oxi_agent/tools/browse/
browse_tool.rs1use super::config::BrowseConfig;
7use super::engine::BrowserEngine;
8use super::helpers;
9use super::tab_guard::TabGuard;
10use crate::tools::{AgentTool, AgentToolResult, ToolContext, ToolError, ToolExecutionMode};
11use async_trait::async_trait;
12use parking_lot::Mutex;
13use serde_json::{json, Value};
14use std::sync::Arc;
15use tokio::sync::oneshot;
16
17pub struct BrowseTool {
21 engine: Arc<dyn BrowserEngine>,
22 config: BrowseConfig,
23 callbacks: super::callback_mixin::BrowseCallbacks,
25 tab_id_slot: Mutex<Arc<parking_lot::Mutex<Option<uuid::Uuid>>>>,
29}
30
31impl BrowseTool {
32 pub fn new(engine: Arc<dyn BrowserEngine>) -> Self {
34 Self {
35 engine,
36 config: BrowseConfig::default(),
37 callbacks: super::callback_mixin::BrowseCallbacks::new(),
38 tab_id_slot: Mutex::new(Arc::new(parking_lot::Mutex::new(None))),
39 }
40 }
41
42 pub fn with_config(engine: Arc<dyn BrowserEngine>, config: BrowseConfig) -> Self {
44 Self {
45 engine,
46 config,
47 callbacks: super::callback_mixin::BrowseCallbacks::new(),
48 tab_id_slot: Mutex::new(Arc::new(parking_lot::Mutex::new(None))),
49 }
50 }
51}
52
53#[async_trait]
54impl AgentTool for BrowseTool {
55 fn name(&self) -> &str {
56 "browse"
57 }
58
59 fn label(&self) -> &str {
60 "Browse"
61 }
62
63 fn description(&self) -> &str {
64 "Browse a web page with a built-in headless browser. Renders JavaScript-powered \
65 pages and returns content as markdown (default), html, or links. Use when \
66 web_search results are insufficient and you need to read the actual page content. \
67 Supports waiting for dynamic content via CSS selectors."
68 }
69
70 fn parameters_schema(&self) -> Value {
71 json!({
72 "type": "object",
73 "properties": {
74 "url": {
75 "type": "string",
76 "description": "URL to browse"
77 },
78 "format": {
79 "type": "string",
80 "enum": ["markdown", "html", "text", "links"],
81 "default": "markdown",
82 "description": "Output format: markdown (default), html, plain text, or list of links"
83 },
84 "selector": {
85 "type": "string",
86 "description": "CSS selector to extract only matching elements"
87 },
88 "wait_for": {
89 "type": "string",
90 "description": "CSS selector to wait for before extracting (for JS-rendered content)"
91 },
92 "screenshot": {
93 "type": "boolean",
94 "default": false,
95 "description": "Include a PNG screenshot as an image block"
96 }
97 },
98 "required": ["url"]
99 })
100 }
101
102 fn on_progress(&self, callback: crate::tools::ProgressCallback) {
103 self.callbacks.store_progress(callback);
104 }
105
106 fn on_browse_progress(&self, callback: Arc<dyn Fn(super::BrowseProgress) + Send + Sync>) {
107 self.callbacks.store_browse(callback);
108 }
109
110 fn execution_mode(&self) -> ToolExecutionMode {
117 ToolExecutionMode::SequentialOnly
118 }
119
120 fn current_tab_id(&self) -> Option<uuid::Uuid> {
121 *self.tab_id_slot.lock().lock()
122 }
123
124 fn set_tab_id_slot(&self, slot: Arc<parking_lot::Mutex<Option<uuid::Uuid>>>) {
125 *self.tab_id_slot.lock() = slot;
126 }
127
128 async fn execute(
129 &self,
130 _tool_call_id: &str,
131 params: Value,
132 _signal: Option<oneshot::Receiver<()>>,
133 _ctx: &ToolContext,
134 ) -> Result<AgentToolResult, ToolError> {
135 let url = params["url"]
136 .as_str()
137 .ok_or_else(|| "Missing required parameter: url".to_string())?;
138
139 let format = params["format"].as_str().unwrap_or("markdown");
140 let selector = params["selector"].as_str();
141 let wait_for = params["wait_for"].as_str();
142 let want_screenshot = params["screenshot"].as_bool().unwrap_or(false);
143
144 tracing::info!(url = %url, format = %format, "browsing page");
145
146 let raw_tab = self
148 .engine
149 .new_tab()
150 .await
151 .map_err(|e| format!("Failed to open browser tab: {}", e))?;
152
153 let tab_id = raw_tab.tab_id();
156 *self.tab_id_slot.lock().lock() = Some(tab_id);
157
158 self.callbacks.register_on_tab(raw_tab.as_ref());
160
161 let guard = TabGuard::new(raw_tab);
162 let tab = guard.tab();
163
164 let page = tab
166 .goto(url)
167 .await
168 .map_err(|e| format!("Navigation failed: {}", e))?;
169
170 if let Some(sel) = wait_for {
172 tab.wait_for(sel, self.config.default_wait_timeout_ms)
173 .await
174 .map_err(|e| format!("wait_for '{}' failed: {}", sel, e))?;
175 }
176
177 let output = match format {
179 "html" => {
180 if let Some(sel) = selector {
181 tab.query_all(sel)
182 .await
183 .map_err(|e| e.to_string())?
