interfluence: Ingest Writing Samples

interfluence: Ingest Writing Samples

You are ingesting writing samples into the interfluence corpus for voice analysis.

Process

1. Determine the source type:

   • File path: Read the file directly and use corpus_add with the file path
   • Directory: Glob for text/markdown files and ingest each one using corpus_add
   • URL: Use WebFetch to extract the text content, then use corpus_add_text with source "url"

2. For URLs (especially Substack, blogs, Notion exports):

   • Fetch the page and extract the main article text
   • Strip navigation, footers, comments, and other non-article content
   • Preserve the author's formatting: paragraph breaks, headers, lists, emphasis
   • Set the title from the page title or first heading

3. After ingestion:

   • Show the user what was added (title, word count, sample ID)
   • If this is the first sample, suggest running /interfluence:voice-analyze to generate a voice profile
   • If a voice profile already exists, suggest re-running analysis to incorporate the new sample

Important

• The projectDir parameter should be the current working directory
• Always preserve the author's original text faithfully — do not clean up, edit, or summarize
• For directories, confirm with the user before ingesting many files