combicode 1.7.3 → 2.0.0

package/index.js

@@ -8,25 +8,34 @@ const ignore = require("ignore");
 
 const { version } = require("./package.json");
 
+// ---------------------------------------------------------------------------
+// System Prompts (v2.0.0)
+// ---------------------------------------------------------------------------
+
 const DEFAULT_SYSTEM_PROMPT = `You are an expert software architect. The user is providing you with the complete source code for a project, contained in a single file. Your task is to meticulously analyze the provided codebase to gain a comprehensive understanding of its structure, functionality, dependencies, and overall architecture.
 
-A file tree is provided below to give you a high-level overview. The subsequent sections contain the full content of each file, clearly marked with a file header.
+A code map with expanded tree structure \`<code_index>\` is provided below to give you a high-level overview. The subsequent section \`<merged_code>\` contain the full content of each file (read using the command \`sed -n '<ML_START>,<ML_END>p' combicode.txt\`), clearly marked with a file header.
 
 Your instructions are:
-1.  **Analyze Thoroughly:** Read through every file to understand its purpose and how it interacts with other files.
-2.  **Identify Key Components:** Pay close attention to configuration files (like package.json, pyproject.toml), entry points (like index.js, main.py), and core logic.
+1.  Analyze Thoroughly: Read through every file to understand its purpose and how it interacts with other files.
+2.  Identify Key Components: Pay close attention to configuration files (like package.json, pyproject.toml), entry points (like index.js, main.py), and core logic.
+3.  Use the Code Map: The code map shows classes, functions, loops with their line numbers (OL = Original Line, ML = Merged Line) and sizes for precise navigation.
 `;
 
-const LLMS_TXT_SYSTEM_PROMPT = `You are an expert software architect. The user is providing you with the full documentation for a project, sourced from the project's 'llms.txt' file. This file contains the complete context needed to understand the project's features, APIs, and usage for a specific version. Your task is to act as a definitive source of truth based *only* on this provided documentation.
+const LLMS_TXT_SYSTEM_PROMPT = `You are an expert software architect. The user is providing you with the full documentation for a project. This file contains the complete context needed to understand the project's features, APIs, and usage for a specific version. Your task is to act as a definitive source of truth based *only* on this provided documentation.
 
 When answering questions or writing code, adhere strictly to the functions, variables, and methods described in this context. Do not use or suggest any deprecated or older functionalities that are not present here.
 
-A file tree of the documentation source is provided below for a high-level overview. The subsequent sections contain the full content of each file, clearly marked with a file header.
+A code map with expanded tree structure is provided below for a high-level overview.
 `;
 
-// Minimal safety ignores that should always apply
+// Minimal safety ignores
 const SAFETY_IGNORES = [".git", ".DS_Store"];
 
+// ---------------------------------------------------------------------------
+// Utility helpers
+// ---------------------------------------------------------------------------
+
 function isLikelyBinary(filePath) {
   const buffer = Buffer.alloc(512);
   let fd;
@@ -42,17 +51,1093 @@ function isLikelyBinary(filePath) {
 }
 
