser-based compliance dashboard showing reviewer adjudication diffs, source lineage provenance trees, and dispute escalation triggers.

## Data
Use open-source yente FollowTheMoney entity format definitions and generate synthetic screening alert files containing contradictory reviewer dispositions.
Starter examples below are synthetic. They are not downloaded records.

## Source
[OpenSanctions data toolkit](https://raw.githubusercontent.com/opensanctions/yente/main/README.md)
Extracts FollowTheMoney entity and match structures defined in the yente open-source data toolkit documentation to model screening hits and watchlist provenance.

## Tools
llamaindex, pydantic, fastapi.
LlamaIndex indexes reviewer adjudication notes, discrepancy criteria, and entity schemas to retrieve relevant precedents during arbitration synthesis.
Fast hosted instruction-tuned LLM accessed via API with at least 8k context window for reasoning over multi-analyst narratives.

## Deliverables
Submit runnable source, pinned dependencies, setup commands, a working interface, example outputs, and a recorded end-to-end demo. Include tests and an action log.

## Acceptance
- Successfully retrieves updated entity properties from a simulated yente service.
- Identifies material profile changes that invalidate previous false positive rationales.
- Outputs a prioritized re-screening queue in JSON format.

## Starter Examples
```json
[
  {
    "input": {
      "whitelist_id": "WL-992",
      "target_id": "ftm-per-123",
      "cleared_date": "2023-01-15",
      "original_rationale": "Different nationality (Spain vs target's unknown).",
      "current_yente_profile": {
        "schema": "Person",
        "properties": {
          "nationality": [
            "Spain",
            "Russia"
          ],
          "sanctions": [
            "Updated 2024-02: Added Spanish dual citizenship"
          ]
        }
      }
    },
    "expected_output": {
      "whitelist_id": "WL-992",
      "target_id": "ftm-per-123",
      "status": "DECAYED_REQUIRES_REVIEW",
      "material_changes_detected": [
        "nationality",
        "sanctions"
      ],
      "reason": "New dual citizenship matches previously cleared discrepancy."
    }
  },
  {
    "input": {
      "whitelist_id": "WL-993",
      "target_id": "ftm-per-456",
      "cleared_date": "2023-05-10",
      "original_rationale": "Distinct date of birth.",
      "current_yente_profile": {
        "schema": "Person",
        "properties": {
          "birthDate": [
            "1980-01-01"
          ]
        }
      }
    },
    "expected_output": {
      "whitelist_id": "WL-993",
      "target_id": "ftm-per-456",
      "status": "VALID",
      "material_changes_detected": [],
      "reason": "No material changes to entity profile since clearance."
    }
  }
]
```

## Safety
Processes mock screening decisions only; no live external sanction list updates or financial blocking actions are executed.
Use mock external writes. Require human review before any real deployment.