fix: limit grep results to prevent context_length_exceeded errors (#2510)

e2e-tests/snapshots/local_agent_ask.spec.ts_local-agent-ask-mode-1.txt

@@ -96,7 +96,7 @@
         "type": "function",
         "function": {
           "name": "grep",
-          "description": "Search for a regex pattern in the codebase using ripgrep.\n\n- Returns matching lines with file paths and line numbers\n- By default, the search is case-insensitive\n- Use include_pattern to filter by file type (e.g. '*.tsx')\n- Use exclude_pattern to skip certain files (e.g. '*.test.ts')",
+          "description": "Search for a regex pattern in the codebase using ripgrep.\n\n- Returns matching lines with file paths and line numbers\n- By default, the search is case-insensitive\n- Use include_pattern to filter by file type (e.g. '*.tsx')\n- Use exclude_pattern to skip certain files (e.g. '*.test.ts')\n- Results are limited to 100 matches by default (max 250). If results are truncated, narrow your search with include_pattern or a more specific query.",
           "parameters": {
             "$schema": "http://json-schema.org/draft-07/schema#",
             "type": "object",
@@ -116,6 +116,12 @@
               "case_sensitive": {
                 "description": "Whether the search should be case sensitive (default: false)",
                 "type": "boolean"
+              },
+              "limit": {
+                "description": "Maximum number of matches to return (default: 100, max: 250). Use include_pattern to narrow results if limit is reached.",
+                "type": "number",
+                "minimum": 1,
+                "maximum": 250
               }
             },
             "required": [

e2e-tests/snapshots/local_agent_auto.spec.ts_local-agent---auto-model-1.txt

@@ -232,7 +232,7 @@
       {
         "type": "function",
         "name": "grep",
-        "description": "Search for a regex pattern in the codebase using ripgrep.\n\n- Returns matching lines with file paths and line numbers\n- By default, the search is case-insensitive\n- Use include_pattern to filter by file type (e.g. '*.tsx')\n- Use exclude_pattern to skip certain files (e.g. '*.test.ts')",
+        "description": "Search for a regex pattern in the codebase using ripgrep.\n\n- Returns matching lines with file paths and line numbers\n- By default, the search is case-insensitive\n- Use include_pattern to filter by file type (e.g. '*.tsx')\n- Use exclude_pattern to skip certain files (e.g. '*.test.ts')\n- Results are limited to 100 matches by default (max 250). If results are truncated, narrow your search with include_pattern or a more specific query.",
         "parameters": {
           "$schema": "http://json-schema.org/draft-07/schema#",
           "type": "object",
@@ -252,6 +252,12 @@
             "case_sensitive": {
               "description": "Whether the search should be case sensitive (default: false)",
               "type": "boolean"
+            },
+            "limit": {
+              "description": "Maximum number of matches to return (default: 100, max: 250). Use include_pattern to narrow results if limit is reached.",
+              "type": "number",
+              "minimum": 1,
+              "maximum": 250
             }
           },
           "required": [

e2e-tests/snapshots/local_agent_basic.spec.ts_local-agent---dump-request-1.txt

@@ -231,7 +231,7 @@
         "type": "function",
         "function": {
           "name": "grep",
-          "description": "Search for a regex pattern in the codebase using ripgrep.\n\n- Returns matching lines with file paths and line numbers\n- By default, the search is case-insensitive\n- Use include_pattern to filter by file type (e.g. '*.tsx')\n- Use exclude_pattern to skip certain files (e.g. '*.test.ts')",
+          "description": "Search for a regex pattern in the codebase using ripgrep.\n\n- Returns matching lines with file paths and line numbers\n- By default, the search is case-insensitive\n- Use include_pattern to filter by file type (e.g. '*.tsx')\n- Use exclude_pattern to skip certain files (e.g. '*.test.ts')\n- Results are limited to 100 matches by default (max 250). If results are truncated, narrow your search with include_pattern or a more specific query.",
           "parameters": {
             "$schema": "http://json-schema.org/draft-07/schema#",
             "type": "object",
@@ -251,6 +251,12 @@
               "case_sensitive": {
                 "description": "Whether the search should be case sensitive (default: false)",
                 "type": "boolean"
+              },
+              "limit": {
+                "description": "Maximum number of matches to return (default: 100, max: 250). Use include_pattern to narrow results if limit is reached.",
+                "type": "number",
+                "minimum": 1,
+                "maximum": 250
               }
             },
             "required": [

e2e-tests/snapshots/local_agent_grep.spec.ts_local-agent---grep-search-1.aria.yml

