Import Content

Scenario: Used WebFetch to retrieve content, but Kurt DB has ERROR records.

Steps:

1. List ERROR records:

   kurt content list --status ERROR
   

2. Find corresponding markdown files:

   find sources -name "*.md" -type f
   

3. For each ERROR record with a matching file:

   python .claude/scripts/import_markdown.py \
     --document-id <doc-id> \
     --file-path <file-path>
   
   # Then extract metadata
   kurt index <doc-id>
   

4. Verify success:

   kurt content get-metadata <doc-id>
   

Workflow 2: Bulk Import All ERROR Records

Scenario: Multiple ERROR records with existing markdown files.

Create bash script:

#!/bin/bash
# Fix all ERROR records with matching files

while read -r doc_id url status; do
  if [ "$status" = "ERROR" ]; then
    # Try to find corresponding file
    # (Implement file finding logic based on URL)

    if [ -f "$file_path" ]; then
      echo "Importing: $doc_id"
      python .claude/scripts/import_markdown.py \
        --document-id "$doc_id" \
        --file-path "$file_path"

      kurt index "$doc_id"
    fi
  fi
done < <(kurt content list)

Workflow 3: Import Manually Created Files

Scenario: You created markdown files directly without using kurct content.

Steps:

1. Verify file exists and has content

2. Check if document record exists:

   kurt content list --url-contains <domain>
   

3. If ERROR record exists, import:

   python .claude/scripts/import_markdown.py \
     --document-id <doc-id> \
     --file-path <file-path>
   

4. If no record exists, use kurct content:

   # Create record and import
   kurct content add <url>
   # Then import file content
   python .claude/scripts/import_markdown.py \
     --document-id <new-doc-id> \
     --file-path <file-path>
   

Auto-Import Hook

Files written to /sources/ or projects/*/sources/ are automatically imported via PostToolUse hook.

How it works:

1. Claude writes markdown file to sources
2. PostToolUse hook triggers
3. Script maps file path → URL
4. Finds ERROR record for URL
5. Updates record to FETCHED
6. Extracts metadata
7. Shows confirmation message

Hook location: .claude/settings.json

Script: .claude/scripts/auto-import-source.sh

Logs: .claude/logs/auto-import.log

YAML Frontmatter & Metadata Extraction

The import script automatically parses YAML frontmatter from markdown files and populates database metadata fields.

Supported Metadata Fields

The following frontmatter fields are automatically extracted and stored in Kurt database:

Frontmatter Format

Use YAML frontmatter at the start of markdown files:

---