Name the CV step

README.md

@@ -6,12 +6,12 @@ colorTo: gray
 sdk: static
 pinned: false
 license: cc-by-sa-4.0
-short_description: "P&ID image to tables and a graph, with pixel connections"
+short_description: "P&ID to a graph: VLM symbols, classic CV connections"
 ---
 
 # pid-parser showcase
 
-Precomputed results of [pid-parser](https://github.com/dzakybd/pid-parser) on one very dense P&ID sheet (190 symbols, 732 connected pairs), run two ways, the skill on Claude Sonnet 5.5 and the Deep Agents runner on gpt-6-luna: the drawing with the symbols the model found, the connections read from the pixels, the graph, the tables, and the SQL, CSV, GraphML and Cypher exports. The page calls no model.
+Precomputed results of [pid-parser](https://github.com/dzakybd/pid-parser) on one very dense P&ID sheet (190 symbols, 732 connected pairs), run two ways, the skill on Claude Sonnet 5.5 and the Deep Agents runner on gpt-6-luna: the drawing with the symbols the model found, the connections found by traditional computer vision, the graph, the tables, and the SQL, CSV, GraphML and Cypher exports. The page calls no model.
 
 Install the skill in your own agent with `npx skills add dzakybd/pid-parser`. How it works, step by step: [the write-up](https://dzaky.ai.id/en/blog/pid-parser-pid-to-graph).
 

index.html

@@ -4,7 +4,7 @@
 <meta charset="utf-8">
 <meta name="viewport" content="width=device-width, initial-scale=1">
 <title>pid-parser showcase</title>
-<meta name="description" content="P&ID image to tables and graph with a vision-language model for the symbols and pixels for the connections: precomputed results on one very dense PID2Graph sheet.">
+<meta name="description" content="P&ID image to tables and graph with a vision-language model for the symbols and traditional computer vision for the connections: precomputed results on one very dense PID2Graph sheet.">
 <script src="https://cdnjs.cloudflare.com/ajax/libs/cytoscape/3.30.2/cytoscape.min.js"></script>
 <style>
 :root {
@@ -54,7 +54,7 @@ footer { margin-top: 28px; font-size: 13px; color: var(--muted); }
 <body>
 <main>
   <h1>pid-parser</h1>
-  <p class="muted">A P&amp;ID image in, tables and a graph out. A vision-language model finds the symbols tile by tile; the connections are read from the pixels (no model); small scripts check the graph and export it for SQL or a graph database. No GPU, no training.</p>
+  <p class="muted">A P&amp;ID image in, tables and a graph out. A vision-language model finds the symbols tile by tile; the connections come from traditional computer vision (thresholding and connected-component labeling, no model); small scripts validate the graph and export it for SQL or a graph database. No GPU, no training.</p>
   <pre>npx skills add dzakybd/pid-parser</pre>
   <p class="muted" style="margin-top:8px">Works in Claude Code, Codex and other skill-aware agents, or as a LangChain Deep Agent. Source: <a href="https://github.com/dzakybd/pid-parser">github.com/dzakybd/pid-parser</a>. How it works, step by step: <a href="https://dzaky.ai.id/en/blog/pid-parser-pid-to-graph">the write-up</a>. The sheet is one very dense P&amp;ID (190 symbols, 732 connected symbol pairs); the two tabs are two runs on it. These results were computed ahead of time; this page calls no model.</p>