browser integration

2026-10-18 07:47:10 -07:00
parent e7beca74f2
commit 1020d7eb15
4 changed files with 770 additions and 4 deletions
@@ -96,4 +96,36 @@ let tools: [Tool] = [
        """,
        run: browserTool
    ),
+    Tool(
+        name: "chrome",
+        description: "Drive Chrome on this Mac through its DevTools protocol: pages, DOM, network, console and performance traces. "
+            + "It is Chrome with byre's own profile, started on first use and kept between calls, so a site the human has logged into there (passkeys included) stays logged in; when a site needs a login, ask the human to sign in in that Chrome window. "
+            + "ops: tabs (list tabs: id, title, url), navigate (url, optionally new_tab; waits for the page to load and clears the tab's network and console logs), "
+            + "js (run code in the tab and return its value: the last expression's, promises awaited, objects as JSON; use it to explore the DOM), select (tab), close (tab), "
+            + "network (the tab's requests since this tool first touched the tab or its last navigate: number, status, type, start and duration in ms, bytes, method, url; with request, one request's headers and bodies; filter keeps urls containing the text), "
+            + "console (console messages, exceptions and browser log entries), "
+            + "trace_start and trace_stop (a performance trace: start, navigate or act, stop; stop saves the trace file and reports FCP, LCP, DOMContentLoaded, load and the main thread's long tasks), "
+            + "cdp (send any DevTools protocol method with params to the tab, e.g. DOM.getDocument, CSS.getComputedStyleForNode, Emulation.setCPUThrottlingRate, Network.setCacheDisabled; returns its result as JSON). "
+            + "tab is a tab id from tabs; without it, the tab this tool last used, else Chrome's first tab. capture screenshots the tab's page. "
+            + "It drives Chrome without the mouse or keyboard; select, close, navigate with new_tab, cdp, capture and the first call that starts Chrome can change what is in front, so those hold the shared input lock like act does.",
+        schema: """
+        {"type": "object", "properties": {
+            "op": {"type": "string", "enum": ["tabs", "navigate", "js", "select", "close", "network", "console", "trace_start", "trace_stop", "cdp"]},
+            "tab": {"type": "string", "description": "Tab id from tabs (default: the tab this tool last used, else Chrome's first tab)."},
+            "url": {"type": "string", "description": "For navigate."},
+            "new_tab": {"type": "boolean", "description": "For navigate: open the url in a new tab."},
+            "code": {"type": "string", "description": "For js."},
+            "request": {"type": "integer", "description": "For network: the request number to show in full."},
+            "filter": {"type": "string", "description": "For network: only requests whose url contains this text (case-insensitive)."},
+            "method": {"type": "string", "description": "For cdp: the protocol method, e.g. DOM.getDocument."},
+            "params": {"type": "object", "description": "For cdp: the method's parameters."},
+            "wait_for": {"type": "string", "description": "Then wait until this text (case-insensitive) is in the tab's title or page text, polling every 200ms."},
+            "wait_ms": {"type": "integer", "description": "Wait this long after the op, before wait_for and capture."},
+            "timeout_ms": {"type": "integer", "description": "How long navigate waits for the load (default 15000) and wait_for waits (default 10000). Giving up is reported, not an error."},
+            "capture": {"type": "boolean", "description": "Finish with a screenshot of the tab's page."},
+            "max_edge": {"type": "number", "description": "Longest edge of the screenshot in pixels, 16 to 8192 (default 1568, which is as much as you can see)."}
+        }, "required": ["op"]}
+        """,
+        run: chromeTool
+    ),
 ]