 function formatBytes(bytes, decimals = 1) {
-  if (bytes === 0) return "0 B";
+  if (bytes === 0) return "0B";
   const k = 1024;
   const dm = decimals < 0 ? 0 : decimals;
   const sizes = ["B", "KB", "MB", "GB", "TB"];
   const i = Math.floor(Math.log(bytes) / Math.log(k));
-  return parseFloat((bytes / Math.pow(k, i)).toFixed(dm)) + "" + sizes[i];
+  return parseFloat((bytes / Math.pow(k, i)).toFixed(dm)) + sizes[i];
+}
+
+// ---------------------------------------------------------------------------
+// Code Parsers (regex-based for all languages)
+// ---------------------------------------------------------------------------
+
+/**
+ * Parse code structure from file content. Returns array of elements:
+ * { type, label, startLine, endLine, startByte, endByte }
+ *
+ * type: "class" | "fn" | "async" | "ctor" | "loop" | "impl" | "test" | "describe"
+ */
+function parseCodeStructure(filePath, content) {
+  const ext = path.extname(filePath).toLowerCase();
+  const lines = content.split("\n");
+
+  switch (ext) {
+    case ".py":
+      return parsePython(lines);
+    case ".js":
+    case ".jsx":
+    case ".mjs":
+    case ".cjs":
+      return parseJavaScript(lines);
+    case ".ts":
+    case ".tsx":
+    case ".mts":
+    case ".cts":
+      return parseTypeScript(lines);
+    case ".go":
+      return parseGo(lines);
+    case ".rs":
+      return parseRust(lines);
+    case ".java":
+      return parseJava(lines);
+    case ".c":
+    case ".h":
+    case ".cpp":
+    case ".hpp":
+    case ".cc":
+    case ".cxx":
+      return parseCCpp(lines);
+    case ".cs":
+      return parseCSharp(lines);
+    case ".php":
+      return parsePHP(lines);
+    case ".rb":
+      return parseRuby(lines);
+    case ".swift":
+      return parseSwift(lines);
+    case ".kt":
+    case ".kts":
+      return parseKotlin(lines);
+    case ".scala":
+    case ".sc":
+      return parseScala(lines);
+    case ".lua":
+      return parseLua(lines);
+    case ".pl":
+    case ".pm":
+      return parsePerl(lines);
+    case ".sh":
+    case ".bash":
+    case ".zsh":
+      return parseBash(lines);
+    default:
+      return [];
+  }
+}
+
+/**
+ * Find the end of a block that starts at `startLine` using brace/indent counting.
+ * For brace-based languages.
+ */
+function findBraceBlockEnd(lines, startLine) {
+  let depth = 0;
+  let foundOpen = false;
+  for (let i = startLine; i < lines.length; i++) {
+    const line = lines[i];
+    for (const ch of line) {
+      if (ch === "{") {
+        depth++;
+        foundOpen = true;
+      } else if (ch === "}") {
+        depth--;
+        if (foundOpen && depth === 0) {
+          return i;
+        }
+      }
+    }
+  }
+  return lines.length - 1;
+}
+
+/**
+ * Find block end for Python (indent-based).
+ */
+function findIndentBlockEnd(lines, startLine) {
+  if (startLine >= lines.length) return startLine;
+  const defLine = lines[startLine];
+  const baseIndent = defLine.match(/^(\s*)/)[1].length;
+
+  for (let i = startLine + 1; i < lines.length; i++) {
+    const line = lines[i];
+    if (line.trim() === "") continue; // skip blank lines
+    const indent = line.match(/^(\s*)/)[1].length;
+    if (indent <= baseIndent) {
+      return i - 1;
+    }
+  }
+  return lines.length - 1;
+}
+
+/**
+ * Find block end for Ruby-like (def/end, class/end, module/end).
+ */
+function findRubyBlockEnd(lines, startLine) {
+  let depth = 0;
+  for (let i = startLine; i < lines.length; i++) {
+    const trimmed = lines[i].trim();
+    // Keywords that open blocks
+    if (
+      /^(class|module|def|do|if|unless|case|while|until|for|begin)\b/.test(
+        trimmed,
+      ) ||
+      /\bdo\s*(\|[^|]*\|)?\s*$/.test(trimmed)
+    ) {
+      depth++;
+    }
+    if (/^end\b/.test(trimmed)) {
+      depth--;
+      if (depth === 0) return i;
+    }
+  }
+  return lines.length - 1;
+}
+
+/**
+ * Find block end for Lua (function/end).
+ */
+function findLuaBlockEnd(lines, startLine) {
+  let depth = 0;
+  for (let i = startLine; i < lines.length; i++) {
+    const trimmed = lines[i].trim();
+    if (/\b(function|if|for|while|repeat)\b/.test(trimmed)) depth++;
+    if (
+      /^end\b/.test(trimmed) ||
+      /\bend\s*[,)\]]/.test(trimmed) ||
+      trimmed === "end"
+    ) {
+      depth--;
+      if (depth === 0) return i;
+    }
+  }
+  return lines.length - 1;
+}
+
+function computeByteSize(lines, startLine, endLine) {
+  let size = 0;
+  for (let i = startLine; i <= endLine && i < lines.length; i++) {
+    size += Buffer.byteLength(lines[i], "utf8") + 1; // +1 for newline
+  }
+  return size;
+}
+
+function buildElement(type, label, startLine, endLine, lines) {
+  return {
+    type,
+    label,
+    startLine: startLine + 1, // 1-indexed
+    endLine: endLine + 1,
+    size: computeByteSize(lines, startLine, endLine),
+  };
+}
+
+// --- Language Parsers ---
+
+function parsePython(lines) {
+  const elements = [];
+  for (let i = 0; i < lines.length; i++) {
+    const line = lines[i];
+    const trimmed = line.trim();
+
+    // class
+    let m = trimmed.match(/^class\s+(\w+)(\(.*?\))?\s*:/);
+    if (m) {
+      const end = findIndentBlockEnd(lines, i);
+      elements.push(buildElement("class", `class ${m[1]}`, i, end, lines));
+      continue;
+    }
+
+    // async def
+    m = trimmed.match(/^async\s+def\s+(\w+)\s*\((.*?)\)(\s*->.*?)?\s*:/);
+    if (m) {
+      const end = findIndentBlockEnd(lines, i);
+      const sig = `${m[1]}(${m[2]})${m[3] || ""}`;
+      const type = m[1] === "__init__" ? "ctor" : "async";
+      elements.push(
+        buildElement(
+          type,
+          `${type === "ctor" ? "ctor" : "async"} ${sig}`,
+          i,
+          end,
+          lines,
+        ),
+      );
+      continue;
+    }
+
+    // def
+    m = trimmed.match(/^def\s+(\w+)\s*\((.*?)\)(\s*->.*?)?\s*:/);
+    if (m) {
+      const end = findIndentBlockEnd(lines, i);
+      const sig = `${m[1]}(${m[2]})${m[3] || ""}`;
+      let type = "fn";
+      if (m[1] === "__init__") type = "ctor";
+      else if (m[1].startsWith("test_")) type = "test";
+      const label =
+        type === "ctor"
+          ? `ctor ${sig}`
+          : type === "test"
+            ? `test ${sig}`
+            : `fn ${sig}`;
+      elements.push(buildElement(type, label, i, end, lines));
+      continue;
+    }
+
+    // for/while loops (> 5 lines)
+    m = trimmed.match(/^(for|while)\s+(.+):\s*$/);
+    if (m) {
+      const end = findIndentBlockEnd(lines, i);
+      if (end - i + 1 > 5) {
+        elements.push(
+          buildElement("loop", `loop ${m[1]} ${m[2]}`, i, end, lines),
+        );
+      }
+      continue;
+    }
+  }
+  return elements;
+}
+
+function parseJavaScript(lines) {
+  const elements = [];
+  for (let i = 0; i < lines.length; i++) {
+    const line = lines[i];
+    const trimmed = line.trim();
+
+    // class
+    let m = trimmed.match(/^(export\s+)?(default\s+)?class\s+(\w+)/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("class", `class ${m[3]}`, i, end, lines));
+      continue;
+    }
+
+    // describe (test suite)
+    m = trimmed.match(/^describe\s*\(\s*['"`]([^'"`]+)['"`]/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(
+        buildElement("describe", `describe ${m[1]}`, i, end, lines),
+      );
+      continue;
+    }
+
+    // test/it blocks
+    m = trimmed.match(/^(it|test)\s*\(\s*['"`]([^'"`]+)['"`]/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("test", `test ${m[2]}`, i, end, lines));
+      continue;
+    }
+
+    // async function
+    m = trimmed.match(
+      /^(export\s+)?(default\s+)?async\s+function\s+(\w+)\s*\((.*?)\)/,
+    );
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(
+        buildElement("async", `async ${m[3]}(${m[4]})`, i, end, lines),
+      );
+      continue;
+    }
+
+    // function
+    m = trimmed.match(/^(export\s+)?(default\s+)?function\s+(\w+)\s*\((.*?)\)/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      const type = m[3] === "constructor" ? "ctor" : "fn";
+      elements.push(
+        buildElement(type, `${type} ${m[3]}(${m[4]})`, i, end, lines),
+      );
+      continue;
+    }
+
+    // arrow functions assigned to const/let/var (with explicit function body)
+    m = trimmed.match(
+      /^(export\s+)?(const|let|var)\s+(\w+)\s*=\s*(async\s+)?\(?(.*?)\)?\s*=>/,
+    );
+    if (m && (trimmed.includes("{") || i + 1 < lines.length)) {
+      // Only include if it has a block body
+      if (
+        trimmed.includes("{") ||
+        (i + 1 < lines.length && lines[i + 1].trim().startsWith("{"))
+      ) {
+        const end = findBraceBlockEnd(lines, i);
+        if (end > i) {
+          const isAsync = !!m[4];
+          const type = isAsync ? "async" : "fn";
+          elements.push(
+            buildElement(type, `${type} ${m[3]}(${m[5] || ""})`, i, end, lines),
+          );
+        }
+      }
+      continue;
+    }
+
+    // for/while loops (> 5 lines)
+    m = trimmed.match(/^(for|while)\s*\((.+)\)\s*\{?/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      if (end - i + 1 > 5) {
+        elements.push(
+          buildElement("loop", `loop ${m[1]} ${m[2]}`, i, end, lines),
+        );
+      }
+      continue;
+    }
+  }
+  return elements;
+}
+
+function parseTypeScript(lines) {
+  const elements = [];
+  for (let i = 0; i < lines.length; i++) {
+    const line = lines[i];
+    const trimmed = line.trim();
+
+    // interface
+    let m = trimmed.match(/^(export\s+)?(default\s+)?interface\s+(\w+)/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("class", `interface ${m[3]}`, i, end, lines));
+      continue;
+    }
+
+    // class
+    m = trimmed.match(/^(export\s+)?(default\s+)?(abstract\s+)?class\s+(\w+)/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("class", `class ${m[4]}`, i, end, lines));
+      continue;
+    }
+
+    // describe
+    m = trimmed.match(/^describe\s*\(\s*['"`]([^'"`]+)['"`]/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(
+        buildElement("describe", `describe ${m[1]}`, i, end, lines),
+      );
+      continue;
+    }
+
+    // test/it
+    m = trimmed.match(/^(it|test)\s*\(\s*['"`]([^'"`]+)['"`]/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("test", `test ${m[2]}`, i, end, lines));
+      continue;
+    }
+
+    // async function
+    m = trimmed.match(
+      /^(export\s+)?(default\s+)?async\s+function\s+(\w+)\s*\((.*?)\)/,
+    );
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(
+        buildElement("async", `async ${m[3]}(${m[4]})`, i, end, lines),
+      );
+      continue;
+    }
+
+    // function
+    m = trimmed.match(/^(export\s+)?(default\s+)?function\s+(\w+)\s*\((.*?)\)/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("fn", `fn ${m[3]}(${m[4]})`, i, end, lines));
+      continue;
+    }
+
+    // arrow functions
+    m = trimmed.match(
+      /^(export\s+)?(const|let|var)\s+(\w+)\s*=\s*(async\s+)?\(?(.*?)\)?\s*=>/,
+    );
+    if (m && trimmed.includes("{")) {
+      const end = findBraceBlockEnd(lines, i);
+      if (end > i) {
+        const isAsync = !!m[4];
+        const type = isAsync ? "async" : "fn";
+        elements.push(
+          buildElement(type, `${type} ${m[3]}(${m[5] || ""})`, i, end, lines),
+        );
+      }
+      continue;
+    }
+
+    // for/while loops (> 5 lines)
+    m = trimmed.match(/^(for|while)\s*\((.+)\)\s*\{?/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      if (end - i + 1 > 5) {
+        elements.push(
+          buildElement("loop", `loop ${m[1]} ${m[2]}`, i, end, lines),
