Build the complete application

## User
A stock administration team coordinating appraiser and counsel review.

## Build
Compare an appraisal information cutoff with later financing, contract, and operating events. Prepare a review queue without automatically declaring an appraisal valid or invalid.

## Input Contract
- Key: event_id + evidence_version + review_cutoff.
- Fields: occurred_at, known_at, document_date, proposed_materiality, reviewer_decision, appraisal_asof.
- Separate when an event happened from when it became known.

## Workflow
1. Construct an evidence-linked event timeline.
2. Identify information absent from the appraisal cutoff.
3. Route uncertain or potentially material changes to qualified reviewers.
4. Export open questions and a dated review packet.

## Interface
An appraisal timeline with event evidence, disputed dates, and appraiser/counsel review assignments.

## Deliverables
- An event-to-appraisal comparison service.
- A review packet that never substitutes an age threshold for legal analysis.
- Runnable local application, setup commands, and a dependency lockfile.
- Versioned input data, data dictionary, acquisition instructions, and rights notes.
- Calculation traces, source references, assumption records, and exported review artifacts.
- A deterministic baseline and tests for normal, incomplete, contradictory, and failure-recovery cases.
- A recorded complete workflow with a reproducible manifest and explicit limitations.

## Acceptance Criteria
- A known potentially material event creates a review item.
- Unknown materiality remains unresolved.
- Do not treat an appraisal under a stated age as automatically valid.
- Later knowledge is not attributed to an earlier decision without evidence.

## Contrast Case
Keep appraisal age unchanged and add a potentially material event. The review routing changes, not an automatic legal verdict.

## Dataset Scope
The attached starter files are prototype inputs. Expected starter answers are visible. They are not independent holdouts.

Full target: Author fictional appraisal dates, financing notices, customer losses, and event evidence. Include conflicting effective dates and uncertain materiality.

Complete target contracts and expert review remain pending. Public observations are not historical release vintages.

## Tools
DuckDB, Pydantic, Streamlit

Use an available model appropriate to the task. Record provider, model identifier, version or retrieval date, settings, cost, and latency. Use deterministic code for calculations. Models may extract, retrieve, or explain evidence. They must not invent missing records, override approval rules, or claim professional authority.

## Constraints
No output determines safe-harbor eligibility, grant validity, or tax consequences.

- No fabricated records, source citations, professional endorsements, or expert-adjudication claims.
- No future-data leakage into dated decisions, model fitting, or historical valuation inputs.
- No unauthorized payments, account restrictions, regulatory filings, or production changes.
- No automated appraisal certification, tax-compliance determination, or unsupported fraud accusation.
- No concealed uncertainty or silently invented inputs required for a numerical conclusion.

## Scoring
- Correct domain calculations and justified methods with documented assumptions. Weight: 30.
- Source lineage, temporal integrity, rights checks, and explicit synthetic-data labels. Weight: 25.
- A working application with review, export, error handling, and reproducible state. Weight: 20.
- Adversarial cases, uncertainty, missing-data behavior, and sensitivity analysis. Weight: 15.
- Evidence-linked conclusions that distinguish facts, assumptions, and unresolved judgments. Weight: 10.
Score each dimension from zero to two. Weighted score = sum(weight * score / 2). A failed safety gate blocks acceptance.