Skip to content

Searching

docbank search insurance
docbank search tax 2026
docbank search return --tag taxes
docbank search report --mime-type application/pdf
docbank search receipt --under /taxes/2026
docbank search report --modified-since 2026-01-01T00:00:00Z
docbank search report --modified-before 2026-04-01T00:00:00Z
docbank search report --limit 200
docbank search report --json

SELECTOR   MATCH    PATH
id:231     name     /taxes/2026/insurance-renewal.pdf
id:198     content  /taxes/2026/car-insurance-notes.md

Semantics

  • Prefix matching. Every whitespace-separated term matches word prefixes: insur finds insurance-renewal.pdf. Multiple terms must all match.
  • Operator-safe. Query text is escaped before reaching FTS5, so AND, OR, quotes, and parentheses in a query are searched for literally, never interpreted. Any string is a safe query.
  • Ranked and bounded. Name matches retain their BM25 order and appear first. Content-only matches follow in their own BM25 order; both groups use deterministic name/ID tie-breaks. The default limit is 50; --limit accepts 1–1000, and truncation is always reported.
  • Live nodes only. Trashed documents don't appear; restore returns them to the index. Renames update the index immediately.
  • Current content only. Retained prior versions stay available through docbank versions, but ordinary search matches the current selected version of each live file.
  • Stable tag filtering. --tag <name-or-id> requires one current tag assignment without changing name-before-content ranking. The CLI resolves a tag name to its stable UUID before searching; JSON echoes that UUID as tag_id, so a later rename cannot change what the request meant.
  • Current media-type filtering. --mime-type <type/subtype> requires the current file version to have that base media type. The filter is case-insensitive and ignores stored parameters, so text/plain also matches text/plain; charset=utf-8. The request itself must be parameter-free; directories and historical versions never match it.
  • Stable directory scoping. --under <path-or-id> restricts results to descendants of one live directory. The CLI resolves a path or id:N selector to its stable node ID before searching; JSON echoes that ID as under_node_id. The selected directory itself is not a result, and moving or renaming it does not change its identity.
  • Current modification time. --modified-since <timestamp> includes nodes modified at or after an absolute RFC3339 timestamp; --modified-before <timestamp> excludes that timestamp and everything after it. Together they form a half-open interval, so adjacent searches do not duplicate a boundary result. Inputs with an explicit offset are normalized to UTC and echoed in JSON. The bounds apply to the live node's current modified_at, not filesystem provenance or the age of retained versions.

For scripts, --json returns hits, limit, and truncated without table formatting. hits is always an array, including when nothing matches.

Text extraction

The daemon's extract:plain-text background job indexes UTF-8 content whose media type is text/*, application/json, application/x-ndjson, or application/jsonl. This covers plain text, Markdown, CSV, JSON, and JSONL. The source blob is read through Docbank's verified loose/packed interface, and text becomes searchable only after the complete stream reaches verified EOF.

Extraction is bounded to 16 MiB per blob. Larger documents, invalid UTF-8, and text containing NUL bytes remain stored and readable but are not body-indexed. Newly ingested or replaced content may take a few seconds to appear while the daemon job reaches it. A transient open, read, or verification error leaves the item queued and is retried on a bounded delay; it does not become a permanent extraction failure. docbank jobs shows whether that worker is running.

PDF text layers, office formats, and OCR are unsupported. Their absence never changes name-search results or document authority.

Next: organize documents beyond paths with Organizing & Tagging, or see every search flag in the CLI Reference.