@@ -34,7 +34,7 @@
 - button "Copy":
   - img
 - text: log
-- code: "src/main.tsx:2: import App from \"./App.tsx\"; src/main.tsx:4: createRoot(document.getElementById(\"root\")!).render(<App />); src/App.tsx:1: const App = () => <div>Minimal imported app</div>; src/App.tsx:3: export default App;"
+- code: "src/App.tsx:1: const App = () => <div>Minimal imported app</div>; src/App.tsx:3: export default App; src/main.tsx:2: import App from \"./App.tsx\"; src/main.tsx:4: createRoot(document.getElementById(\"root\")!).render(<App />);"
 - paragraph: I found the matches! The React app is initialized in src/main.tsx using createRoot, and the App component is defined in src/App.tsx and imported in src/main.tsx.
 - button "Copy":
   - img

e2e-tests/snapshots/local_agent_grep.spec.ts_local-agent---grep-search-3.aria.yml

@@ -5,4 +5,4 @@
 - button "Copy":
   - img
 - text: log
-- code: "src/main.tsx:2: import App from \"./App.tsx\"; src/main.tsx:4: createRoot(document.getElementById(\"root\")!).render(<App />); src/App.tsx:1: const App = () => <div>Minimal imported app</div>; src/App.tsx:3: export default App;"
\ No newline at end of file
+- code: "src/App.tsx:1: const App = () => <div>Minimal imported app</div>; src/App.tsx:3: export default App; src/main.tsx:2: import App from \"./App.tsx\"; src/main.tsx:4: createRoot(document.getElementById(\"root\")!).render(<App />);"
\ No newline at end of file

src/components/chat/DyadGrep.tsx

@@ -21,6 +21,8 @@ interface DyadGrepProps {
       exclude?: string;
       "case-sensitive"?: string;
       count?: string;
+      total?: string;
+      truncated?: string;
     };
   };
 }
@@ -39,6 +41,8 @@ export const DyadGrep: React.FC<DyadGrepProps> = ({ children, node }) => {
   const excludePattern = node?.properties?.exclude || "";
   const caseSensitive = node?.properties?.["case-sensitive"] === "true";
   const count = node?.properties?.count || "";
+  const total = node?.properties?.total || "";
+  const truncated = node?.properties?.truncated === "true";
   const hasResults = count !== "" && count !== "0";
 
   // Build description
@@ -55,7 +59,9 @@ export const DyadGrep: React.FC<DyadGrepProps> = ({ children, node }) => {
 
   // Build result summary
   const resultSummary = count
-    ? `${count} match${count === "1" ? "" : "es"}`
+    ? truncated && total
+      ? `${count} of ${total} matches`
+      : `${count} match${count === "1" ? "" : "es"}`
     : "";
 
   // Dynamic border styling

src/components/chat/DyadMarkdownParser.tsx

@@ -523,6 +523,8 @@ function renderCustomTag(
               exclude: attributes.exclude || "",
               "case-sensitive": attributes["case-sensitive"] || "",
               count: attributes.count || "",
+              total: attributes.total || "",
+              truncated: attributes.truncated || "",
             },
           }}
         >

src/pro/main/ipc/handlers/local_agent/tools/grep.ts

@@ -15,6 +15,10 @@ import log from "electron-log";
 
 const logger = log.scope("grep");
 
+const DEFAULT_LIMIT = 100;
+const MAX_LIMIT = 250;
+const MAX_LINE_LENGTH = 500;
+
 const grepSchema = z.object({
   query: z.string().describe("The regex pattern to search for"),
   include_pattern: z
@@ -31,6 +35,14 @@ const grepSchema = z.object({
     .boolean()
     .optional()
     .describe("Whether the search should be case sensitive (default: false)"),
+  limit: z
+    .number()
+    .min(1)
+    .max(MAX_LIMIT)
+    .optional()
+    .describe(
+      `Maximum number of matches to return (default: ${DEFAULT_LIMIT}, max: ${MAX_LIMIT}). Use include_pattern to narrow results if limit is reached.`,
+    ),
 });
 
 interface RipgrepMatch {
@@ -42,6 +54,7 @@ interface RipgrepMatch {
 function buildGrepAttributes(
   args: Partial<z.infer<typeof grepSchema>>,
   count?: number,
+  totalCount?: number,
 ): string {
   const attrs: string[] = [];
   if (args.query) {
@@ -59,9 +72,20 @@ function buildGrepAttributes(
   if (count !== undefined) {
     attrs.push(`count="${count}"`);
   }
+  if (totalCount !== undefined && totalCount > (count ?? 0)) {
+    attrs.push(`total="${totalCount}"`);
+    attrs.push(`truncated="true"`);
+  }
   return attrs.join(" ");
 }
 
