Workflows

A workflow is a family of adapters that together accomplish one job — scraping a site, converting HTML to segments, embedding text, translating, synthesising knowledge. Configs wire these adapters into routes; this section is the recipe book, one page per domain. Each workflow page links to its exact adapter types in the Adapter catalog.

The canonical chain

Most ingestion configs walk the same spine, picking up where the previous one left off via storage:

Sitemap discovery → URL distribution → content fetching → storage
→ Markdown conversion → segmentation → embedding → vector storage
→ (optional) knowledge synthesis → Avalon

A config rarely runs the whole chain. Operators compose the slice they need — crawl only, crawl + markdown, markdown + embeddings, or the full pipeline — by selecting routes.

The use-cases, by domain

You want to…	Workflow	Engines / adapters
Pull pages off a website	Web ingest	web_scraper (HTTP), web_crawler (browser)
Turn HTML into LLM-ready segments	Markdown & segmentation	html_to_markdown, segmenters
Vectorise content	Embeddings & vectors	embedding_generator
Translate while preserving structure	Translation	translator adapters
Ingest from SharePoint	SharePoint ingest	Graph fetch + converter
Ingest a Confluence space	Confluence ingest	Confluence fetch + converter
Process Boost.AI exports	Boost.AI knowledge	boost catalog sub-pipeline
Strip PII / classify sensitivity	Hygiene	pii_redactor, sensitivity_classifier, sc_redactor
Synthesise a concept map → Avalon	Knowledge & Avalon synthesis	concept detection, knowledge_diff

Workflows never import each other — they hand off through queues and storage. See Adapters & routes for why.