184 .join("\n\n")
185 } else {
186 page.html.clone()
187 }
188 }
189 "links" => {
190 let links = helpers::extract_links(tab).await?;
191 helpers::format_links(&links)
192 }
193 "text" => {
194 if let Some(sel) = selector {
195 tab.query_all(sel)
196 .await
197 .map_err(|e| e.to_string())?
198 .join("\n")
199 } else {
200 page.markdown.clone()
201 }
202 }
203 _ => {
204 if let Some(sel) = selector {
206 tab.query_all(sel)
207 .await
208 .map_err(|e| e.to_string())?
209 .join("\n\n")
210 } else {
211 page.markdown.clone()
212 }
213 }
214 };
215
216 let title = page.title.clone();
217 let final_url = page.url.clone();
218 let status = page.status;
219
220 let screenshot_blocks = if want_screenshot {
222 match tab.screenshot(self.config.screenshot_width).await {
223 Ok(png) => {
224 let b64 =
225 base64::Engine::encode(&base64::engine::general_purpose::STANDARD, &png);
226 let img =
227 oxi_ai::ContentBlock::Image(oxi_ai::ImageContent::new(b64, "image/png"));
228 Some(vec![img])
229 }
230 Err(e) => {
231 tracing::warn!("screenshot failed for {}: {}", final_url, e);
232 None
233 }
234 }
235 } else {
236 None
237 };
238
239 guard.close().await;
241 *self.tab_id_slot.lock().lock() = None;
242
243 let mut result = AgentToolResult::success(output).with_metadata(json!({
244 "url": final_url,
245 "title": title,
246 "status": status,
247 }));
248
249 if let Some(blocks) = screenshot_blocks {
250 result = result.with_content_blocks(blocks);
251 }
252
253 Ok(result)
254 }
255}
256
257#[cfg(test)]
258mod tests {
259 use super::*;
260 use crate::tools::browse::engine::{BrowserError, BrowserTab};
261 use async_trait::async_trait;
262
263 struct MockEngine;
267
268 #[async_trait]
269 impl BrowserEngine for MockEngine {
270 async fn new_tab(&self) -> Result<Box<dyn BrowserTab>, BrowserError> {
271 Err(BrowserError::Backend("MockEngine: no real browser".into()))
272 }
273
274 async fn close(&self) -> Result<(), BrowserError> {
275 Ok(())
276 }
277
278 async fn is_alive(&self) -> bool {
279 false
280 }
281 }
282
283 #[test]
284 fn browse_tool_is_sequential_only() {
285 let tool = BrowseTool::new(std::sync::Arc::new(MockEngine));
286 assert!(matches!(
287 tool.execution_mode(),
288 crate::tools::ToolExecutionMode::SequentialOnly
289 ));
290 }
291
292 #[test]
293 fn browse_tool_tab_id_slot_receives_id_from_agent_loop() {
294 let tool = BrowseTool::new(std::sync::Arc::new(MockEngine));
297
298 assert!(tool.current_tab_id().is_none());
300
301 let slot: Arc<parking_lot::Mutex<Option<uuid::Uuid>>> =
303 Arc::new(parking_lot::Mutex::new(None));
304 tool.set_tab_id_slot(Arc::clone(&slot));
305
306 let tab_id = uuid::Uuid::new_v4();
308 *slot.lock() = Some(tab_id);
309
310 assert_eq!(tool.current_tab_id(), Some(tab_id));
312
313 *slot.lock() = None;
315 assert!(tool.current_tab_id().is_none());
316 }
317}