+        );
+      }
+    }
+  }
+  return elements;
+}
+
+function parseGo(lines) {
+  const elements = [];
+  for (let i = 0; i < lines.length; i++) {
+    const trimmed = lines[i].trim();
+
+    // struct
+    let m = trimmed.match(/^type\s+(\w+)\s+struct\b/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("class", `struct ${m[1]}`, i, end, lines));
+      continue;
+    }
+
+    // interface
+    m = trimmed.match(/^type\s+(\w+)\s+interface\b/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("class", `interface ${m[1]}`, i, end, lines));
+      continue;
+    }
+
+    // func
+    m = trimmed.match(/^func\s+(\(.*?\)\s*)?(\w+)\s*\((.*?)\)/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      const receiver = m[1] ? m[1].trim() + " " : "";
+      const name = m[2];
+      const type = name.startsWith("Test") ? "test" : "fn";
+      elements.push(
+        buildElement(
+          type,
+          `${type === "test" ? "test" : "fn"} ${receiver}${name}(${m[3]})`,
+          i,
+          end,
+          lines,
+        ),
+      );
+      continue;
+    }
+
+    // for loops (> 5 lines) - Go only has for
+    m = trimmed.match(/^for\s+(.+)\s*\{/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      if (end - i + 1 > 5) {
+        elements.push(buildElement("loop", `loop for ${m[1]}`, i, end, lines));
+      }
+    }
+  }
+  return elements;
+}
+
+function parseRust(lines) {
+  const elements = [];
+  for (let i = 0; i < lines.length; i++) {
+    const trimmed = lines[i].trim();
+
+    // struct
+    let m = trimmed.match(/^(pub\s+)?struct\s+(\w+)/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("class", `struct ${m[2]}`, i, end, lines));
+      continue;
+    }
+
+    // enum
+    m = trimmed.match(/^(pub\s+)?enum\s+(\w+)/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("class", `enum ${m[2]}`, i, end, lines));
+      continue;
+    }
+
+    // trait
+    m = trimmed.match(/^(pub\s+)?trait\s+(\w+)/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("class", `trait ${m[2]}`, i, end, lines));
+      continue;
+    }
+
+    // impl
+    m = trimmed.match(/^impl\s+(.+?)\s*\{/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("impl", `impl ${m[1]}`, i, end, lines));
+      continue;
+    }
+
+    // fn
+    m = trimmed.match(/^(pub\s+)?(async\s+)?fn\s+(\w+)\s*\((.*?)\)/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      const isAsync = !!m[2];
+      const isTest = m[3].startsWith("test_");
+      const type = isTest ? "test" : isAsync ? "async" : "fn";
+      elements.push(
+        buildElement(type, `${type} ${m[3]}(${m[4]})`, i, end, lines),
+      );
+      continue;
+    }
+
+    // loop/for/while (> 5 lines)
+    m = trimmed.match(/^(for|while|loop)\b(.*)?\{/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      if (end - i + 1 > 5) {
+        elements.push(
+          buildElement(
+            "loop",
+            `loop ${m[1]}${m[2] ? " " + m[2].trim() : ""}`,
+            i,
+            end,
+            lines,
+          ),
+        );
+      }
+    }
+  }
+  return elements;
+}
+
+function parseJava(lines) {
+  const elements = [];
+  for (let i = 0; i < lines.length; i++) {
+    const trimmed = lines[i].trim();
+
+    // class
+    let m = trimmed.match(
+      /^(public\s+|private\s+|protected\s+)?(static\s+)?(abstract\s+)?(final\s+)?class\s+(\w+)/,
+    );
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("class", `class ${m[5]}`, i, end, lines));
+      continue;
+    }
+
+    // interface
+    m = trimmed.match(/^(public\s+|private\s+|protected\s+)?interface\s+(\w+)/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("class", `interface ${m[2]}`, i, end, lines));
+      continue;
+    }
+
+    // enum
+    m = trimmed.match(/^(public\s+|private\s+|protected\s+)?enum\s+(\w+)/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("class", `enum ${m[2]}`, i, end, lines));
+      continue;
+    }
+
+    // method (including constructors)
+    m = trimmed.match(
+      /^(public\s+|private\s+|protected\s+)?(static\s+)?(abstract\s+)?(final\s+)?(synchronized\s+)?(\w+\s+)?(\w+)\s*\((.*?)\)\s*(\{|throws)/,
+    );
+    if (
+      m &&
+      !["if", "for", "while", "switch", "catch", "return"].includes(m[7])
+    ) {
+      const end = findBraceBlockEnd(lines, i);
+      const name = m[7];
+      // Constructor: return type is absent and name matches class-like pattern
+      const hasReturnType = m[6] && m[6].trim();
+      const type = !hasReturnType
+        ? "ctor"
+        : name.startsWith("test")
+          ? "test"
+          : "fn";
+      elements.push(
+        buildElement(type, `${type} ${name}(${m[8]})`, i, end, lines),
+      );
+      continue;
+    }
+
+    // for/while loops (> 5 lines)
+    m = trimmed.match(/^(for|while)\s*\((.+)\)\s*\{?/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      if (end - i + 1 > 5) {
+        elements.push(
+          buildElement("loop", `loop ${m[1]} ${m[2]}`, i, end, lines),
+        );
+      }
+    }
+  }
+  return elements;
+}
+
+function parseCCpp(lines) {
+  const elements = [];
+  for (let i = 0; i < lines.length; i++) {
+    const trimmed = lines[i].trim();
+
+    // class
+    let m = trimmed.match(/^class\s+(\w+)/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("class", `class ${m[1]}`, i, end, lines));
+      continue;
+    }
+
+    // struct
+    m = trimmed.match(/^(typedef\s+)?struct\s+(\w+)/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("class", `struct ${m[2]}`, i, end, lines));
+      continue;
+    }
+
+    // function (C-style: return_type name(...))
+    m = trimmed.match(/^(\w[\w\s*&]+?)\s+(\w+)\s*\(([^)]*)\)\s*(\{|$)/);
+    if (
+      m &&
+      ![
+        "if",
+        "for",
+        "while",
+        "switch",
+        "return",
+        "typedef",
+        "struct",
+        "class",
+        "enum",
+      ].includes(m[2])
+    ) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("fn", `fn ${m[2]}(${m[3]})`, i, end, lines));
+      continue;
+    }
+
+    // for/while loops (> 5 lines)
+    m = trimmed.match(/^(for|while)\s*\((.+)\)\s*\{?/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      if (end - i + 1 > 5) {
+        elements.push(
+          buildElement("loop", `loop ${m[1]} ${m[2]}`, i, end, lines),
+        );
+      }
+    }
+  }
+  return elements;
+}
+
+function parseCSharp(lines) {
+  const elements = [];
+  for (let i = 0; i < lines.length; i++) {
+    const trimmed = lines[i].trim();
+
+    // class / struct / interface / enum / record
+    let m = trimmed.match(
+      /^(public\s+|private\s+|protected\s+|internal\s+)?(static\s+)?(abstract\s+|sealed\s+)?(partial\s+)?(class|struct|interface|enum|record)\s+(\w+)/,
+    );
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("class", `${m[5]} ${m[6]}`, i, end, lines));
+      continue;
+    }
+
+    // method
+    m = trimmed.match(
+      /^(public\s+|private\s+|protected\s+|internal\s+)?(static\s+)?(async\s+)?(virtual\s+|override\s+|abstract\s+)?(\w[\w<>\[\],\s]*?)\s+(\w+)\s*\((.*?)\)\s*\{?/,
+    );
+    if (
+      m &&
+      ![
+        "if",
+        "for",
+        "while",
+        "switch",
+        "catch",
+        "return",
+        "class",
+        "struct",
+        "interface",
+        "enum",
+      ].includes(m[6])
+    ) {
+      const end = findBraceBlockEnd(lines, i);
+      const isAsync = !!m[3];
+      const type = isAsync ? "async" : "fn";
+      elements.push(
+        buildElement(type, `${type} ${m[6]}(${m[7]})`, i, end, lines),
+      );
+      continue;
+    }
+
+    // for/while loops (> 5 lines)
+    m = trimmed.match(/^(for|foreach|while)\s*\((.+)\)\s*\{?/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      if (end - i + 1 > 5) {
+        elements.push(
+          buildElement("loop", `loop ${m[1]} ${m[2]}`, i, end, lines),
+        );
+      }
+    }
+  }
+  return elements;
+}
+
+function parsePHP(lines) {
+  const elements = [];
+  for (let i = 0; i < lines.length; i++) {
+    const trimmed = lines[i].trim();
+
+    // class / interface / trait
+    let m = trimmed.match(
+      /^(abstract\s+)?(final\s+)?(class|interface|trait)\s+(\w+)/,
+    );
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("class", `${m[3]} ${m[4]}`, i, end, lines));
+      continue;
+    }
+
+    // function
+    m = trimmed.match(
+      /^(public\s+|private\s+|protected\s+)?(static\s+)?function\s+(\w+)\s*\((.*?)\)/,
+    );
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      const type =
+        m[3] === "__construct"
+          ? "ctor"
+          : m[3].startsWith("test")
+            ? "test"
+            : "fn";
+      elements.push(
+        buildElement(type, `${type} ${m[3]}(${m[4]})`, i, end, lines),
+      );
+      continue;
+    }
+
+    // for/while loops (> 5 lines)
+    m = trimmed.match(/^(for|foreach|while)\s*\((.+)\)\s*\{?/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      if (end - i + 1 > 5) {
+        elements.push(
+          buildElement("loop", `loop ${m[1]} ${m[2]}`, i, end, lines),
+        );
+      }
+    }
+  }
+  return elements;
+}
+
+function parseRuby(lines) {
+  const elements = [];
+  for (let i = 0; i < lines.length; i++) {
+    const trimmed = lines[i].trim();
+
+    // class
+    let m = trimmed.match(/^class\s+(\w+)/);
+    if (m) {
+      const end = findRubyBlockEnd(lines, i);
+      elements.push(buildElement("class", `class ${m[1]}`, i, end, lines));
+      continue;
+    }
+
+    // module
+    m = trimmed.match(/^module\s+(\w+)/);
+    if (m) {
+      const end = findRubyBlockEnd(lines, i);
+      elements.push(buildElement("class", `module ${m[1]}`, i, end, lines));
+      continue;
+    }
+
+    // def
+    m = trimmed.match(/^def\s+(self\.)?(\w+[?!=]?)\s*(\(.*?\))?/);
+    if (m) {
+      const end = findRubyBlockEnd(lines, i);
+      const prefix = m[1] || "";
+      const type =
+        m[2] === "initialize"
+          ? "ctor"
+          : m[2].startsWith("test_")
+            ? "test"
+            : "fn";
+      elements.push(
+        buildElement(
+          type,
+          `${type} ${prefix}${m[2]}${m[3] || ""}`,
+          i,
+          end,
+          lines,
+        ),
+      );
+      continue;
+    }
+  }
+  return elements;
+}
+
+function parseSwift(lines) {
+  const elements = [];
+  for (let i = 0; i < lines.length; i++) {
+    const trimmed = lines[i].trim();
+
+    // class / struct / enum / protocol
+    let m = trimmed.match(
+      /^(public\s+|private\s+|internal\s+|open\s+|fileprivate\s+)?(final\s+)?(class|struct|enum|protocol)\s+(\w+)/,
+    );
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("class", `${m[3]} ${m[4]}`, i, end, lines));
+      continue;
+    }
+
+    // func
+    m = trimmed.match(
+      /^(public\s+|private\s+|internal\s+|open\s+)?(static\s+|class\s+)?(override\s+)?func\s+(\w+)\s*\((.*?)\)/,
+    );
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      const name = m[4];
+      const type =
+        name === "init" ? "ctor" : name.startsWith("test") ? "test" : "fn";
+      elements.push(
+        buildElement(type, `${type} ${name}(${m[5]})`, i, end, lines),
+      );
+      continue;
+    }
+
+    // for/while loops (> 5 lines)
+    m = trimmed.match(/^(for|while)\s+(.+)\s*\{/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      if (end - i + 1 > 5) {
+        elements.push(
+          buildElement("loop", `loop ${m[1]} ${m[2]}`, i, end, lines),
+        );
+      }
+    }
+  }
+  return elements;
 }
 
