Web fetch tool (#31)

* Adding a web fetch tool.

2 files changed, 177 insertions, 0 deletions

diff --git a/packages/cli/src/gemini.ts b/packages/cli/src/gemini.ts
index 8762d39b..f6de65c3 100644
--- a/packages/cli/src/gemini.ts
+++ b/packages/cli/src/gemini.ts
@@ -9,6 +9,7 @@ import { GlobTool } from './tools/glob.tool.js';
 import { EditTool } from './tools/edit.tool.js';
 import { TerminalTool } from './tools/terminal.tool.js';
 import { WriteFileTool } from './tools/write-file.tool.js';
+import { WebFetchTool } from './tools/web-fetch.tool.js';
 import { globalConfig } from './config/config.js';
 
 async function main() {
@@ -77,6 +78,7 @@ function registerTools(targetDir: string) {
   const editTool = new EditTool(targetDir);
   const terminalTool = new TerminalTool(targetDir);
   const writeFileTool = new WriteFileTool(targetDir);
+  const webFetchTool = new WebFetchTool();
 
   toolRegistry.registerTool(lsTool);
   toolRegistry.registerTool(readFileTool);
@@ -85,4 +87,5 @@ function registerTools(targetDir: string) {
   toolRegistry.registerTool(editTool);
   toolRegistry.registerTool(terminalTool);
   toolRegistry.registerTool(writeFileTool);
+  toolRegistry.registerTool(webFetchTool);
 }
diff --git a/packages/cli/src/tools/web-fetch.tool.ts b/packages/cli/src/tools/web-fetch.tool.ts
new file mode 100644
index 00000000..74467605
--- /dev/null
+++ b/packages/cli/src/tools/web-fetch.tool.ts
@@ -0,0 +1,174 @@
+import { SchemaValidator } from '../utils/schemaValidator.js';
+import { BaseTool, ToolResult } from './tools.js';
+import { ToolCallConfirmationDetails } from '../ui/types.js'; // Added for shouldConfirmExecute
+
+/**
+ * Parameters for the WebFetch tool
+ */
+export interface WebFetchToolParams {
+  /**
+   * The URL to fetch content from.
+   */
+  url: string;
+}
+
+/**
+ * Standardized result from the WebFetch tool
+ */
+export interface WebFetchToolResult extends ToolResult {}
+
+/**
+ * Implementation of the WebFetch tool that reads content from a URL.
+ */
+export class WebFetchTool extends BaseTool<
+  WebFetchToolParams,
+  WebFetchToolResult
+> {
+  static readonly Name: string = 'web_fetch';
+
+  /**
+   * Creates a new instance of the WebFetchTool
+   */
+  constructor() {
+    super(
+      WebFetchTool.Name,
+      'WebFetch',
+      'Fetches text content from a given URL. Handles potential network errors and non-success HTTP status codes.',
+      {
+        properties: {
+          url: {
+            description:
+              "The URL to fetch. Must be an absolute URL (e.g., 'https://example.com/file.txt').",
+            type: 'string',
+          },
+        },
+        required: ['url'],
+        type: 'object',
+      },
+    );
+    // No rootDirectory needed for web fetching
+  }
+
+  /**
+   * Validates the parameters for the WebFetch tool
+   * @param params Parameters to validate
+   * @returns An error message string if invalid, null otherwise
+   */
+  invalidParams(params: WebFetchToolParams): string | null {
+    // 1. Validate against the basic schema first
+    if (
+      this.schema.parameters &&
+      !SchemaValidator.validate(
+        this.schema.parameters as Record<string, unknown>,
+        params,
+      )
+    ) {
+      return 'Parameters failed schema validation.';
+    }
+
+    // 2. Validate the URL format and protocol
+    try {
+      const parsedUrl = new URL(params.url);
+      // Ensure it's an HTTP or HTTPS URL
+      if (!['http:', 'https:'].includes(parsedUrl.protocol)) {
+        return `Invalid URL protocol: "${parsedUrl.protocol}". Only 'http:' and 'https:' are supported.`;
+      }
+    } catch (error) {
+      // The URL constructor throws if the format is invalid
+      return `Invalid URL format: "${params.url}". Please provide a valid absolute URL (e.g., 'https://example.com').`;
+    }
+
+    // If all checks pass, the parameters are valid
+    return null;
+  }
+
+  /**
+   * Gets a description of the web fetch operation.
+   * @param params Parameters for the web fetch.
+   * @returns A string describing the operation.
+   */
+  getDescription(params: WebFetchToolParams): string {
+    // Shorten long URLs for display
+    const displayUrl =
+      params.url.length > 80
+        ? params.url.substring(0, 77) + '...'
+        : params.url;
+    return `Fetching content from ${displayUrl}`;
+  }
+
+  /**
+   * Determines if the tool should prompt for confirmation before execution.
+   * Web fetches are generally safe, so default to false.
+   * @param params Parameters for the tool execution
+   * @returns Whether execute should be confirmed.
+   */
+  async shouldConfirmExecute(
+    params: WebFetchToolParams,
+  ): Promise<ToolCallConfirmationDetails | false> {
+    // Could add logic here to confirm based on domain, etc. if needed
+    return Promise.resolve(false);
+  }
+
+  /**
+   * Fetches content from the specified URL.
+   * @param params Parameters for the web fetch operation.
+   * @returns Result with the fetched content or an error message.
+   */
+  async execute(params: WebFetchToolParams): Promise<WebFetchToolResult> {
+    const validationError = this.invalidParams(params);
+    if (validationError) {
+      return {
+        llmContent: `Error: Invalid parameters provided. Reason: ${validationError}`,
+        returnDisplay: `**Error:** Invalid parameters. ${validationError}`,
+      };
+    }
+
+    const url = params.url;
+
+    try {
+      const response = await fetch(url, {
+        headers: {
+          'User-Agent': 'GeminiCode-CLI/1.0',
+        },
+        signal: AbortSignal.timeout(15000) // 15 seconds timeout
+      });
+
+      if (!response.ok) {
+        // fetch doesn't throw on bad HTTP status codes (4xx, 5xx)
+        const errorText = `Failed to fetch data from ${url}. Status: ${response.status} ${response.statusText}`;
+        return {
+          llmContent: `Error: ${errorText}`,
+          returnDisplay: `**Error:** ${errorText}`,
+        };
+      }
+
+      // Assuming the response is text. Add checks for content-type if needed.
+      const data = await response.text();
+      let llmContent = '';
+      // Truncate very large responses for the LLM context
+      const MAX_LLM_CONTENT_LENGTH = 100000;
+      if (data) {
+        llmContent = `Fetched data from ${url}:\n\n${
+          data.length > MAX_LLM_CONTENT_LENGTH
+            ? data.substring(0, MAX_LLM_CONTENT_LENGTH) +
+              '\n... [Content truncated]'
+            : data
+        }`;
+      } else {
+        llmContent = `No data fetched from ${url}. Status: ${response.status}`;
+      }
+      return {
+        llmContent,
+        returnDisplay: `Fetched content from ${url}`, // Simple display message
+      };
+    } catch (error: any) {
+      // This catches network errors (DNS resolution, connection refused, etc.)
+      // and errors from the URL constructor if somehow bypassed validation (unlikely)
+      const errorMessage = `Failed to fetch data from ${url}. Error: ${error instanceof Error ? error.message : String(error)}`;
+      return {
+        llmContent: `Error: ${errorMessage}`,
+        returnDisplay: `**Error:** ${errorMessage}`,
+      };
+    }
+  }
+}