+function truncateLineText(text: string): string {
+  if (text.length <= MAX_LINE_LENGTH) {
+    return text;
+  }
+  return text.slice(0, MAX_LINE_LENGTH) + "...";
+}
+
 async function runRipgrep({
   appPath,
   query,
@@ -82,7 +106,6 @@ async function runRipgrep({
       "--no-config",
       "--max-filesize",
       `${MAX_FILE_SEARCH_SIZE}`,
-      ...RIPGREP_EXCLUDED_GLOBS.flatMap((glob) => ["--glob", glob]),
     ];
 
     // Case sensitivity: default is case-insensitive
@@ -90,8 +113,9 @@ async function runRipgrep({
       args.push("--ignore-case");
     }
 
-    // Include pattern
-    if (includePat) {
+    // Include pattern (skip no-op "*" which would override exclusion globs
+    // and .gitignore rules since --glob always takes precedence over ignore logic)
+    if (includePat && includePat !== "*") {
       args.push("--glob", includePat);
     }
 
@@ -100,6 +124,10 @@ async function runRipgrep({
       args.push("--glob", `!${excludePat}`);
     }
 
+    // Exclusion globs come LAST so they always take precedence over any
+    // include pattern (later --glob flags override earlier ones in ripgrep)
+    args.push(...RIPGREP_EXCLUDED_GLOBS.flatMap((glob) => ["--glob", glob]));
+
     args.push("--", query, ".");
 
     const rg = spawn(getRgExecutablePath(), args, { cwd: appPath });
@@ -168,7 +196,8 @@ export const grepTool: ToolDefinition<z.infer<typeof grepSchema>> = {
 - Returns matching lines with file paths and line numbers
 - By default, the search is case-insensitive
 - Use include_pattern to filter by file type (e.g. '*.tsx')
-- Use exclude_pattern to skip certain files (e.g. '*.test.ts')`,
+- Use exclude_pattern to skip certain files (e.g. '*.test.ts')
+- Results are limited to ${DEFAULT_LIMIT} matches by default (max ${MAX_LIMIT}). If results are truncated, narrow your search with include_pattern or a more specific query.`,
   inputSchema: grepSchema,
   defaultConsent: "always",
 
@@ -192,7 +221,9 @@ export const grepTool: ToolDefinition<z.infer<typeof grepSchema>> = {
   },
 
   execute: async (args, ctx: AgentContext) => {
-    const matches = await runRipgrep({
+    const includePatWasWildcard = args.include_pattern === "*";
+
+    const allMatches = await runRipgrep({
       appPath: ctx.appPath,
       query: args.query,
       includePat: args.include_pattern,
@@ -200,18 +231,37 @@ export const grepTool: ToolDefinition<z.infer<typeof grepSchema>> = {
       caseSensitive: args.case_sensitive,
     });
 
-    const attrs = buildGrepAttributes(args, matches.length);
+    const totalCount = allMatches.length;
+    const limit = Math.min(args.limit ?? DEFAULT_LIMIT, MAX_LIMIT);
+    // Sort for deterministic output (ripgrep's parallel execution can produce varying order)
+    const sortedMatches = [...allMatches].sort(
+      (a, b) => a.path.localeCompare(b.path) || a.lineNumber - b.lineNumber,
+    );
+    const matches = sortedMatches.slice(0, limit);
+    const wasTruncated = totalCount > limit;
+
+    const attrs = buildGrepAttributes(args, matches.length, totalCount);
 
     if (matches.length === 0) {
       ctx.onXmlComplete(`<dyad-grep ${attrs}>No matches found.</dyad-grep>`);
       return "No matches found.";
     }
 
-    // Format output: path:line: content
+    // Format output: path:line: content (with truncated line text)
     const lines = matches.map(
-      (m) => `${m.path}:${m.lineNumber}: ${m.lineText}`,
+      (m) => `${m.path}:${m.lineNumber}: ${truncateLineText(m.lineText)}`,
     );
-    const resultText = lines.join("\n");
+    let resultText = lines.join("\n");
+
+    // Add truncation notice for the AI
+    if (wasTruncated) {
+      resultText += `\n\n[TRUNCATED: Showing ${matches.length} of ${totalCount} matches. Use include_pattern to narrow your search (e.g., include_pattern="*.tsx") or use a more specific query.]`;
+    }
+
+    // Warn the LLM that "*" was ignored so it doesn't retry with the same pattern
+    if (includePatWasWildcard) {
+      resultText += `\n\n[NOTE: include_pattern="*" was ignored because it matches all files including git-ignored files! Omit include_pattern to search all files, or use a specific glob like "*.ts".]`;
+    }
 
     ctx.onXmlComplete(
       `<dyad-grep ${attrs}>\n${escapeXmlContent(resultText)}\n</dyad-grep>`,