Merge branch 'main' into xw/bash-perf

openhands/agenthub/codeact_agent/codeact_agent.py

@@ -303,7 +303,6 @@ def get_observation_message(
                 and len(obs.set_of_marks) > 0
                 and self.config.enable_som_visual_browsing
                 and self.llm.vision_is_active()
-                and self.llm.is_visual_browser_tool_supported()
             ):
                 text += 'Image: Current webpage screenshot (Note that only visible portion of webpage is present in the screenshot. You may need to scroll to view the remaining portion of the web-page.)\n'
                 message = Message(

openhands/llm/llm.py

@@ -75,16 +75,6 @@
     'o3-mini',
 ]
 
-# visual browsing tool supported models
-# This flag is needed since gpt-4o and gpt-4o-mini do not allow passing image_urls with role='tool'
-VISUAL_BROWSING_TOOL_SUPPORTED_MODELS = [
-    'claude-3-5-sonnet',
-    'claude-3-5-sonnet-20240620',
-    'claude-3-5-sonnet-20241022',
-    'o1-2024-12-17',
-]
-
-
 REASONING_EFFORT_SUPPORTED_MODELS = [
     'o1-2024-12-17',
     'o1',
@@ -495,15 +485,6 @@ def is_function_calling_active(self) -> bool:
         """
         return self._function_calling_active
 
-    def is_visual_browser_tool_supported(self) -> bool:
-        return (
-            self.config.model in VISUAL_BROWSING_TOOL_SUPPORTED_MODELS
-            or self.config.model.split('/')[-1] in VISUAL_BROWSING_TOOL_SUPPORTED_MODELS
-            or any(
-                m in self.config.model for m in VISUAL_BROWSING_TOOL_SUPPORTED_MODELS
-            )
-        )
-
     def _post_completion(self, response: ModelResponse) -> float:
         """Post-process the completion response.
 

pyproject.toml

@@ -67,7 +67,7 @@ runloop-api-client = "0.22.0"
 libtmux = ">=0.37,<0.40"
 pygithub = "^2.5.0"
 joblib = "*"
-openhands-aci = "^0.2.0"
+openhands-aci = "^0.2.1"
 python-socketio = "^5.11.4"
 redis = "^5.2.0"
 sse-starlette = "^2.1.3"