+function parseKotlin(lines) {
+  const elements = [];
+  for (let i = 0; i < lines.length; i++) {
+    const trimmed = lines[i].trim();
+
+    // class / interface / object
+    let m = trimmed.match(
+      /^(open\s+|abstract\s+|data\s+|sealed\s+)?(class|interface|object)\s+(\w+)/,
+    );
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("class", `${m[2]} ${m[3]}`, i, end, lines));
+      continue;
+    }
+
+    // fun
+    m = trimmed.match(
+      /^(public\s+|private\s+|protected\s+|internal\s+)?(override\s+)?(suspend\s+)?fun\s+(\w+)\s*\((.*?)\)/,
+    );
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      const isSuspend = !!m[3];
+      const type = m[4].startsWith("test")
+        ? "test"
+        : isSuspend
+          ? "async"
+          : "fn";
+      elements.push(
+        buildElement(type, `${type} ${m[4]}(${m[5]})`, i, end, lines),
+      );
+      continue;
+    }
+
+    // for/while loops (> 5 lines)
+    m = trimmed.match(/^(for|while)\s*\((.+)\)\s*\{?/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      if (end - i + 1 > 5) {
+        elements.push(
+          buildElement("loop", `loop ${m[1]} ${m[2]}`, i, end, lines),
+        );
+      }
+    }
+  }
+  return elements;
+}
+
+function parseScala(lines) {
+  const elements = [];
+  for (let i = 0; i < lines.length; i++) {
+    const trimmed = lines[i].trim();
+
+    // class / object / trait
+    let m = trimmed.match(/^(case\s+)?(class|object|trait)\s+(\w+)/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(
+        buildElement("class", `${m[1] || ""}${m[2]} ${m[3]}`, i, end, lines),
+      );
+      continue;
+    }
+
+    // def
+    m = trimmed.match(/^(override\s+)?def\s+(\w+)\s*(\(.*?\))?/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      const type = m[2].startsWith("test") ? "test" : "fn";
+      elements.push(
+        buildElement(type, `${type} ${m[2]}${m[3] || ""}`, i, end, lines),
+      );
+      continue;
+    }
+  }
+  return elements;
+}
+
+function parseLua(lines) {
+  const elements = [];
+  for (let i = 0; i < lines.length; i++) {
+    const trimmed = lines[i].trim();
+
+    // function / local function
+    let m = trimmed.match(/^(local\s+)?function\s+([\w.:]+)\s*\((.*?)\)/);
+    if (m) {
+      const end = findLuaBlockEnd(lines, i);
+      elements.push(buildElement("fn", `fn ${m[2]}(${m[3]})`, i, end, lines));
+      continue;
+    }
+
+    // for/while loops (> 5 lines)
+    m = trimmed.match(/^(for|while)\s+(.+)\s+do/);
+    if (m) {
+      const end = findLuaBlockEnd(lines, i);
+      if (end - i + 1 > 5) {
+        elements.push(
+          buildElement("loop", `loop ${m[1]} ${m[2]}`, i, end, lines),
+        );
+      }
+    }
+  }
+  return elements;
+}
+
+function parsePerl(lines) {
+  const elements = [];
+  for (let i = 0; i < lines.length; i++) {
+    const trimmed = lines[i].trim();
+
+    // package
+    let m = trimmed.match(/^package\s+([\w:]+)/);
+    if (m) {
+      elements.push(buildElement("class", `package ${m[1]}`, i, i, lines));
+      continue;
+    }
+
+    // sub
+    m = trimmed.match(/^sub\s+(\w+)/);
+    if (m) {
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("fn", `fn ${m[1]}`, i, end, lines));
+      continue;
+    }
+  }
+  return elements;
+}
+
+function parseBash(lines) {
+  const elements = [];
+  for (let i = 0; i < lines.length; i++) {
+    const trimmed = lines[i].trim();
+
+    // function keyword or name()
+    let m = trimmed.match(/^(function\s+)?(\w+)\s*\(\s*\)\s*\{?/);
+    if (m && m[1]) {
+      // function keyword form
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("fn", `fn ${m[2]}`, i, end, lines));
+      continue;
+    }
+    if (m && !m[1] && trimmed.includes("()")) {
+      // name() form
+      const end = findBraceBlockEnd(lines, i);
+      elements.push(buildElement("fn", `fn ${m[2]}`, i, end, lines));
+      continue;
+    }
+
+    // for/while loops (> 5 lines)
+    m = trimmed.match(/^(for|while)\s+(.+?);\s*do/);
+    if (!m) m = trimmed.match(/^(for|while)\s+(.+)/);
+    if (m) {
+      // Look for done
+      let end = i;
+      for (let j = i + 1; j < lines.length; j++) {
+        if (lines[j].trim() === "done") {
+          end = j;
+          break;
+        }
+      }
+      if (end - i + 1 > 5) {
+        elements.push(
+          buildElement("loop", `loop ${m[1]} ${m[2]}`, i, end, lines),
+        );
+      }
+    }
+  }
+  return elements;
+}
+
+// ---------------------------------------------------------------------------
+// Nesting + Tree Building
+// ---------------------------------------------------------------------------
+
 /**
- * Recursively walks directories, respecting .gitignore files at each level.
+ * Nest flat elements into a tree based on line ranges.
+ * Elements that fall within the range of a parent become children.
  */
+function nestElements(elements) {
+  if (!elements.length) return [];
+
+  // Sort by start line, then by larger range first (parents before children)
+  const sorted = [...elements].sort((a, b) => {
+    if (a.startLine !== b.startLine) return a.startLine - b.startLine;
+    return b.endLine - b.startLine - (a.endLine - a.startLine);
+  });
+
+  const root = [];
+  const stack = []; // stack of { element, children }
+
+  for (const el of sorted) {
+    const node = { ...el, children: [] };
+
+    // Pop from stack if current element is outside of parent's range
+    while (stack.length > 0) {
+      const parent = stack[stack.length - 1];
+      if (el.startLine >= parent.startLine && el.endLine <= parent.endLine) {
+        break;
+      }
+      stack.pop();
+    }
+
+    if (stack.length > 0) {
+      stack[stack.length - 1].children.push(node);
+    } else {
+      root.push(node);
+    }
+
+    stack.push(node);
+  }
+
+  return root;
+}
+
+// ---------------------------------------------------------------------------
+// Directory Walker (unchanged from v1)
+// ---------------------------------------------------------------------------
+
 function walkDirectory(
   currentDir,
   rootDir,
@@ -60,12 +1145,11 @@ function walkDirectory(
   allowedExts,
   absoluteOutputPath,
   useGitIgnore,
-  stats // { scanned: 0, ignored: 0 }
+  stats,
 ) {
   let results = [];
   let currentIgnoreManager = null;
 
