Content Pipeline Hub: Multi-Agent Workflows, Fact Verification, and Automated Publishing

The content pipeline is a set of three small tools. Each lives in its own GitHub repository under jakefearsd, and none of them is part of the Wikantik codebase. Wikantik does not call them: they hand work to each other, and to a wiki or a blog, through files and publishing APIs.


1. Quick-Reference: The Three Projects

ProjectRoleStackOutput
AI Discovery (aidiscovery)Plans a topic universe: the topics, relationships and priorities for a wikiJava 21, Spring Boot CLI, LangChain4j; Anthropic Claude or Ollama; Wikidata/Wikipedia grounding~/.aipublisher/universes/<id>.universe.json
AI Publisher (aipublisher)Writes articles through a five-agent pipelineJava 25, Spring Boot 3.5, LangChain4j; Ollama (default) or Anthropic ClaudeJSPWiki-syntax .txt files in ./output
Blogger Agent (blogger-agent)Researches, drafts, illustrates and publishes posts to four Google Blogger blogsAgent Skills (SKILL.md) plus Python scripts over the Blogger API v3Live Blogger posts, plus content_calendar.md and social_drafts.md

2. How the Pieces Connect

3. The AI Publisher Agents

AgentResponsibilityRuns by default?
ResearchGathers facts and sources, builds an outline, grounded with WikidataYes
WriterDrafts the article in JSPWiki syntaxYes
Fact CheckerVerifies claims against the sources; approves, asks for a revision, or rejectsNo (pipeline.skip-fact-check=true)
EditorPolishes prose, integrates wiki links, scores qualityYes
CriticReviews JSPWiki syntax, structure and readabilityNo (pipeline.skip-critique=true)

4. Core Pipeline Guides