Module Generators/Pipeline/AnalysisStage

Description

Analysis-first pre-generation pipeline stage.

Executes between the Fetch and Generate stages, consuming already-fetched European Parliament data and running the full suite of political intelligence analysis methods. Produces structured markdown analysis files that article generation strategies then consume to produce higher-quality, deeply-analysed news articles in all 14 languages.

This stage is side-effect-only: it writes analysis markdown and a manifest.json to disk under analysis-output/{date}/. The returned AnalysisContext is informational and currently not consumed by the generate stage; strategies read the analysis output from disk instead. Analysis artifacts are committed to the repository for review and political intelligence improvement.

Analysis methods are grouped into four categories:

Classification (Issues #804): significance, impact-matrix, actor-mapping, forces
Threat Assessment (Issues #805): political-stride, actor-threat, consequence-trees, disruption
Risk Scoring (Issues #806): risk-matrix, capital-risk, quantitative-swot, velocity-risk, agent-workflow
Existing (current codebase): deep-analysis, stakeholder-analysis, coalition-analysis, voting-patterns, cross-session-intelligence

Each method writes a markdown file; failures are isolated so other methods can continue. A AnalysisManifest JSON file is written at the end.

Example

const ctx = await runAnalysisStage(fetchedData, {
  articleTypes: [ArticleCategory.WEEK_AHEAD],
  date: '2026-03-26',
  outputDir: 'analysis-output',
});
console.log(ctx.completedMethods);