-  // 1. Check for local .gitignore and add to chain for this scope
   if (useGitIgnore) {
     const gitignorePath = path.join(currentDir, ".gitignore");
     if (fs.existsSync(gitignorePath)) {
@@ -73,13 +1157,10 @@ function walkDirectory(
         const content = fs.readFileSync(gitignorePath, "utf8");
         const ig = ignore().add(content);
         currentIgnoreManager = { manager: ig, root: currentDir };
-      } catch (e) {
-        // Warning could go here
-      }
+      } catch (e) {}
     }
   }
 
-  // Create a new chain for this directory and its children
   const nextIgnoreChain = currentIgnoreManager
     ? [...ignoreChain, currentIgnoreManager]
     : ignoreChain;
@@ -94,25 +1175,17 @@ function walkDirectory(
   for (const entry of entries) {
     const fullPath = path.join(currentDir, entry.name);
 
-    // SKIP CHECK: Output file
     if (path.resolve(fullPath) === absoluteOutputPath) continue;
 
-    // SKIP CHECK: Ignore Chain
     let shouldIgnore = false;
     for (const item of nextIgnoreChain) {
-      // Calculate path relative to the specific ignore manager's root
-      // IMPORTANT: Normalize to POSIX slashes for 'ignore' package compatibility
       let relToIgnoreRoot = path.relative(item.root, fullPath);
-
       if (path.sep === "\\") {
         relToIgnoreRoot = relToIgnoreRoot.replace(/\\/g, "/");
       }
-
-      // If checking a directory, ensure trailing slash for proper 'ignore' directory matching
       if (entry.isDirectory() && !relToIgnoreRoot.endsWith("/")) {
         relToIgnoreRoot += "/";
       }
-
       if (item.manager.ignores(relToIgnoreRoot)) {
         shouldIgnore = true;
         break;
@@ -125,7 +1198,6 @@ function walkDirectory(
     }
 
     if (entry.isDirectory()) {
-      // Recurse
       results = results.concat(
         walkDirectory(
           fullPath,
@@ -134,22 +1206,18 @@ function walkDirectory(
           allowedExts,
           absoluteOutputPath,
           useGitIgnore,
-          stats
-        )
+          stats,
+        ),
       );
     } else if (entry.isFile()) {
-      // SKIP CHECK: Binary
       if (isLikelyBinary(fullPath)) {
         stats.ignored++;
         continue;
       }
-
-      // SKIP CHECK: Extensions
       if (allowedExts && !allowedExts.has(path.extname(entry.name))) {
         stats.ignored++;
         continue;
       }
-
       try {
         const fileStats = fs.statSync(fullPath);
         const relativeToRoot = path.relative(rootDir, fullPath);
@@ -160,62 +1228,175 @@ function walkDirectory(
           size: fileStats.size,
           formattedSize: formatBytes(fileStats.size),
         });
-      } catch (e) {
-        // Skip inaccessible files
-      }
+      } catch (e) {}
     }
   }
 
   return results;
 }
 
-function generateFileTree(filesWithSize, root, skipContentSet = null) {
+// ---------------------------------------------------------------------------
+// Code Index Tree Generator (v2.0.0)
+// ---------------------------------------------------------------------------
+
+/**
+ * Build the <code_index> tree with expanded code elements.
+ */
+function generateCodeIndex(filesWithMeta, root, skipContentSet, noParse) {
   let tree = `${path.basename(root)}/\n`;
-  const structure = {};
 
