deep-read-summarize

deep-read-summarize is designed for the common problem of long content being too time-consuming to read deeply and too messy to turn into useful notes. It accepts a book, paper, video link, or web page and runs a DSH workflow that splits the material into chunks, assigns them to parallel sub-agents for close reading, and merges the results into a structured Obsidian note.

The workflow emphasizes traceability and reliability rather than just producing a short summary. Key conclusions are anchored to source locations such as pages, chapters, or paragraphs; sub-task outputs are constrained by JSON Schema and retried when invalid; and a final quality check is applied before the note is finished. If one chunk fails, the workflow can skip it and mark the gap instead of silently producing a misleading complete summary.

It is a good fit for researchers, students, writers, and knowledge-management users who regularly process papers, ebooks, video transcripts, or long web articles. If you use Obsidian and want deep-reading results saved as Markdown with YAML frontmatter, this plugin helps turn scattered sources into reviewable, searchable notes.