Fix UI layout, AI renaming logic, bank statement extraction, and duplicate detection entity-skipping bug

Author: ptkvaibhav

package.json

@@ -26,6 +26,7 @@
     "node": ">=22"
   },
   "scripts": {
+    "start": "node src/cli.js",
     "plan": "node src/cli.js plan",
     "doctor": "node src/cli.js doctor",
     "demo": "node src/cli.js demo",

src/core/entity-detector.js

@@ -54,12 +54,16 @@ export async function detectCohesiveEntity(absoluteDirPath) {
     }
 
     const prompt = `You are a strict, deterministic file system analysis AI. Decide if a directory is a "Cohesive Entity".
-A Cohesive Entity is a folder that MUST NOT have its internal files separated or moved.
+A Cohesive Entity is a strict, machine-generated folder that MUST NOT have its internal files separated or moved.
 Examples:
-- Software projects (contain code, package.json)
-- Installed applications (contain .exe, .dll)
+- Software projects (contain code, package.json, build configs)
+- Installed applications (contain .exe, .dll, binary assets)
 - System caches or hidden metadata folders (e.g., .thumbnails, .cache, .git)
-- Device backups or OS images
+
+CRITICAL RULES:
+- A folder containing mostly user documents (PDFs, Word docs, Excel, media, loose files) is NEVER a cohesive entity.
+- If it looks like a user's personal organization folder, return false.
+- Do NOT flag device backups as entities unless they are raw disk images.
 
 Example 1:
 Directory Name: ".thumbnails"

src/organization/purpose-rules.js

@@ -128,6 +128,16 @@ export function inferPurposeDetails({ absolutePath = "", baseName = "", extensio
     };
   }
 
+  // Specific Deep Content Rule: Bank Statement
+  if (normalizedBaseName.includes("statement") || /account\s*statement/i.test(extractedText) || /bank\s*statement/i.test(extractedText) || (/account\s*summary/i.test(extractedText) && /balance/i.test(extractedText))) {
+     return {
+       purpose: "finance",
+       expectedFolders: ["Finance/Bank_Statements"],
+       matchedByRule: true,
+       renameLabel: "Bank_Statement"
+     };
+  }
+
   // Specific Deep Content Rule: Aadhaar Card
   if (normalizedBaseName.includes("aadhaar") || /unique identification authority of india/i.test(extractedText) || (/government of india/i.test(extractedText) && /aadhaar/i.test(extractedText))) {
      return {

src/organization/structure.js

@@ -3,7 +3,7 @@ import { classifyFile } from "../core/classify.js";
 import { inferPurposeDetails } from "./purpose-rules.js";
 
 const GENERIC_NAME_PATTERNS = [
-  /^(file|document|scan|copy|duplicate|new|untitled)[-_ ]?\d*$/i,
+  /^(file|document|scan|copy|duplicate|new|untitled)[-_ ]?[a-z0-9]*$/i,
   /^(img|image|photo|vid|video|media|whatsapp|screenshot)[-_ ]?[a-z0-9-_() ]+$/i,
   /^\d{8,}$/, // long numeric strings
   /^[a-f0-9]{16,}$/i // long hex hashes

src/server.js

@@ -187,12 +187,20 @@ Give me ONLY a raw JSON string like {"exclusions": ["folder1"], "reasoning": "wh
   // Step 5: AI Renaming reasoning
   app.post("/api/ai/rename", async (req, res) => {
     try {
-      const { fileInfo } = req.body;
-      const prompt = `I have a file with info ${JSON.stringify(fileInfo)}. Propose a rename and give reasoning.
-Ensure names are deterministic (e.g., lowercase, underscores).
-If it's a Finance/Form_16, ensure the year is prominent.
-Return ONLY a raw JSON string like {"proposedName": "file.pdf", "reasoning": "why"}. No markdown.`;
-      const aiResponse = await askAI(prompt, "You are a highly deterministic file renaming assistant. You must follow standard naming conventions and prioritize semantic clarity based on file content and metadata.");
+      const { fileInfo, subjectPath } = req.body;
+      const file = catalog.getFileByPath(subjectPath);
+      const textSample = file?.extractedText ? file.extractedText.slice(0, 800) : "No text available.";
+      
+      const prompt = `I have a file named "${fileInfo.currentName}" with category "${fileInfo.category}" and purpose "${fileInfo.purpose}".
+Here is a sample of its extracted text content:
+---
+${textSample}
+---
+Analyze the text to determine exactly what this file is (e.g. Bank Statement, Aadhaar Card, Offer Letter, etc.) and who it belongs to if applicable.
+Propose a highly descriptive and structured file name. Use Spaces, Title Case, and clear descriptors (e.g., "Pratik Vaibhav - Aadhaar Card.pdf" or "HDFC Bank Statement - Jan 2024.pdf").
+Do not just return the original name or "document_123.pdf".
+Return ONLY a raw JSON string like {"proposedName": "New Name.pdf", "reasoning": "why"}. No markdown.`;
+      const aiResponse = await askAI(prompt, "You are a highly intelligent file renaming assistant. You must analyze the text content to extract the semantic meaning of the document and propose a human-readable, descriptive name.");
       const clean = cleanJSON(aiResponse);
       res.json(JSON.parse(clean));
     } catch (error) {

test/local-audit.test.js

@@ -90,5 +90,5 @@ Important folder candidates once structure exists:
 
