Skip to content
FSS Knowledge Base

Ideas-Workflow — Upgrades

Section titled “Ideas-Workflow — Upgrades”

Ideas discovered but not committed to a phase. Promote into ROADMAP.md when concrete.

  • Clipping enrichment on save: rate each clipping 0–10, summarize key ideas + how they benefit (from Later/UPGRADES.md → Idea Capture)
  • Mobile capture: Obsidian Web Clipper works on mobile without pro subscription; Syncthing + Mobius Sync recommended for reliable, private, free (single-vault) bidirectional Android/iOS/desktop sync — share with Sean once working; fallback: try Obsidian mobile 1 min if the sync alternative isn’t good enough (from Later/UPGRADES.md)
  • Key-phrase highlighting: bold/color key phrases inside kept clippings during triage
  • Automation processing lists: batch-process a list of URLs (load via clipboard?) (Dynalist import)
  • d-squared YouTube summaries via Web Clipper (Dynalist import)
  • Coconote-style processing test: run one Medium article through the pipeline end-to-end as the acceptance test — e.g. the Coconote $6.7M-ARR article (URL in idea-capture-workflow § Web content test)
  • Briefing inspiration: “AI daily newspaper” Medium article (URL in idea-capture-workflow § Background 2) — rejected the newspaper format, kept the curation concept
  • Globe PDF backlog: convert all existing Globe_*.pdf in D:\FSS\Static\Info\Finance\ (naming pattern Globe_BeatingTheTSX-2025AP03.pdf) to searchable KB text once P4 PDF path exists — “much better workflow: convert into searchable KB text, like clippings”. Deterministic extraction: Python PyMuPDF or pdfplumber → markdown gives granular control + high accuracy for text PDFs (https://artifex.com/blog/rag-llm-and-pdf-conversion-to-markdown-text-with-pymupdf). Original flow was specified-page → PDF → .md (from Later/Automations.md)
  • General Summary clip template alongside YouTube/Medium ones — community templates may exist (Dynalist import)
  • Triage dedupe-by-source (v0.1): /triage-inbox Step 5 should check the target Inbox for an existing bullet on the clip’s source URL before appending — acceptance test nearly duplicated the Coconote bullet routed earlier from the origin task (caught manually, merged instead; 2026-07-17)
  • Re-evaluate PDF extraction against MarkItDown: ~/ai-config/scripts/pdf-to-clippings.py (built P4, 2026-07-18) uses PyMuPDF with a custom glyph-spacing fix for Globe & Mail’s subsetted fonts. MarkItDown (installed 2026-07-28, Core/IT/Tasks/MarkItDown.md) is a general-purpose doc→markdown converter that may handle this font issue natively and cover other formats (DOCX/PPTX/XLSX/images) for free. Worth a side-by-side test against the Globe backlog before extending the pipeline further — don’t rebuild on faith, verify output quality first.