The latest articles on DEV Community by gpt ai clips (@gptaiclips). https://dev.to/gptaiclips https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3934477%2F49cc3ae8-3d31-449f-8ce9-de663f7a78a2.gif https://dev.to/gptaiclips en gpt ai clips Sat, 16 May 2026 11:14:40 +0000 https://dev.to/gptaiclips/from-abandoned-repos-to-a-87k-obsidian-vault-a-three-pass-extraction-pattern-16f6 https://dev.to/gptaiclips/from-abandoned-repos-to-a-87k-obsidian-vault-a-three-pass-extraction-pattern-16f6 <p>Most of us have a folder full of repos that haven't been opened in months. Here is a pattern I have been using to turn those repos into a packaged developer product instead of guilt.</p> <h2> The three-pass pipeline </h2> <p>The core idea is to extract <em>decisions</em>, not descriptions. A file-level summary is useless to a future reader — what they want is the implicit reasoning the original author was preserving.</p> <h3> Pass 1 — file-level extraction </h3> <p>For every file, ask the model for four things: purpose, public surface, hidden invariants, and a risk score from 1 to 5. The risk score is the secret ingredient — it forces the model to find load-bearing logic.</p> <h3> Pass 2 — module-level clustering </h3> <p>Feed all of pass 1 back in and ask for clusters of files that share invariants. Each cluster becomes an Architecture Decision Record (ADR) with status, context, decision, and consequences.</p> <h3> Pass 3 — architecture-level graph </h3> <p>Apply Leiden clustering across the ADRs to surface the cross-cutting concepts. Each graph node carries a one-paragraph 'why this matters to a maintainer' note. Leiden gives stabler cluster boundaries than vanilla modularity on small graphs.</p> <h2> Why Sonnet 4.6 matters here </h2> <p>With a 1M-token context window I can run the whole-repo pass without first summarizing per file. Per-file summarization is where cross-references die — once you compress, you lose the links the graph step depends on.</p> <h2> Packaging into Obsidian </h2> <p>Graphify (the Claude Code skill, ~37K stars) has an <code>--obsidian</code> flag that writes the graph as a markdown vault with backlinks already wired up. Add ADR templates and you have a product, not a dump.</p> <h2> Quick demo </h2> <p> <iframe src="https://www.youtube.com/embed/j1x-wi8CzRM"> </iframe> </p> <h2> Three-tier pricing </h2> <ul> <li>$0 — sampler vault with two ADRs and the graph view</li> <li>$49 — full vault with all ADRs and concept notes</li> <li>$149 — full vault plus raw prompts and ADR templates so the buyer can run the pipeline on their own code</li> </ul> <p>Top developer vaults on Gumroad clear 3K+ copies a year, so the ceiling is real.</p> <h2> Try it on your own repo </h2> <p>Checkout the <a href="https://github.com/" rel="noopener noreferrer">Graphify project</a> and the longer walkthroughs over at <a href="https://cptdigital.com" rel="noopener noreferrer">cptdigital.com</a>. The bottleneck is almost always the prompt for pass 1, not the model.</p> ai productivity tutorial webdev