index

name	index
description	LogosDB — incremental index of a file or directory (changed/new files only; fast refresh, concise output).
argument-hint	<path> [--namespace=name] [-n name]
model	sonnet

/index

User invoked with: $ARGUMENTS

Instructions

Index a file or directory into LogosDB for semantic search using incremental mode (requires logosdb-mcp-server ≥ 0.7.11).

Parse the arguments:

• First positional argument is the path to index (file or directory).
• --namespace=<name> or -n <name> sets the collection name (default: code).

Call the logosdb_index_file MCP tool with the resolved path, namespace, and incremental: true.

Incremental: new or changed files only; changed files have old chunks removed first; unchanged files are skipped; directory runs also drop chunks for files removed from disk (see LogosDB mcp/README.md). Cost: proportional to changed chunks, not the whole tree — suitable for session-start refresh or after edits when the user wants memory aligned with disk.

Session load (semantic-memory plugin): the model-invoked semantic-memory skill requires /index . at every session start when the plugin is active.

Output format (strict)

Respond with one line only. Do not echo the raw tool JSON. Use the tool result fields indexed, indexed_files, skipped_files, pruned_files, and namespace:

Indexed {indexed} chunks · {indexed_files} updated · {skipped_files} skipped · {pruned_files} pruned → '{namespace}'

If indexed_files == 0 && skipped_files > 0, append in the same line: (no changes since last index — touch the file or delete the namespace to force re-index).

Related

Setup, namespaces, and CLAUDE.md template: skills/semantic-memory/SKILL.md and skills/semantic-memory/references/claude-md-template.md.