Transcribe handwritten multi-source research notes from reMarkable PDFs or paper photos. Trigger phrase "process deep dive notes". Handles research compilations with multiple sources per document. Converts handwriting to markdown, applies formatting conventions, generates metadata, creates AI summaries, and appends quotes to a central quotes file. Notes may be in Czech or English.
Transcribe handwritten multi-source research notes into structured Obsidian notes.
User says "process deep dive notes" and provides either:
Notes may be in Czech or English. Preserve the original language throughout. Do not translate.
For PDF input:
from pdf2image import convert_from_path
images = convert_from_path('notes.pdf', dpi=150)
for i, img in enumerate(images):
img.save(f'page_{i+1}.png', 'PNG')
For image input: Accept images directly and process in order.
Then visually read and transcribe each page.
Deep dive notes contain multiple sources (multiple H1 headings), each representing a different article, video, or resource.
If user provides single-source notes, suggest: "This looks like single-source book notes. Should I use 'process book notes' instead?"
See ../shared/formatting-rules.md for complete conventions. Key rules:
! at bullet start → italics!!! at bullet start → bold#) marks source titlesAuthor: *"Quote text"*Ask user for:
Then query Obsidian for existing tags:
obsidian-mcp-tools:search_vault_simple with relevant keywords
Select 3-5 tags that match the vault's existing taxonomy.
If transcribed content exceeds ~1 A4 of text, generate a 1-3 paragraph summary covering:
Place summary after frontmatter, before raw transcription.
---