   const unstructuredGeneric = audit.unstructuredFiles.find((file) => file.relativePath === "file123.pdf");
   assert.ok(unstructuredGeneric);
-  assert.equal(unstructuredGeneric.renameRecommended, false);
+  assert.equal(unstructuredGeneric.renameRecommended, true);
 });

test/review-queue.test.js

@@ -18,8 +18,8 @@ test("buildOrganizationProposals creates approval-gated move and rename proposal
       category: "document"
     },
     structure: {
-      purpose: "document",
-      expectedFolders: ["Documents"],
+      purpose: "finance",
+      expectedFolders: ["Finance"],
       moveRecommended: true,
       renameRecommended: true,
       reasons: [
@@ -39,13 +39,13 @@ test("buildOrganizationProposals creates approval-gated move and rename proposal
   assert.ok(move);
   assert.equal(move.status, "pending_user_approval");
   assert.equal(move.approvalGate, "moving files in batch");
-  assert.equal(move.evidence.proposedRelativePath, "Documents/file123.pdf");
+  assert.equal(move.evidence.proposedRelativePath, "Finance/file123.pdf");
 
   const rename = proposals.find((proposal) => proposal.action === "rename_file");
   assert.ok(rename);
   assert.equal(rename.status, "pending_user_approval");
   assert.equal(rename.approvalGate, "renaming files");
-  assert.equal(rename.proposedName, "Document_abcdef12.pdf");
+  assert.equal(rename.proposedName, "Finance_Record_file123.pdf");
   assert.equal(rename.evidence.sha256, "abcdef1234567890");
 });
 

ui/dist/assets/index-B8Ts6tKM.js ui/dist/assets/index-Bef_zoun.jsui/dist/assets/index-B8Ts6tKM.js renamed to ui/dist/assets/index-Bef_zoun.js

@@ -5,7 +5,7 @@
     <link rel="icon" type="image/svg+xml" href="/favicon.svg" />
     <meta name="viewport" content="width=device-width, initial-scale=1.0" />
     <title>ui</title>
-    <script type="module" crossorigin src="/assets/index-B8Ts6tKM.js"></script>
+    <script type="module" crossorigin src="/assets/index-Bef_zoun.js"></script>
     <link rel="stylesheet" crossorigin href="/assets/index-BMzdelOc.css">
   </head>
   <body>

ui/dist/index.html

@@ -518,13 +518,13 @@ function App() {
                       </div>
                       <div>
                         <div className="text-xs font-bold text-green-500 uppercase tracking-widest mb-3 flex items-center gap-2"><CheckCircle className="w-4 h-4"/> Keep Original</div>
-                        <div className="p-5 bg-green-500/5 border border-green-500/20 rounded-xl text-sm font-mono truncate shadow-inner" title={item.evidence.proposedKeepPath || item.evidence.keptPath}>
+                        <div className="p-5 bg-green-500/5 border border-green-500/20 rounded-xl text-sm font-mono break-all shadow-inner" title={item.evidence.proposedKeepPath || item.evidence.keptPath}>
                           {(item.evidence.proposedKeepPath || item.evidence.keptPath)?.split('\\').pop()}
                         </div>
                       </div>
                       <div>
                         <div className="text-xs font-bold text-rose-500 uppercase tracking-widest mb-3 flex items-center gap-2"><Trash2 className="w-4 h-4"/> Delete Duplicate</div>
-                        <div className="p-5 bg-rose-500/5 border border-rose-500/20 rounded-xl text-sm font-mono truncate opacity-60 line-through decoration-rose-500/50 shadow-inner" title={item.evidence.proposedDeletePaths?.[0] || item.evidence.deletedPaths?.[0]}>
+                        <div className="p-5 bg-rose-500/5 border border-rose-500/20 rounded-xl text-sm font-mono break-all opacity-60 line-through decoration-rose-500/50 shadow-inner" title={item.evidence.proposedDeletePaths?.[0] || item.evidence.deletedPaths?.[0]}>
                           {(item.evidence.proposedDeletePaths?.[0] || item.evidence.deletedPaths?.[0])?.split('\\').pop()}
                         </div>
                       </div>
@@ -582,11 +582,11 @@ function App() {
                           {item.action.replace('_', ' ')}
                         </span>
                       </td>
-                      <td className="px-6 py-5 pr-8 truncate">
+                      <td className="px-6 py-5 pr-8">
                         <div className="flex flex-col gap-3 mb-4">
                           <div className="flex items-center gap-3 text-slate-500 line-through decoration-rose-500/50 overflow-hidden">
                             <div className="p-2 bg-slate-950 rounded border border-slate-800 shrink-0"><FileMinus className="w-4 h-4 text-rose-400"/></div>
-                            <span className="text-xs font-mono truncate w-full" title={item.subjectPath}>{item.subjectPath}</span>
+                            <span className="text-xs font-mono break-all w-full" title={item.subjectPath}>{item.subjectPath}</span>
                             <a href={`/api/file?path=${encodeURIComponent(item.subjectPath)}`} target="_blank" rel="noreferrer" className="text-sky-400 hover:text-sky-300 transition-colors ml-auto shrink-0" title="View File">
                               <Eye className="w-4 h-4" />
                             </a>
@@ -610,7 +610,7 @@ function App() {
                               </div>
                             ) : (
                               <>
-                                <span className="text-sm font-bold text-sky-400 truncate w-full" title={item.proposedPath || item.evidence?.proposedName}>
+                                <span className="text-sm font-bold text-sky-400 break-all w-full" title={item.proposedPath || item.evidence?.proposedName}>
                                   {item.proposedPath || item.evidence?.proposedName}
                                 </span>
                                 <button