Skip to content

Open-source Python library for AI browser agents, plus Browser Use Cloud: stealth browsers and a hosted agent API.

Switch all to jpeg

Magnus Müllera8e6af254e30
diff --git a/browser_use/agent/cloud_events.py b/browser_use/agent/cloud_events.py
index ed7b3c4b3..4ff893df4 100644
--- a/browser_use/agent/cloud_events.py
+++ b/browser_use/agent/cloud_events.py
@@ -155,7 +155,7 @@ class CreateAgentStepEvent(BaseEvent):
# Capture screenshot as base64 data URL if available
screenshot_url = None
if browser_state_summary.screenshot:
- screenshot_url = f'data:image/png;base64,{browser_state_summary.screenshot}'
+ screenshot_url = f'data:image/jpeg;base64,{browser_state_summary.screenshot}'
import logging
logger = logging.getLogger(__name__)
diff --git a/browser_use/agent/prompts.py b/browser_use/agent/prompts.py
index 42ef64926..2ea6cfa0d 100644
--- a/browser_use/agent/prompts.py
+++ b/browser_use/agent/prompts.py
@@ -367,8 +367,8 @@ Available tabs:
content_parts.append(
ContentPartImageParam(
image_url=ImageURL(
- url=f'data:image/png;base64,{screenshot}',
- media_type='image/png',
+ url=f'data:image/jpeg;base64,{screenshot}',
+ media_type='image/jpeg',
detail=self.vision_detail_level,
),
)
diff --git a/browser_use/llm/google/serializer.py b/browser_use/llm/google/serializer.py
index 2e85e580d..ff8dce6d0 100644
--- a/browser_use/llm/google/serializer.py
+++ b/browser_use/llm/google/serializer.py
@@ -101,13 +101,13 @@ class GoogleMessageSerializer:
# Handle images
url = part.image_url.url
- # Format: data:image/png;base64,<data>
+ # Format: data:image/jpeg;base64,<data>
header, data = url.split(',', 1)
# Decode base64 to bytes
image_bytes = base64.b64decode(data)
# Add image part
- image_part = Part.from_bytes(data=image_bytes, mime_type='image/png')
+ image_part = Part.from_bytes(data=image_bytes, mime_type='image/jpeg')
message_parts.append(image_part)
diff --git a/browser_use/llm/ollama/serializer.py b/browser_use/llm/ollama/serializer.py
index f7e0740da..afaf838a7 100644
--- a/browser_use/llm/ollama/serializer.py
+++ b/browser_use/llm/ollama/serializer.py
@@ -47,7 +47,7 @@ class OllamaMessageSerializer:
url = part.image_url.url
if url.startswith('data:'):
# Handle base64 encoded images
- # Format: data:image/png;base64,<data>
+ # Format: data:image/jpeg;base64,<data>
_, data = url.split(',', 1)
# Decode base64 to bytes
image_bytes = base64.b64decode(data)
diff --git a/browser_use/llm/tests/test_gemini_image.py b/browser_use/llm/tests/test_gemini_image.py
index 75c3e6e73..5d7dcbcc5 100644
--- a/browser_use/llm/tests/test_gemini_image.py
+++ b/browser_use/llm/tests/test_gemini_image.py
@@ -42,10 +42,10 @@ def create_random_text_image(text: str = 'hello world', width: int = 4000, heigh
# Convert to base64
buffer = io.BytesIO()
- image.save(buffer, format='PNG')
+ image.save(buffer, format='JPEG')
img_data = base64.b64encode(buffer.getvalue()).decode()
- return f'data:image/png;base64,{img_data}'
+ return f'data:image/jpeg;base64,{img_data}'
async def test_gemini_image_vision():
diff --git a/examples/features/add_image_context.py b/examples/features/add_image_context.py
index 0ec9875cc..2f759c8b6 100644
--- a/examples/features/add_image_context.py
+++ b/examples/features/add_image_context.py
@@ -77,8 +77,8 @@ def create_sample_images() -> list[ContentPartTextParam | ContentPartImageParam]
content_parts.append(
ContentPartImageParam(
image_url=ImageURL(
- url=f'data:image/png;base64,{item["value"]}',
- media_type='image/png',
+ url=f'data:image/jpeg;base64,{item["value"]}',
+ media_type='image/jpeg',
),
)
)