Follow information into a wiki, then follow a question back to evidence.
These four flows explain example personal and work systems. New pages, changes to existing pages, human review, retrieval, and answer generation have different responsibilities. Start with a flow, then follow the labeled steps and branches.
1. Turn a source into knowledge
Preserve raw evidence, search related pages, create a seedling or propose changes, and review the result.
Open source → page flow →Personal · query2. Find evidence for a question
Choose exact lookup or hybrid search. Graph expansion runs only when its evidence gate permits it.
Open question → evidence flow →Work · ingestion3. Preserve evidence and project context
Separate raw material, source cards, project context, and review artifacts. Existing-source updates require explicit review.
Open source → project flow →Work · query4. Retrieve work evidence
Inspect the two query entry points, graph reranking, preview-based model answers, and the separate source-support check.
Open question → answer flow →Read each flow by responsibility
Human triggers start capture, approve changes, and ask questions. Agent or tool calls carry out a deliberate action. Automatic processing transforms and ranks data. External systems supply information or model output.
Solid teal means the example implementation was checked in code. Dashed amber means a configured or manual workflow step. These labels describe inspected examples and do not prove that your installation implements them or that the workflow ran successfully.
The work example contains useful gaps: existing-source updates need deliberate version review; short model summaries can lose detail; a ranked hit still needs a source-support check. Treat these as checks to build into your own wiki.
These public derivatives preserve the reviewed nodes and directed connections. Internal source paths, source hashes, private evidence inventories, tool logs, and scan copies are excluded.