Project Structure
After running swarmvault init, your workspace looks like this:
my-vault/
|-- swarmvault.config.json
|-- swarmvault.schema.md
|-- inbox/
|-- raw/
| |-- sources/
| `-- assets/
|-- wiki/
| |-- index.md
| |-- log.md
| |-- candidates/
| |-- graph/
| |-- insights/
| |-- context/
| |-- exports/
| |-- memory/
| |-- sources/
| |-- code/
| |-- concepts/
| |-- entities/
| |-- projects/
| `-- outputs/
| |-- assets/
| |-- chat-sessions/
| `-- index.md
|-- state/
| |-- manifests/
| |-- extracts/
| |-- analyses/
| |-- code-index.json
| |-- embeddings.json
| |-- graph.json
| |-- context-packs/
| |-- chat-sessions/
| |-- memory/
| |-- retrieval/
| |-- compile-state.json
| |-- sessions/
| |-- approvals/
| |-- schedules/
| `-- jobs.ndjson
|-- .obsidian/
`-- agent/Key Directories
- `swarmvault.schema.md` - vault-specific markdown instructions used by compile and query
- `inbox/` - staging area for browser clips, imported markdown and HTML bundles, and other capture workflows
- `raw/sources/` - immutable canonical copies of ingested source files
- `raw/assets/` - copied sidecar assets referenced by inbox markdown bundles and localized remote URL images
- `wiki/` - generated markdown pages for sources, code modules, concepts, entities, indexes, saved answers, and exploration hubs
- `wiki/log.md` - append-only activity log for ingest, compile, query, lint, review, and candidate operations
- `wiki/candidates/` - staged concept and entity pages awaiting confirmation on a later compile
- `wiki/insights/` - human-authored pages that SwarmVault can read during compile and query but does not rewrite
- `wiki/graph/` - graph orientation pages, including
report.md,share-card.md,share-card.svg,share-kit/,report.json, and per-community summaries - `wiki/context/` - markdown companions for saved context packs
- `wiki/exports/ai/` - default static AI export pack with
llms.txt, full text, JSON-LD graph data, manifest metadata, and optional page siblings - `wiki/memory/` - task ledger index and markdown task pages
- `wiki/code/` - module pages generated from ingested 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, Svelte, Julia, Verilog/SystemVerilog, R, and SQL sources, with explicit diagnostics when a detected language lacks a safe packaged parser
- `wiki/projects/` - project rollups over canonical pages when
projectsare configured - `wiki/outputs/assets/` - local chart/image artifacts plus JSON manifests for saved visual outputs
- `wiki/outputs/chat-sessions/` - markdown transcripts for persisted
swarmvault chatsessions - `state/manifests/` - source metadata, content hashes, storage locations, and attachment references
- `state/extracts/` - extracted markdown plus JSON sidecars with extractor metadata, warnings, PDF page counts, Word family (
.docx/.docm/.dotx/.dotm) metadata, RTF paragraph counts, OpenDocument metadata, EPUB chapter details, CSV/TSV and Excel family (.xlsx/.xlsm/.xlsb/.xls/.xltx/.xltm) tabular previews, PowerPoint family (.pptx/.pptm/.potx/.potm) slide metadata, Jupyter notebook cell/output counts, BibTeX/Org-mode/AsciiDoc structure, structured config/data schema hints, and image-vision details - `state/analyses/` - provider-generated analysis output
- `state/code-index.json` - repo-aware code aliases, module names, namespaces/packages, and local import resolution data
- `state/embeddings.json` - cached graph-semantic vectors used by
swarmvault graph queryand embedding-backed similarity enrichment - `state/context-packs/` - JSON context-pack artifacts with citations, token-budget accounting, included items, and omitted items
- `state/chat-sessions/` - structured chat state used by
swarmvault chat --resume - `state/memory/tasks/` - JSON task ledger records with decisions, linked context packs, changed paths, outcomes, and follow-ups
When SWARMVAULT_OUT=<dir> is set, generated artifact directories such as raw/, wiki/, state/, agent/, and inbox/ live under that output root. swarmvault.config.json and swarmvault.schema.md remain in the project root.
- `state/benchmark.json` - latest graph-guided context benchmark summary
- `state/compile-state.json` - incremental compile metadata, source hashes, root and project-effective schema hashes, source project assignments, saved-output hashes, and candidate history
- `state/sessions/` - canonical session artifacts for compile, query, explore, lint, and watch
- `state/approvals/` - staged review bundles produced by
swarmvault compile --approve - `state/schedules/` - persisted state for configured schedule jobs
- `state/watch/` - watch-status and pending semantic refresh artifacts for tracked repo automation
- `state/jobs.ndjson` - automation run log written by
swarmvault watch - `.git/hooks/` - optional local git hooks managed by
swarmvault hook installwhen the vault lives inside a git repository - `.obsidian/` - optional Obsidian workspace files written by
swarmvault init --obsidian - `agent/` - agent-specific files and future agent integrations
Main Artifacts
- `state/graph.json` - compiled graph data
- `state/retrieval/` - local retrieval index directory, including the SQLite FTS shard and manifest
- `wiki/index.md` - entry point into the compiled markdown wiki
- `wiki/log.md` - human-readable recent activity trail for the vault
- `wiki/projects/index.md` - entry point into project-aware rollups
- `wiki/outputs/index.md` - index of saved answers and exploration runs
- `wiki/outputs/chat-sessions/<id>.md` / `state/chat-sessions/<id>.json` - persisted multi-turn conversations generated by
swarmvault chat - `wiki/context/<id>.md` / `state/context-packs/<id>.json` - bounded agent context packs generated by
swarmvault context build - `wiki/exports/ai/` - static AI handoff files generated by
swarmvault export ai - `wiki/memory/tasks/<id>.md` / `state/memory/tasks/<id>.json` - durable task ledgers generated by
swarmvault memory - `wiki/graph/share-card.md` / `wiki/graph/share-card.svg` / `wiki/graph/share-kit/` - compact post-ready text, visual graph summaries, and portable HTML-preview bundles generated by
swarmvault graph share - Generated page frontmatter - lifecycle fields such as
status,created_at,updated_at,compiled_from,managed_by,schema_hash, andproject_ids, plussource_typeon research-aware captures andsource_classon repo-aware page outputs
Configuration File
swarmvault.config.json controls workspace paths, provider configuration, task routing, viewer settings, and installed agent types. swarmvault.schema.md controls vault-specific naming, categorization, and grounding behavior. See Configuration and Schema for details.