smolbren index — build and update the search index

The index command scans every Markdown file in your vault, parses frontmatter and wikilinks, and writes the results into the search index used by all other commands. It must be run at least once before search, query, get, links, backlinks, types, or edges will work. After the initial index, run it again whenever your notes change to keep results current.

Synopsis

smolbren index [--full] [--vault <name>]

Flags

boolean

Build a complete replacement index from scratch. Every note is re-read, re-hashed, and re-written, and all edge resolutions are recomputed. The replacement is built and validated in a sibling directory, then atomically swapped into place. Any failure leaves the last good index readable. Use this after renaming or moving many files. For a damaged index, smolbren repair is the explicit recovery command.

string

Name of the vault to index. Defaults to the configured default vault. Use smolbren vault list to see registered vault names.

Output

index returns a single JSON object summarising what happened during the run.

{
  "scanned": 9,
  "unchanged": 0,
  "added": 9,
  "updated": 0,
  "removed": 0,
  "edges": 15,
  "unresolved_edges": 3,
  "duration_ms": 42
}

Field	Type	Description
`scanned`	integer	Total number of `.md` files found in the vault directory (hidden directories such as `.obsidian` are excluded)
`unchanged`	integer	Files whose content and metadata matched the stored state and were skipped
`added`	integer	Files that did not exist in the previous index and were newly added
`updated`	integer	Files whose content changed since the last index run
`removed`	integer	Files that were previously indexed but no longer exist on disk
`edges`	integer	Total number of wikilink edges present in the index after this run
`unresolved_edges`	integer	Edges whose target wikilink could not be matched to any indexed note
`duration_ms`	integer	Wall-clock time the index run took, in milliseconds

Incremental vs full rebuild

Incremental (default) — smolbren checks each file’s modification time and size before reading it. Files with unchanged metadata are skipped immediately. Files that are read but whose content hash matches the stored hash are updated in place (mtime converges) without triggering an edge recomputation. Only genuinely changed, added, or removed files are written to the dataset. This makes subsequent runs very fast on large vaults. Full rebuild (--full) — smolbren ignores all stored note and edge state, builds fresh datasets beside the live index, validates them, preserves existing embeddings, and performs a same-filesystem directory swap. Use a full rebuild when:

You have renamed or moved many files and want to recompute all edge resolutions from a clean state.
The incremental state appears corrupted or out of sync.
You want to verify the unresolved_edges count is accurate across the whole vault.

A full rebuild is always slower than an incremental run but guarantees a consistent, authoritative result without putting the last good index at risk.

Commands

Output & Errors

smolbren index — build and update the search index

Synopsis

Flags

Output

Incremental vs full rebuild

​Synopsis

​Flags

​Output

​Incremental vs full rebuild

Synopsis

Flags

Output

Incremental vs full rebuild