Add pydoll frameid example

examples/pydoll_tab_example.py

@@ -0,0 +1,207 @@
+"""
+Example: use a Pydoll Tab with the Stagehand Python SDK.
+
+What this demonstrates:
+- Start a Stagehand session (remote Stagehand API / Browserbase browser)
+- Attach Pydoll to the same browser via CDP (`cdp_url`)
+- Use Pydoll to navigate
+- Fetch the current page's `frame_id` via CDP `Page.getFrameTree`
+- Fetch the current page's `frame_id` via CDP ePage.getFrameTree`
+- Pass `frame_id` into `session.observe/act/extract`
+
+Environment variables required:
+- MODEL_API_KEY
+- BROWSERBASE_API_KEY
+- BROWSERBASE_PROJECT_ID
+
+Optional:
+- STAGEHAND_BASE_URL (defaults to https://api.stagehand.browserbase.com)
+
+Notes:
+- This example requires Python 3.10+ because `pydoll-python` requires Python 3.10+.
+- If this repo is pinned to an older Python via `.python-version`, run with:
+  - `uv run --python 3.12 python examples/pydoll_tab_example.py`
+"""
+
+from __future__ import annotations
+
+import os
+import sys
+import asyncio
+from typing import Any
+
+from stagehand import AsyncStagehand
+
+
+def _normalize_ws_address_for_pydoll(cdp_url: str) -> str:
+    # Pydoll currently validates the address strictly as `ws://...` (not `wss://...`).
+    if cdp_url.startswith("ws://"):
+        return cdp_url
+    if cdp_url.startswith("wss://"):
+        return "ws://" + cdp_url.removeprefix("wss://")
+    if cdp_url.startswith("http://"):
+        return "ws://" + cdp_url.removeprefix("http://")
+    if cdp_url.startswith("https://"):
+        return "ws://" + cdp_url.removeprefix("https://")
+    raise RuntimeError(f"Unsupported CDP URL scheme for Pydoll: {cdp_url!r}")
+
+
+async def _pydoll_attach_to_tab_session(*, chrome: Any, tab: Any) -> tuple[Any, str]:
+    """
+    Attach to the tab target via CDP Target.attachToTarget (flatten mode) and return (handler, session_id).
+
+    For some CDP proxies (including Browserbase connect URLs), the `/devtools/page/<id>` endpoints may not
+    behave like local Chrome's endpoints. Attaching and sending commands with `sessionId` is the most
+    compatible approach.
+    """
+    handler = getattr(chrome, "_connection_handler", None)
+    if handler is None:
+        raise RuntimeError("Could not find Pydoll browser connection handler on `chrome`.")
+
+    target_id = getattr(tab, "_target_id", None) or getattr(tab, "target_id", None)
+    if not target_id:
+        raise RuntimeError("Could not find a target id on the tab (expected `tab._target_id`).")
+
+    attached = await handler.execute_command(
+        {
+            "method": "Target.attachToTarget",
+            "params": {"targetId": target_id, "flatten": True},
+        },
+        timeout=60,
+    )
+    try:
+        return handler, attached["result"]["sessionId"]
+    except Exception as e:  # noqa: BLE001
+        raise RuntimeError("Failed to attach to target and get sessionId") from e
+
+
+async def _pydoll_execute_on_session(*, handler: Any, session_id: str, command: dict[str, Any]) -> dict[str, Any]:
+    cmd = dict(command)
+    cmd["sessionId"] = session_id
+    return await handler.execute_command(cmd, timeout=60)
+
+
+async def _pydoll_session_to_frame_id(*, handler: Any, session_id: str) -> str:
+    response = await _pydoll_execute_on_session(
+        handler=handler,
+        session_id=session_id,
+        command={"method": "Page.getFrameTree", "params": {}},
+    )
+    try:
+        return response["result"]["frameTree"]["frame"]["id"]
+    except Exception as e:  # noqa: BLE001
+        raise RuntimeError("Failed to extract frame id from CDP Page.getFrameTree response") from e
+
+
+async def main() -> None:
+    model_api_key = os.environ.get("MODEL_API_KEY")
+    if not model_api_key:
+        sys.exit("Set the MODEL_API_KEY environment variable to run this example.")
