Wiki Pages

The compiled wiki in wiki/ contains Markdown pages organized by type.

Page Kinds

KindDirectoryDescription
indexwiki/index.mdHome page with vault overview
sourcewiki/sources/One page per ingested source
modulewiki/code/One page per ingested code source with repo path, imports, exports, symbols, resolved local links, and diagnostics
conceptwiki/concepts/One page per extracted concept
entitywiki/entities/One page per extracted entity
outputwiki/outputs/Persisted query answers and exploration hub pages
outputwiki/context/Markdown companions for saved agent context packs
memory_taskwiki/memory/tasks/Durable agent task ledger pages with decisions, evidence, outcomes, and follow-ups
insightwiki/insights/Human-authored pages that SwarmVault can read but does not rewrite

Candidate pages for concepts and entities live under wiki/candidates/ with status: candidate until they are promoted.

Project rollups live under wiki/projects/, but they link to the canonical pages above rather than duplicating them.

Graph orientation pages live under wiki/graph/. wiki/graph/report.md is the full trust and orientation report, while wiki/graph/share-card.md, wiki/graph/share-card.svg, and wiki/graph/share-kit/ are the compact post-ready text, visual card, and portable HTML-preview bundle generated by swarmvault graph share.

Page Content

Each page includes:

  • Title and description from the extracted analysis
  • Backlinks to related pages
  • Source citations with provenance links
  • Related concepts and entities
  • Lifecycle metadata such as status, created_at, updated_at, compiled_from, and managed_by
  • Schema and project metadata such as schema_hash and project_ids
  • Semantic tags such as project/<id> and category/<slug> when the schema and source analysis support them

Output pages also carry related page, node, and source metadata. SwarmVault uses that metadata to refresh Related Outputs sections on compiled source, concept, entity, and module pages immediately after saved query and exploration writes. Context-pack output pages additionally carry context_pack_id, goal, target, budget_tokens, estimated_tokens, and artifact_path so the markdown page can be traced back to its saved JSON bundle. Task pages carry task_id, task_status, compatibility memory_task_id / memory_status, context_pack_ids, related_page_ids, related_node_ids, related_source_ids, git_refs, and changed_paths so task handoffs can stay linked to context packs and graph evidence.

Module pages also surface code-specific metadata such as the detected language, repoRelativePath, module name, declared namespace/package, resolved local imports as wikilinks when available, and parser diagnostics when the source is malformed or only partially understood. The current parser-backed language set includes JavaScript, JSX, TypeScript, TSX, Bash, Python, Go, Rust, Java, Kotlin, Scala, Dart, Lua, Zig, C#, C, C++, PHP, Ruby, PowerShell, Elixir, OCaml, Objective-C, ReScript, Solidity, HTML, CSS, Vue single-file components, Svelte single-file components, Julia, and Verilog/SystemVerilog. JavaScript and TypeScript capture static imports plus dynamic import() edges. R files are detected and produce explicit parser-asset diagnostics until a safe packaged grammar is available.

Insight pages are different: they stay human-owned. SwarmVault indexes them into search and query context, but compile does not rewrite files under wiki/insights/.

Version Control

Wiki pages are plain Markdown, making them ideal for version control. Track changes to your knowledge base with git:

git add wiki/
git commit -m "Recompile after adding new sources"