Implement full-text analysis, department impact, and debate context for propositions

[Copilot]

propositions.ts was generating articles from get_propositioner metadata alone — no policy substance, no department context, no debate reactions. REQUIRED_TOOLS had been reverted to 1 tool to silence validation failures caused by the missing implementations.

MCP Tool Integrations

• search_dokument_fulltext — fetches full policy text for the top proposition; extracts title/summary for a "Policy Substance" section
• analyze_g0v_by_department — queries a 7-day rolling window of government documents by department; renders a "Department Impact" list
• search_anforanden — retrieves parliamentary speeches mentioning the proposition; renders attributed blockquotes in a "Parliamentary Debate" section

All three calls use non-fatal try/catch — articles generate successfully even when enrichment tools are unavailable.

Article Template Additions (content-generators.ts)

Three new sections appended to generatePropositionsContent(), each with 14-language heading maps and consistent display limits:

const MAX_DISPLAY_ITEMS = 3;
const MAX_SPEECH_PREVIEW_LENGTH = 200;
// sections: .policy-substance, .department-impact, .debate-context

Data Flow

• ArticleContentData extended with fullTextResults, departmentAnalysis, speechDebates
• REQUIRED_TOOLS restored to full 4-tool list: get_propositioner, search_dokument_fulltext, analyze_g0v_by_department, search_anforanden
• generateSources() updated to reflect all 4 tools

Test Coverage

• Mock extended with request and searchSpeeches on MCPClient
• New tests: 4-tool REQUIRED_TOOLS assertion, all-tools-called assertion, graceful degradation when enrichment tools throw

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Improve Propositions: Add full-text analysis, department impact, and debate context</issue_title>
<issue_description>## 📋 Issue Type
Enhancement — Implement missing MCP tools for government propositions: full-text analysis, department impact, and debate context

🎯 Objective

Implement the three TODO MCP tools in scripts/news-types/propositions.ts: search_dokument_fulltext (full policy text analysis), analyze_g0v_by_department (government department impact mapping), and search_anforanden (parliamentary debate context). These tools will transform proposition articles from basic metadata summaries into comprehensive policy analysis pieces.

📊 Current State

• File: scripts/news-types/propositions.ts (401 lines)
• Implemented tools: get_propositioner ✅ (only tool)
• Missing tools: search_dokument_fulltext ❌, analyze_g0v_by_department ❌, search_anforanden ❌
• REQUIRED_TOOLS: Only 1 tool (get_propositioner) — reduced from original 4 to avoid validation failures
• Impact: Proposition articles only show basic metadata (title, date, document ID) without policy substance, department context, or debate reactions

🚀 Desired State

• Full-text analysis extracts key policy provisions, budget allocations, and regulatory changes
• Department impact mapping shows which ministries are affected and their responsibilities
• Debate context includes ministerial presentations and opposition reactions
• Articles explain what propositions actually propose (not just that they exist)
• Budget impact estimates where available
• Timeline for implementation (when laws take effect)
• Cross-reference with related motions and committee assignments

🔧 Implementation Approach

1. Add search_dokument_fulltext integration (~80 lines)

// Extract policy substance from proposition full text
const fullText = await mcpClient.call('search_dokument_fulltext', {
  query: proposition.titel,
  limit: 3
});

• Parse key policy provisions from full text
• Extract budget figures and regulatory changes
• Identify implementation timelines and affected populations
• Use scripts/data-transformers/policy-analysis.ts for domain detection

2. Add analyze_g0v_by_department integration (~60 lines)

// Map proposition to responsible departments
const departmentAnalysis = await mcpClient.call('analyze_g0v_by_department', {
  dateFrom: propositionDate,
  dateTo: propositionDate
});

• Identify lead ministry and supporting departments
• Track ministerial ownership of policy areas
• Add department context to article (who's responsible for implementation)

3. Add search_anforanden integration (~60 lines)

// Find debate speeches about this proposition
const debates = await mcpClient.call('search_anforanden', {
  text: proposition.titel,
  rm: currentRiksmote,
  limit: 10
});

• Extract ministerial presentation quotes
• Capture opposition reaction highlights
• Identify party positions on the proposition

4. Update REQUIRED_TOOLS

Restore full 4-tool list: get_propositioner, search_dokument_fulltext, analyze_g0v_by_department, search_anforanden

5. Enhanced article templates

• Add "Policy Substance" section with key provisions
• Add "Department Impact" sidebar
• Add "Parliamentary Debate" quotes section
• Add "Budget Impact" where applicable
• Update all 14 language templates

🤖 Recommended Agent

code-quality-engineer — TypeScript implementation matching existing MCP integration patterns

✅ Acceptance Criteria

• search_dokument_fulltext integrated for policy substance extraction
• analyze_g0v_by_department integrated for department impact mapping
• search_anforanden integrated for debate context
• REQUIRED_TOOLS restored to full 4-tool specification
• REQUIRED_TOOLS validation passes without warnings
• Article templates include "Policy Substance", "Department Impact", "Debate" sections
• All 14 language templates updated
• No regression in existing proposition generation
• Performance within 20-second batch target for all languages
• Error handling for tools that return empty results (graceful degradation)

📚 References

• File: scripts/news-types/propositions.ts:49-52 (TODO comments)
• File: scripts/news-types/propositions.ts:187-196 (REQUIRED_TOOLS update comment)
• MCP Tools: search_dokument_fulltext, analyze_g0v_by_department, search_anforanden
• Policy analysis: scripts/data-transformers/policy-analysis.ts
• Document analysis: scripts/data-transformers/document-analysis.ts
• Pattern reference: scripts/news-types/weekly-review.ts (full enrichment pipeline example)
  </issue_description>

<agent_instructions>Implement the three TODO MCP tools in scripts/news-types/propositions.ts: search_dokument_fulltext, analyze_g0v_by_department, search_anforanden. Resto...

• Fixes Improve Propositions: Add full-text analysis, department impact, and debate context #600

---

🔒 GitHub Advanced Security automatically protects Copilot coding agent pull requests. You can protect all pull requests by enabling Advanced Security for your repositories. Learn more about Advanced Security.