Merge pull request #30 from google/ericpts/image_part

Fix context issues.

Author: ericpts

agent.py

@@ -44,12 +44,7 @@ def multiply_numbers(x: float, y: float) -> dict:
 
 
 class BrowserAgent:
-    def __init__(
-        self,
-        browser_computer: Computer,
-        query: str,
-        model_name: str
-    ):
+    def __init__(self, browser_computer: Computer, query: str, model_name: str):
         self._browser_computer = browser_computer
         self._query = query
         self._model_name = model_name
@@ -283,15 +278,14 @@ def run_one_iteration(self) -> Literal["COMPLETE", "CONTINUE"]:
                 function_responses.append(
                     FunctionResponse(
                         name=function_call.name,
-                        response={
-                            "image": {
-                                "mimetype": "image/png",
-                                "data": base64.b64encode(fc_result.screenshot).decode(
-                                    "utf-8"
-                                ),
-                            },
-                            "url": fc_result.url,
-                        },
+                        response={"url": fc_result.url},
+                        data=[
+                            types.Part(
+                                inline_data=types.Blob(
+                                    mime_type="image/png", data=fc_result.screenshot
+                                )
+                            )
+                        ],
                     )
                 )
             elif isinstance(fc_result, dict):

computers/playwright/playwright.py

@@ -13,6 +13,7 @@
 # limitations under the License.
 import termcolor
 import time
+import os
 import sys
 from ..computer import (
     Computer,
@@ -88,11 +89,9 @@ def __enter__(self):
         print("Creating session...")
         self._playwright = sync_playwright().start()
         self._browser = self._playwright.chromium.launch(
-            args=[
-                '--disable-blink-features=AutomationControlled'
-            ],
-            headless=False,
-            )
+            args=["--disable-blink-features=AutomationControlled"],
+            headless=bool(os.environ.get("PLAYWRIGHT_HEADLESS", False)),
+        )
         self._context = self._browser.new_context(
             viewport={
                 "width": self._screen_size[0],

requirements.txt

@@ -1,6 +1,6 @@
 termcolor==3.1.0
 pydantic==2.11.4
-./sdk/google_genai-1.25.0-py3-none-any.whl
+./sdk/google_genai-1.27.0-py3-none-any.whl
 playwright==1.52.0
 browserbase==1.3.0
 rich