-  filesWithSize.forEach(({ relativePath, formattedSize }) => {
-    const parts = relativePath.split(path.sep);
+  // Build directory structure
+  const structure = {};
+  for (const file of filesWithMeta) {
+    const parts = file.relativePath.split(path.sep);
     let currentLevel = structure;
-    parts.forEach((part, index) => {
-      const isFile = index === parts.length - 1;
+    for (let i = 0; i < parts.length; i++) {
+      const part = parts[i];
+      const isFile = i === parts.length - 1;
       if (isFile) {
-        const shouldSkipContent =
-          skipContentSet && skipContentSet.has(relativePath);
-        currentLevel[part] = {
-          size: formattedSize,
-          skipContent: shouldSkipContent,
-        };
+        currentLevel[part] = { __file: file };
       } else {
-        if (!currentLevel[part]) {
-          currentLevel[part] = {};
-        }
+        if (!currentLevel[part]) currentLevel[part] = {};
         currentLevel = currentLevel[part];
       }
-    });
-  });
+    }
+  }
 
-  const buildTree = (level, prefix) => {
-    const entries = Object.keys(level);
-    entries.forEach((entry, index) => {
-      const isLast = index === entries.length - 1;
-      const value = level[entry];
-      const isFile = typeof value === "object" && value.size !== undefined;
-      const connector = isLast ? "└── " : "├── ";
+  function renderTree(level, prefix) {
+    const keys = Object.keys(level);
+    for (let i = 0; i < keys.length; i++) {
+      const key = keys[i];
+      const isLast = i === keys.length - 1;
+      const connector = isLast ? "\u2514\u2500\u2500 " : "\u251c\u2500\u2500 ";
+      const childPrefix = prefix + (isLast ? "    " : "\u2502   ");
+      const value = level[key];
 
-      if (isFile) {
-        const marker = value.skipContent ? " (content omitted)" : "";
-        tree += `${prefix}${connector}[${value.size}] ${entry}${marker}\n`;
+      if (value.__file) {
+        const f = value.__file;
+        const olRange = `OL: 1-${f.lineCount}`;
+        const mlRange = `ML: ${f.mlStart}-${f.mlEnd}`;
+        const sizeStr = f.formattedSize;
+        const isSkipped = skipContentSet && skipContentSet.has(f.relativePath);
+
+        tree += `${prefix}${connector}${key} [${olRange} | ${mlRange} | ${sizeStr}]\n`;
+
+        if (isSkipped) {
+          tree += `${childPrefix}(Content omitted - file size: ${sizeStr})\n`;
+        } else if (!noParse && f.codeElements && f.codeElements.length > 0) {
+          renderCodeElements(f.codeElements, childPrefix, f.mlStart);
+        }
       } else {
-        tree += `${prefix}${connector}${entry}\n`;
-        buildTree(value, `${prefix}${isLast ? "    " : "│   "}`);
+        // Directory
+        tree += `${prefix}${connector}${key}/\n`;
+        renderTree(value, childPrefix);
       }
-    });
-  };
+    }
+  }
+
+  function renderCodeElements(elements, prefix, mlOffset) {
+    for (let i = 0; i < elements.length; i++) {
+      const el = elements[i];
+      const isLast = i === elements.length - 1;
+      const connector = isLast ? "\u2514\u2500\u2500 " : "\u251c\u2500\u2500 ";
+      const childPrefix = prefix + (isLast ? "    " : "\u2502   ");
+
+      const olRange = `OL: ${el.startLine}-${el.endLine}`;
+      const mlStart = mlOffset + el.startLine - 1;
+      const mlEnd = mlOffset + el.endLine - 1;
+      const mlRange = `ML: ${mlStart}-${mlEnd}`;
+      const sizeStr = formatBytes(el.size);
+
+      tree += `${prefix}${connector}${el.label} [${olRange} | ${mlRange} | ${sizeStr}]\n`;
+
+      if (el.children && el.children.length > 0) {
+        renderCodeElements(el.children, childPrefix, mlOffset);
+      }
+    }
+  }
 
-  buildTree(structure, "");
+  renderTree(structure, "");
   return tree;
 }
 
+// ---------------------------------------------------------------------------
+// Recreate functionality (v2.0.0)
+// ---------------------------------------------------------------------------
+
+function recreateFromFile(inputFile, outputDir, dryRun, overwrite) {
+  if (!fs.existsSync(inputFile)) {
+    console.error(`\n\u274c Input file not found: ${inputFile}`);
+    process.exit(1);
+  }
+
+  const content = fs.readFileSync(inputFile, "utf8");
+
+  // Parse files from the merged_code section (or legacy format)
+  const files = [];
+  // Match: # FILE: path [...]  followed by ```` block
+  const fileRegex = /# FILE:\s*(.+?)\s*\[.*?\]\n````\n([\s\S]*?)\n````/g;
+  let match;
+
+  while ((match = fileRegex.exec(content)) !== null) {
+    const filePath = match[1].trim();
+    const fileContent = match[2];
+
+    // Skip content-omitted files
+    if (fileContent.trim().startsWith("(Content omitted")) continue;
+
+    files.push({ path: filePath, content: fileContent });
+  }
+
+  if (files.length === 0) {
+    // Try legacy format: ### **FILE:** `path`
+    const legacyRegex = /### \*\*FILE:\*\*\s*`(.+?)`\n````\n([\s\S]*?)\n````/g;
+    while ((match = legacyRegex.exec(content)) !== null) {
+      const filePath = match[1].trim();
+      const fileContent = match[2];
+      if (fileContent.trim().startsWith("(Content omitted")) continue;
+      files.push({ path: filePath, content: fileContent });
+    }
+  }
+
+  if (files.length === 0) {
+    console.error("\n\u274c No files found in the input file.");
+    process.exit(1);
+  }
+
+  const resolvedOutputDir = path.resolve(outputDir);
+  console.log(`\n\ud83d\udcc2 Output directory: ${resolvedOutputDir}\n`);
+
+  let totalSize = 0;
+
+  for (const file of files) {
+    const fullPath = path.join(resolvedOutputDir, file.path);
+    const size = Buffer.byteLength(file.content, "utf8");
+    totalSize += size;
+
+    console.log(`   ${file.path} (${formatBytes(size)})`);
+
+    if (!dryRun) {
+      if (fs.existsSync(fullPath) && !overwrite) {
+        console.error(`   \u26a0\ufe0f  Skipped (exists): ${file.path}`);
+        continue;
+      }
+      const dir = path.dirname(fullPath);
+      fs.mkdirSync(dir, { recursive: true });
+      fs.writeFileSync(fullPath, file.content, "utf8");
+    }
+  }
+
+  console.log(`\n\ud83d\udcca Summary:`);
+  console.log(
+    `   \u2022 ${dryRun ? "Files to recreate" : "Files recreated"}: ${files.length}`,
+  );
+  console.log(`   \u2022 Total size: ${formatBytes(totalSize)}`);
+  console.log(`\n\u2705 Done!`);
+}
+
+// ---------------------------------------------------------------------------
+// Main
+// ---------------------------------------------------------------------------
+
 async function main() {
   const rawArgv = hideBin(process.argv);
   if (rawArgv.includes("--version") || rawArgv.includes("-v")) {
@@ -228,13 +1409,13 @@ async function main() {
     .usage("$0 [options]")
     .option("o", {
       alias: "output",
-      describe: "Output file name",
+      describe: "Output file (combine) or directory (recreate)",
       type: "string",
       default: "combicode.txt",
     })
     .option("d", {
       alias: "dry-run",
-      describe: "Preview files without creating the output file",
+      describe: "Preview without making changes",
       type: "boolean",
       default: false,
     })
@@ -254,41 +1435,81 @@ async function main() {
       type: "boolean",
       default: false,
     })
-    .option("no-gitignore", {
-      describe: "Ignore the project's .gitignore file",
+    .option("gitignore", {
+      describe: "Use patterns from the project's .gitignore file",
       type: "boolean",
-      default: false,
+      default: true,
     })
-    .option("no-header", {
-      describe: "Omit the introductory prompt and file tree from the output",
+    .option("header", {
+      describe: "Include the introductory prompt and code index in the output",
       type: "boolean",
-      default: false,
+      default: true,
     })
     .option("skip-content", {
       describe:
         "Comma-separated glob patterns for files to include in tree but omit content",
       type: "string",
     })
+    .option("parse", {
+      describe: "Enable code structure parsing",
+      type: "boolean",
+      default: true,
+    })
+    .option("r", {
+      alias: "recreate",
+      describe: "Recreate project from a combicode.txt file",
+      type: "boolean",
+      default: false,
+    })
+    .option("input", {
+      describe: "Input combicode.txt file for recreate",
+      type: "string",
+      default: "combicode.txt",
+    })
+    .option("overwrite", {
+      describe: "Overwrite existing files when recreating",
+      type: "boolean",
+      default: false,
+    })
     .version(version)
     .alias("v", "version")
     .help()
     .alias("h", "help").argv;
 