+
+    bb_api_key = os.environ.get("BROWSERBASE_API_KEY")
+    bb_project_id = os.environ.get("BROWSERBASE_PROJECT_ID")
+    if not bb_api_key or not bb_project_id:
+        sys.exit("Set BROWSERBASE_API_KEY and BROWSERBASE_PROJECT_ID to run this example.")
+
+    try:
+        from pydoll.browser.chromium import Chrome  # type: ignore[import-not-found]
+    except Exception:
+        sys.exit(
+            "Pydoll is not installed. Install it with:\n"
+            "  uv pip install pydoll-python\n"
+            "or:\n"
+            "  pip install pydoll-python\n"
+        )
+
+    async with AsyncStagehand(
+        server="remote",
+        browserbase_api_key=bb_api_key,
+        browserbase_project_id=bb_project_id,
+        model_api_key=model_api_key,
+    ) as client:
+        print("⏳ Starting Stagehand session...")
+        session = await client.sessions.create(
+            model_name="openai/gpt-5-nano",
+            browser={"type": "browserbase"},
+        )
+
+        cdp_url = session.data.cdp_url
+        if not cdp_url:
+            sys.exit("No cdp_url returned from the API for this session; cannot attach Pydoll.")
+
+        print(f"✅ Session started: {session.id}")
+        print("🔌 Connecting Pydoll to the same browser over CDP...")
+
+        chrome = Chrome()
+        try:
+            ws_address = _normalize_ws_address_for_pydoll(cdp_url)
+            if ws_address != cdp_url:
+                print(f"ℹ️ Normalized cdp_url for Pydoll: {ws_address}")
+            tab = await chrome.connect(ws_address)
+
+            handler, session_id = await _pydoll_attach_to_tab_session(chrome=chrome, tab=tab)
+
+            await _pydoll_execute_on_session(
+                handler=handler,
+                session_id=session_id,
+                command={"method": "Page.enable", "params": {}},
+            )
+            await _pydoll_execute_on_session(
+                handler=handler,
+                session_id=session_id,
+                command={"method": "Runtime.enable", "params": {}},
+            )
+
+            # Navigate a bit using CDP (via the attached session).
+            await _pydoll_execute_on_session(
+                handler=handler,
+                session_id=session_id,
+                command={"method": "Page.navigate", "params": {"url": "https://example.com"}},
+            )
+            await asyncio.sleep(2)
+
+            await _pydoll_execute_on_session(
+                handler=handler,
+                session_id=session_id,
+                command={
+                    "method": "Page.navigate",
+                    "params": {"url": "https://www.iana.org/domains/reserved"},
+                },
+            )
+            await asyncio.sleep(2)
+
+            frame_id = await _pydoll_session_to_frame_id(handler=handler, session_id=session_id)
+            print(f"🧩 frame_id: {frame_id}")
+
+            print("👀 Stagehand.observe(frame_id=...) ...")
+            actions = await session.observe(
+                instruction="Find the most relevant click target on this page",
+                frame_id=frame_id,
+            )
+            print(f"Observed {len(actions.data.result)} actions")
+
+            print("🧠 Stagehand.extract(frame_id=...) ...")
+            extracted = await session.extract(
+                instruction="Extract the page title and the primary heading (h1) text",
+                schema={
+                    "type": "object",
+                    "properties": {
+                        "title": {"type": "string"},
+                        "h1": {"type": "string"},
+                    },
+                    "required": ["title", "h1"],
+                    "additionalProperties": False,
+                },
+                frame_id=frame_id,
+            )
+            print("Extracted:", extracted.data.result)
+
+        finally:
+            close = getattr(chrome, "close", None)
+            if callable(close):
+                await close()
+            await session.end()
+
+
+if __name__ == "__main__":
+    asyncio.run(main())