raggrep 0.12.0 → 0.12.2

package/dist/cli/main.js

@@ -14276,7 +14276,7 @@ init_logger();
 // package.json
 var package_default = {
   name: "raggrep",
-  version: "0.12.0",
+  version: "0.12.2",
   description: "Local filesystem-based RAG system for codebases - semantic search using local embeddings",
   type: "module",
   main: "./dist/index.js",
@@ -14778,14 +14778,47 @@ Examples:
         const toolPath = path25.join(toolDir, "raggrep.ts");
         const toolContent = `import { tool } from "@opencode-ai/plugin";
 
+/**
+ * Get the package executor command (pnpx if available, otherwise npx)
+ */
+async function getExecutor(): Promise<string> {
+  try {
+    // Try to find pnpm first (faster)
+    await Bun.spawn(['pnpm', '--version'], { stdout: 'pipe', stderr: 'pipe' }).exited;
+    return 'pnpx';
+  } catch {
+    // Fall back to npx
+    return 'npx';
+  }
+}
+
+/**
+ * Get the installed raggrep version
+ */
+async function getRagrepVersion(executor: string): Promise<string | null> {
+  try {
+    const proc = Bun.spawn([executor, 'raggrep', '--version'], { stdout: 'pipe', stderr: 'pipe' });
+    const output = await new Response(proc.stdout).text();
+    const match = output.match(/v([\\d.]+)/);
+    return match ? match[1] : null;
+  } catch {
+    return null;
+  }
+}
+
 export default tool({
   description:
-    "Search the codebase using semantic RAG (Retrieval-Augmented Generation). Uses raggrep to find relevant code snippets based on natural language queries. The index is managed automatically - first query creates it, changed files are re-indexed, and unchanged files use cached index.",
+    "Semantic code search powered by RAG - understands INTENT, not just literal text. Parses code using AST to extract functions, classes, and symbols with full context. Finds relevant code even when exact keywords don't match. Superior to grep for exploratory searches like 'authentication logic', 'error handling patterns', or 'configuration loading'.\\n\\n\uD83C\uDFAF USE THIS TOOL FIRST when you need to:\\n• Find WHERE code is located (functions, components, services)\\n• Understand HOW code is structured\\n• Discover RELATED code across multiple files\\n• Get a QUICK overview of a topic\\n\\n❌ DON'T read multiple files manually when you can:\\n  raggrep(\\"user authentication\\", { filter: [\\"src/\\"] })\\n\\n✅ INSTEAD of reading files one-by-one, search semantically:\\n  • \\"Find the auth middleware\\" vs read: auth.ts, middleware.ts, index.ts...\\n  • \\"Where are React components?\\" vs read: App.tsx, components/*, pages/*...\\n  • \\"Database connection logic?\\" vs read: db.ts, config.ts, models/*...\\n  • \\"Error handling patterns\\" vs read: error.ts, middleware.ts, handlers/*...\\n\\nThis saves ~10x tool calls and provides BETTER context by showing related code across the entire codebase.",
   args: {
     query: tool.schema
       .string()
       .describe(
-        "Natural language search query (e.g., 'user authentication', 'handle errors')"
+        "Natural language search query describing what you want to find. Be specific: 'auth middleware that checks JWT', 'React hooks for data fetching', 'database connection pool config'. This is MUCH faster than reading files manually."
+      ),
+    filter: tool.schema
+      .array(tool.schema.string())
+      .describe(
+        "Array of path prefixes or glob patterns to narrow search scope (OR logic). If user mentions a directory, use it. Otherwise infer from context. Common patterns: ['src/auth'], ['*.tsx', 'components/'], ['api/', 'routes/'], ['docs/', '*.md'], ['*.test.ts']. For broad search use ['src/'] or ['**/*']."
       ),
     top: tool.schema
       .number()
@@ -14798,15 +14831,18 @@ export default tool({
     type: tool.schema
       .string()
       .optional()
-      .describe("Filter by file extension (e.g., ts, tsx, js)"),
-    filter: tool.schema
-      .array(tool.schema.string())
-      .optional()
       .describe(
-        "Filter by path prefix or glob pattern. Multiple filters use OR logic. Examples: 'src/auth', '*.ts', '*.md', 'src/**/*.test.ts'"
+        "Filter by single file extension without dot (e.g., 'ts', 'tsx', 'js', 'md'). Prefer using 'filter' with glob patterns like '*.ts' for more flexibility."
       ),
   },
   async execute(args) {
+    const executor = await getExecutor();
+    const version = await getRagrepVersion(executor);
+
+    if (!version) {
+      return \`Error: raggrep not found. Install it with: \${executor} install -g raggrep\`;
+    }
+
     const cmdArgs = [args.query];
 
     if (args.top !== undefined) {
@@ -14824,26 +14860,17 @@ export default tool({
       }
     }
 
-    const result = await Bun.$\`raggrep query \${cmdArgs}\`.text();
+    const proc = Bun.spawn([executor, 'raggrep', 'query', ...cmdArgs], { stdout: 'pipe' });
+    const result = await new Response(proc.stdout).text();
     return result.trim();
   },
 });
 `;
         try {
           await fs10.mkdir(toolDir, { recursive: true });
-          let action = "Installed";
-          const backupPath = toolPath + ".backup";
-          try {
-            await fs10.access(toolPath);
-            await fs10.copyFile(toolPath, backupPath);
-            action = "Updated";
-          } catch {}
           await fs10.writeFile(toolPath, toolContent, "utf-8");
-          console.log(`${action} raggrep tool for opencode.`);
+          console.log(`Installed raggrep tool for opencode.`);
           console.log(`  Location: ${toolPath}`);
-          if (action === "Updated") {
-            console.log(`  Backup:   ${backupPath}`);
-          }
           console.log(`
 The raggrep tool is now available in opencode.`);
         } catch (error) {
@@ -14892,4 +14919,4 @@ Run 'raggrep <command> --help' for more information.
 }
 main();
 
-//# debugId=E6A5C2A479D4470864756E2164756E21
+//# debugId=049DF49BCF57EE1664756E2164756E21