   const projectRoot = process.cwd();
-  console.log(`\n✨ Combicode v${version}`);
-  console.log(`📂 Root: ${projectRoot}`);
+  console.log(`\u2728 Combicode v${version}`);
+  console.log(`\ud83d\udcc2 Root: ${projectRoot}`);
 
-  const rootIgnoreManager = ignore();
+  // --- Recreate mode ---
+  if (argv.recreate) {
+    const inputFile = path.resolve(projectRoot, argv.input);
+    const outputDir =
+      argv.output !== "combicode.txt" ? argv.output : projectRoot;
+    recreateFromFile(inputFile, outputDir, argv.dryRun, argv.overwrite);
+    return;
+  }
 
-  // Only add minimal safety ignores + CLI excludes.
-  // No external JSON config is loaded.
+  // --- Combine mode ---
+  const rootIgnoreManager = ignore();
   rootIgnoreManager.add(SAFETY_IGNORES);
 
   if (argv.exclude) {
     rootIgnoreManager.add(argv.exclude.split(","));
   }
 
-  // Create skip-content manager
+  // Parse .gitmodules for submodule paths
+  const gitModulesPath = path.join(projectRoot, ".gitmodules");
+  if (fs.existsSync(gitModulesPath)) {
+    try {
+      const content = fs.readFileSync(gitModulesPath, "utf8");
+      const lines = content.split(/\r?\n/);
+      for (const line of lines) {
+        const m = line.match(/^\s*path\s*=\s*(.+?)\s*$/);
+        if (m) rootIgnoreManager.add([m[1]]);
+      }
+    } catch (e) {}
+  }
+
   const skipContentManager = ignore();
   if (argv.skipContent) {
     skipContentManager.add(argv.skipContent.split(","));
@@ -300,30 +1521,26 @@ async function main() {
     ? new Set(
         argv.includeExt
           .split(",")
-          .map((ext) => (ext.startsWith(".") ? ext : `.${ext}`))
+          .map((ext) => (ext.startsWith(".") ? ext : `.${ext}`)),
       )
     : null;
 
-  // Initialize the ignore chain with the root manager
   const ignoreChain = [{ manager: rootIgnoreManager, root: projectRoot }];
-
-  // Statistics container
   const stats = { scanned: 0, ignored: 0 };
 
-  // Perform Recursive Walk
   const includedFiles = walkDirectory(
     projectRoot,
     projectRoot,
     ignoreChain,
     allowedExtensions,
     absoluteOutputPath,
-    !argv.noGitignore,
-    stats
+    argv.gitignore,
+    stats,
   );
 
   includedFiles.sort((a, b) => a.path.localeCompare(b.path));
 
-  // Determine which files should have content skipped
+  // Determine skip-content set
   const skipContentSet = new Set();
   if (argv.skipContent) {
     includedFiles.forEach((file) => {
@@ -334,96 +1551,219 @@ async function main() {
     });
   }
 
-  // Calculate total size of included files (excluding files with skipped content)
-  const totalSizeBytes = includedFiles.reduce((acc, file) => {
-    // Don't count files with skipped content in the total size
-    if (skipContentSet.has(file.relativePath)) {
-      return acc;
-    }
-    return acc + file.size;
-  }, 0);
-
   if (includedFiles.length === 0) {
     console.error(
-      "\n❌ No files to include. Check your path, .gitignore, or filters."
+      "\n\u274c No files to include. Check your path, .gitignore, or filters.",
     );
     process.exit(1);
   }
 
