docs: update and reorganize documentation (#806)

docs/tools/web-fetch.md

@@ -0,0 +1,25 @@
+# Web Fetch Tool
+
+This document describes the `web_fetch` tool.
+
+## `web_fetch`
+
+- **Purpose:** Processes content from one or more URLs (up to 20) embedded in a prompt. This tool is ideal for tasks that require summarizing, comparing, or extracting specific information from web pages.
+- **Arguments:**
+  - `prompt` (string, required): A comprehensive prompt that includes the URL(s) (up to 20) to fetch and specific instructions on how to process their content. For example: `"Summarize https://example.com/article and extract key points from https://another.com/data"`. The prompt must contain at least one URL starting with `http://` or `https://`.
+- **Behavior:**
+  - The tool sends the prompt and the specified URLs to the Gemini API.
+  - The API fetches the content of the URLs, processes it according to the instructions in the prompt, and returns a consolidated response.
+  - The tool formats the response, including source attribution with citations, and returns it to the user.
+- **Examples:**
+  - Summarizing a single article:
+    ```
+    web_fetch(prompt="Can you summarize the main points of https://example.com/news/latest")
+    ```
+  - Comparing two articles:
+    ```
+    web_fetch(prompt="What are the differences in the conclusions of these two papers: https://arxiv.org/abs/2401.0001 and https://arxiv.org/abs/2401.0002?")
+    ```
+- **Important Notes:**
+  - This tool relies on the Gemini API's ability to access and process the given URLs.
+  - The quality of the output depends on the clarity of the instructions in the prompt.