Exporting

This page walks the four document-set exporters: Obsidian (vault), Anki (deck), wiki (per-subject cited articles), and Logseq (graph). Each writes to the output path you pass on the CLI.

Three more export targets have different shapes and live elsewhere:

• graph — a scoped, self-contained HTML render of the store's epistemics rather than a document set. It has its own page: Graph view.
• notion — an idempotent upsert into a Notion database over the HTTP API, so it writes no files at all (--dry-run plans it with zero API writes). It needs a credential; see Operator guide → configuration → secrets.
• jsonl — line-delimited JSON for downstream tooling.

particles export --help and the command reference carry the full, always-current flag list for each.

Obsidian vault

uv run particles export obsidian ./my-vault

Writes one .md per Subject plus a _index.md. Coin subjects get the Numismatic infobox template; pivot subjects get a minimal linking template; everything else gets the generic callout template.

Useful flags:

• --min-particles N — skip subjects with fewer than N ACTIVE particles after the quality filter.
• --min-links N — skip subjects with fewer than N graph links (in
• out combined). Reduces phantom-node noise in the graph view.
• --with-synthesis — splice an LLM-synthesised prose article into each note. Requires ANTHROPIC_API_KEY. Shares the article-cache with export wiki, so running both pays the LLM cost once per subject. Also emits one cited-prose note per NARRATIVE under Narratives/ — a journal entry rendered as its whole-entry narrative. Disable with obsidian.emit_narrative_notes: false; with no key these notes fall back to a deterministic cited listing.
• --invalidate-stale-links — drop the article-cache hash from any note whose [[X]] wikilinks reference a renamed subject.
• --min-particle-confidence F — cross-exporter quality filter . Drops particles whose effective_confidence is below F from every rendered note. To set a standing floor for every export instead of passing it each run, see Operator guide → cross-exporter quality threshold; for what effective_confidence is composed of, see Concepts → confidence.

Anki deck

uv run particles export anki ./deck.txt --deck-name=Numismatics

Writes a tab-delimited file. One card per particle, grouped into decks by subject. Structured particles emit one card per properties key; descriptive particles emit one card from the content.

Useful flags:

• --deck-name NAME — root deck name prefix.
• --min-particle-confidence F — same cross-exporter filter.

(The exporter also accepts a max_cards_per_subject option via the Python / HTTP API; it is not exposed as a CLI flag.)

Wiki articles

uv run particles export wiki ./my-wiki

Writes a flat directory of per-Subject Markdown articles, each with LLM-synthesised prose and inline citations back to the corpus entries. Plus a top-level index.md.

Useful flags:

• --dry-run — report cache hits + regen count + estimated token spend; no LLM calls or file writes.
• --regenerate-all — bypass the per-subject input-hash cache.
• --invalidate-stale-links — same wikilink-staleness behaviour as Obsidian.
• --subjects "A,B,C" — limit to specific canonical names.
• --min-particles N — minimum post-filter particle count (default 3 per config.wiki.min_particles).
• --min-particle-confidence F — cross-exporter quality filter.

The export also writes one cited article per NARRATIVE under Narratives/, using the same render path and cache as the Obsidian narrative notes. Disable with wiki.emit_narrative_notes: false. Narratives are subject-less, so --subjects suppresses them.

Logseq graph

uv run particles export logseq ./my-graph

Writes pages/<subject>.md in Logseq's native bullet-outline format . Each particle is emitted as a block whose id:: is the particle ID, enabling cross-page citation via ((<particle_id>)) syntax.

Useful flags:

• --with-synthesis — same synthesis splice as Obsidian; shares the article cache, so running multiple synthesising exporters pays the LLM cost once per subject. It also emits one cited-prose page per NARRATIVE in Logseq's Narratives/ page namespace (on disk pages/Narratives___<slug>.md) and adds a ## Narratives backlink block to each subject page whose claims take part in one. Disable with logseq.emit_narrative_notes: false.
• --invalidate-stale-links, --min-particles, --min-links, --min-particle-confidence — same semantics as Obsidian.

The dry-run summary

Each exporter returns a typed Pydantic summary. The CLI prints it as key: value lines after the run. Conditional fields (synthesis counts, articles_written) only appear when the producing step actually ran. For machine consumption, the same model is available as *.model_dump_json() for downstream tooling.

What about other formats?

The plugin registry is designed so an exporter is one file plus one registry line. If the format you want isn't in the list above, writing it is a small job — see Plugin-author guide → writing an exporter, which walks the contract, the worked examples in the tree, and the cross-exporter options every exporter must honour.