+  // --- Read file contents & parse code structure ---
+  for (const fileObj of includedFiles) {
+    const isSkipped = skipContentSet.has(fileObj.relativePath);
+    if (isSkipped) {
+      fileObj.content = null;
+      fileObj.lineCount = 0;
+      fileObj.codeElements = [];
+    } else {
+      try {
+        fileObj.content = fs.readFileSync(fileObj.path, "utf8");
+        // Count actual lines: strings ending with \n get an extra empty element from split
+        const parts = fileObj.content.split("\n");
+        fileObj.lineCount = fileObj.content.endsWith("\n")
+          ? parts.length - 1
+          : parts.length;
+
+        if (argv.parse) {
+          const flat = parseCodeStructure(
+            fileObj.relativePath,
+            fileObj.content,
+          );
+          fileObj.codeElements = nestElements(flat);
+        } else {
+          fileObj.codeElements = [];
+        }
+      } catch (e) {
+        fileObj.content = `... (error reading file: ${e.message}) ...`;
+        fileObj.lineCount = 1;
+        fileObj.codeElements = [];
+      }
+    }
+  }
+
+  // --- Calculate ML (Merged Line) offsets ---
+  // First pass: figure out how many header lines come before merged_code
+  const systemPrompt = argv.llmsTxt
+    ? LLMS_TXT_SYSTEM_PROMPT
+    : DEFAULT_SYSTEM_PROMPT;
+  let headerLines = 0;
+  if (argv.header) {
+    headerLines += systemPrompt.split("\n").length; // system prompt
+    headerLines += 1; // blank line after prompt
+    // <code_index> section will be added but we need to calculate it after ML offsets are known
+    // We'll do a two-pass approach
+  }
+
+  // Two-pass: first compute code_index size, then compute ML offsets
+  // Pass 1: Assign provisional ML offsets (we'll need the code_index line count first)
+  // The structure is: header + code_index_block + merged_code_block
+
+  // Compute provisional line count for each file in merged_code
+  // Each file: "# FILE: path [...]\n````\n" + content + "\n````\n"  = 4 lines + content lines
+  let mergedCodeHeaderLine = 1; // Will be set properly
+  let currentMl = 1;
+
+  // We need to do this iteratively:
+  // 1. Calculate code_index with placeholder MLs
+  // 2. Count code_index lines
+  // 3. Calculate real MLs
+  // 4. Regenerate code_index with real MLs
+
+  // Step 1: Assign temporary ML values
+  let tempMl = 1; // placeholder
+  for (const fileObj of includedFiles) {
+    fileObj.mlStart = tempMl;
+    const isSkipped = skipContentSet.has(fileObj.relativePath);
+    if (isSkipped) {
+      fileObj.mlEnd = tempMl + 1; // placeholder line
+      tempMl += 4 + 1; // header(2) + placeholder(1) + footer(1) = 4
+    } else {
+      fileObj.mlEnd = tempMl + fileObj.lineCount - 1;
+      tempMl += 4 + fileObj.lineCount; // header(2) + content + footer(2)
+    }
+  }
+
+  // Step 2: Generate provisional code_index to count lines
+  let codeIndex = generateCodeIndex(
+    includedFiles,
+    projectRoot,
+    skipContentSet,
+    !argv.parse,
+  );
+  // codeIndex ends with \n, so split produces an extra empty element
+  const codeIndexLineCount = codeIndex.endsWith("\n")
+    ? codeIndex.split("\n").length - 1
+    : codeIndex.split("\n").length;
+
+  // Step 3: Calculate real header line count
+  let totalHeaderLines = 0;
+  if (argv.header) {
+    totalHeaderLines += systemPrompt.split("\n").length; // prompt lines + blank line from extra \n
+    totalHeaderLines += 1; // "<code_index>"
+    totalHeaderLines += codeIndexLineCount; // code index content
+    totalHeaderLines += 1; // "</code_index>"
+    totalHeaderLines += 1; // blank line
+    totalHeaderLines += 1; // "<merged_code>"
+  } else {
+    totalHeaderLines += 1; // "<merged_code>"
+  }
+
+  // Step 4: Recalculate ML offsets with real header
+  currentMl = totalHeaderLines + 1;
+  for (const fileObj of includedFiles) {
+    const isSkipped = skipContentSet.has(fileObj.relativePath);
+    // File header: "# FILE: path [...]\n````\n" = 2 lines
+    fileObj.mlStart = currentMl + 2; // Content starts after header
+    if (isSkipped) {
+      fileObj.mlEnd = fileObj.mlStart; // 1 line placeholder
+      currentMl += 2 + 1 + 2; // header(2) + placeholder(1) + footer(2: "\n````\n")
+    } else {
+      fileObj.mlEnd = fileObj.mlStart + fileObj.lineCount - 1;
+      currentMl += 2 + fileObj.lineCount + 2; // header(2) + content + footer(2)
+    }
+  }
+
+  // Step 5: Regenerate code_index with correct MLs
+  codeIndex = generateCodeIndex(
+    includedFiles,
+    projectRoot,
+    skipContentSet,
+    !argv.parse,
+  );
+
+  // Calculate total content size
+  const totalSizeBytes = includedFiles.reduce((acc, file) => {
+    if (skipContentSet.has(file.relativePath)) return acc;
+    return acc + file.size;
+  }, 0);
+
+  // --- Dry run ---
   if (argv.dryRun) {
-    console.log("\n📋 Files to be included (Dry Run):\n");
-    const tree = generateFileTree(includedFiles, projectRoot, skipContentSet);
-    console.log(tree);
-    console.log("\n📊 Summary (Dry Run):");
-    console.log(
-      `   • Included: ${includedFiles.length} files (${formatBytes(
-        totalSizeBytes
-      )})`
-    );
+    console.log("\n\ud83d\udccb Files to include (dry run):\n");
+    console.log(codeIndex);
+    console.log(`\n\ud83d\udcca Summary:`);
+    console.log(`   \u2022 Total files: ${includedFiles.length}`);
+    console.log(`   \u2022 Total size: ${formatBytes(totalSizeBytes)}`);
     if (skipContentSet.size > 0) {
-      console.log(`   • Content omitted: ${skipContentSet.size} files`);
+      console.log(`   \u2022 Content omitted: ${skipContentSet.size} files`);
     }
-    console.log(`   • Ignored:  ${stats.ignored} files/dirs`);
+    console.log(`\n\u2705 Done!`);
     return;
   }
 
+  // --- Write output ---
   const outputStream = fs.createWriteStream(argv.output);
   let totalLines = 0;
 
-  if (!argv.noHeader) {
-    const systemPrompt = argv.llmsTxt
-      ? LLMS_TXT_SYSTEM_PROMPT
-      : DEFAULT_SYSTEM_PROMPT;
+  if (argv.header) {
     outputStream.write(systemPrompt + "\n");
-    totalLines += systemPrompt.split("\n").length;
+    totalLines += systemPrompt.split("\n").length + 1;
 
-    outputStream.write("## Project File Tree\n\n");
-    outputStream.write("```\n");
-    const tree = generateFileTree(includedFiles, projectRoot, skipContentSet);
-    outputStream.write(tree);
-    outputStream.write("```\n\n");
-    outputStream.write("---\n\n");
+    outputStream.write("<code_index>\n");
+    outputStream.write(codeIndex);
+    outputStream.write("</code_index>\n\n");
+    totalLines += codeIndexLineCount + 3;
 
-    totalLines += tree.split("\n").length + 5;
+    outputStream.write("<merged_code>\n");
+    totalLines += 1;
+  } else {
+    outputStream.write("<merged_code>\n");
+    totalLines += 1;
   }
 
   for (const fileObj of includedFiles) {
     const relativePath = fileObj.relativePath.replace(/\\/g, "/");
-    const shouldSkipContent = skipContentSet.has(fileObj.relativePath);
+    const isSkipped = skipContentSet.has(fileObj.relativePath);
+
+    const olRange = `OL: 1-${fileObj.lineCount}`;
+    const mlRange = `ML: ${fileObj.mlStart}-${fileObj.mlEnd}`;
+    const sizeStr = fileObj.formattedSize;
 
-    outputStream.write(`### **FILE:** \`${relativePath}\`\n`);
+    outputStream.write(
+      `# FILE: ${relativePath} [${olRange} | ${mlRange} | ${sizeStr}]\n`,
+    );
     outputStream.write("````\n");
-    if (shouldSkipContent) {
-      outputStream.write(
-        `(Content omitted - file size: ${fileObj.formattedSize})`
-      );
+    totalLines += 2;
+
+    if (isSkipped) {
+      outputStream.write(`(Content omitted - file size: ${sizeStr})\n`);
       totalLines += 1;
     } else {
-      try {
-        const content = fs.readFileSync(fileObj.path, "utf8");
-        outputStream.write(content);
-        totalLines += content.split("\n").length;
-      } catch (e) {
-        outputStream.write(`... (error reading file: ${e.message}) ...`);
+      outputStream.write(fileObj.content);
+      if (!fileObj.content.endsWith("\n")) {
+        outputStream.write("\n");
       }
+      totalLines += fileObj.lineCount;
     }
-    outputStream.write("\n````\n\n");
-    totalLines += 4; // Headers/footers lines
+
+    outputStream.write("````\n\n");
+    totalLines += 2;
   }
+
+  outputStream.write("</merged_code>\n");
+  totalLines += 1;
   outputStream.end();
 
-  console.log(`\n📊 Summary:`);
+  console.log(`\n\ud83d\udcca Summary:`);
   console.log(
-    `   • Included: ${includedFiles.length} files (${formatBytes(
-      totalSizeBytes
-    )})`
+    `   \u2022 Included: ${includedFiles.length} files (${formatBytes(totalSizeBytes)})`,
   );
   if (skipContentSet.size > 0) {
-    console.log(`   • Content omitted: ${skipContentSet.size} files`);
+    console.log(`   \u2022 Content omitted: ${skipContentSet.size} files`);
   }
-  console.log(`   • Ignored:  ${stats.ignored} files/dirs`);
-  console.log(`   • Output:   ${argv.output} (~${totalLines} lines)`);
-  console.log(`\n✅ Done!`);
+  console.log(`   \u2022 Ignored:  ${stats.ignored} files/dirs`);
+  console.log(`   \u2022 Output:   ${argv.output} (~${totalLines} lines)`);
+  console.log(`\n\u2705 Done!`);
 }
 
 main().catch((err) => {