reindex

name	reindex
description	Update the knowledge graph with new or changed markdown files. Re-ingests documents, extracts from new chunks, and resolves new duplicates. Use when the user has added, edited, or removed markdown files and wants the graph updated.

Reindex Knowledge Graph

Overview

This skill incrementally updates an existing knowledge graph when documents have changed. Only new or modified files are processed.

Tools

Use the kgmd MCP server tools: ingest_documents, get_chunks_for_extraction, store_extractions, get_resolution_candidates, apply_merges.

Workflow

Step 1: Re-ingest

Call ingest_documents(). It automatically:

• Skips unchanged files (hash match)
• Re-chunks modified files
• Adds new files

Report: how many documents were new, updated, or skipped.

Step 2: Extract from new chunks

If new chunks were created:

1. Call get_chunks_for_extraction(batch_size=10)
2. Extract entities and relations from each chunk
3. Call store_extractions() with results
4. Repeat until all new chunks are processed

Step 3: Resolve new duplicates

1. Call get_resolution_candidates() — new entities may duplicate existing ones
2. Review and confirm merges
3. Call apply_merges()

Step 4: Report

Summarize what changed:

• Documents added/updated
• New entities and relations
• Merges applied