cua-agent 0.4.10__py3-none-any.whl → 0.4.12__py3-none-any.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Potentially problematic release.
This version of cua-agent might be problematic. Click here for more details.
- agent/cli.py +4 -4
- agent/computer_handler.py +12 -2
- agent/loops/__init__.py +2 -1
- agent/loops/anthropic.py +601 -504
- agent/loops/glm45v.py +902 -0
- agent/responses.py +45 -0
- {cua_agent-0.4.10.dist-info → cua_agent-0.4.12.dist-info}/METADATA +10 -1
- {cua_agent-0.4.10.dist-info → cua_agent-0.4.12.dist-info}/RECORD +10 -9
- {cua_agent-0.4.10.dist-info → cua_agent-0.4.12.dist-info}/WHEEL +0 -0
- {cua_agent-0.4.10.dist-info → cua_agent-0.4.12.dist-info}/entry_points.txt +0 -0
agent/loops/anthropic.py
CHANGED
|
@@ -23,7 +23,10 @@ from ..responses import (
|
|
|
23
23
|
make_type_item,
|
|
24
24
|
make_wait_item,
|
|
25
25
|
make_input_image_item,
|
|
26
|
-
make_screenshot_item
|
|
26
|
+
make_screenshot_item,
|
|
27
|
+
make_failed_tool_call_items,
|
|
28
|
+
make_left_mouse_down_item,
|
|
29
|
+
make_left_mouse_up_item
|
|
27
30
|
)
|
|
28
31
|
|
|
29
32
|
# Model version mapping to tool version and beta flag
|
|
@@ -115,7 +118,8 @@ async def _prepare_tools_for_anthropic(tool_schemas: List[Dict[str, Any]], model
|
|
|
115
118
|
def _convert_responses_items_to_completion_messages(messages: Messages) -> List[Dict[str, Any]]:
|
|
116
119
|
"""Convert responses_items message format to liteLLM completion format."""
|
|
117
120
|
completion_messages = []
|
|
118
|
-
|
|
121
|
+
call_id_to_fn_name = {}
|
|
122
|
+
|
|
119
123
|
for message in messages:
|
|
120
124
|
msg_type = message.get("type")
|
|
121
125
|
role = message.get("role")
|
|
@@ -193,6 +197,43 @@ def _convert_responses_items_to_completion_messages(messages: Messages) -> List[
|
|
|
193
197
|
"content": reasoning_text
|
|
194
198
|
})
|
|
195
199
|
|
|
200
|
+
elif msg_type == "function_call":
|
|
201
|
+
fn_name = message.get("name")
|
|
202
|
+
fn_args = message.get("arguments", "{}")
|
|
203
|
+
call_id = message.get("call_id", "call_1")
|
|
204
|
+
call_id_to_fn_name[call_id] = fn_name
|
|
205
|
+
openai_tool_calls = [{
|
|
206
|
+
"id": call_id,
|
|
207
|
+
"type": "function",
|
|
208
|
+
"function": {
|
|
209
|
+
"name": fn_name,
|
|
210
|
+
"arguments": fn_args
|
|
211
|
+
}
|
|
212
|
+
}] # If the last completion message is an assistant message, extend the tool_calls
|
|
213
|
+
if completion_messages and completion_messages[-1].get("role") == "assistant":
|
|
214
|
+
if "tool_calls" not in completion_messages[-1]:
|
|
215
|
+
completion_messages[-1]["tool_calls"] = []
|
|
216
|
+
completion_messages[-1]["tool_calls"].extend(openai_tool_calls)
|
|
217
|
+
else:
|
|
218
|
+
# Create new assistant message with tool calls
|
|
219
|
+
completion_messages.append({
|
|
220
|
+
"role": "assistant",
|
|
221
|
+
"content": None,
|
|
222
|
+
"tool_calls": openai_tool_calls
|
|
223
|
+
})
|
|
224
|
+
|
|
225
|
+
elif msg_type == "function_call_output":
|
|
226
|
+
call_id = message.get("call_id", "call_1")
|
|
227
|
+
fn_output = message.get("output", "")
|
|
228
|
+
fn_name = call_id_to_fn_name.get(call_id, "computer")
|
|
229
|
+
|
|
230
|
+
completion_messages.append({
|
|
231
|
+
"role": "function",
|
|
232
|
+
"name": fn_name,
|
|
233
|
+
"tool_call_id": call_id,
|
|
234
|
+
"content": str(fn_output)
|
|
235
|
+
})
|
|
236
|
+
|
|
196
237
|
elif msg_type == "computer_call":
|
|
197
238
|
# Computer call becomes tool use in assistant message
|
|
198
239
|
action = message.get("action", {})
|
|
@@ -527,6 +568,26 @@ def _convert_responses_items_to_completion_messages(messages: Messages) -> List[
|
|
|
527
568
|
"action": "screenshot"
|
|
528
569
|
}
|
|
529
570
|
})
|
|
571
|
+
elif action_type == "left_mouse_down":
|
|
572
|
+
tool_use_content.append({
|
|
573
|
+
"type": "tool_use",
|
|
574
|
+
"id": call_id,
|
|
575
|
+
"name": "computer",
|
|
576
|
+
"input": {
|
|
577
|
+
"action": "left_mouse_down",
|
|
578
|
+
"coordinate": [action.get("x", None), action.get("y", None)]
|
|
579
|
+
}
|
|
580
|
+
})
|
|
581
|
+
elif action_type == "left_mouse_up":
|
|
582
|
+
tool_use_content.append({
|
|
583
|
+
"type": "tool_use",
|
|
584
|
+
"id": call_id,
|
|
585
|
+
"name": "computer",
|
|
586
|
+
"input": {
|
|
587
|
+
"action": "left_mouse_up",
|
|
588
|
+
"coordinate": [action.get("x", None), action.get("y", None)]
|
|
589
|
+
}
|
|
590
|
+
})
|
|
530
591
|
|
|
531
592
|
# Convert tool_use_content to OpenAI tool_calls format
|
|
532
593
|
openai_tool_calls = []
|
|
@@ -611,45 +672,350 @@ def _convert_completion_to_responses_items(response: Any) -> List[Dict[str, Any]
|
|
|
611
672
|
# Action reference:
|
|
612
673
|
# https://docs.anthropic.com/en/docs/agents-and-tools/tool-use/computer-use-tool#available-actions
|
|
613
674
|
|
|
675
|
+
try:
|
|
676
|
+
# Basic actions (all versions)
|
|
677
|
+
if action_type == "screenshot":
|
|
678
|
+
responses_items.append(make_screenshot_item(call_id=call_id))
|
|
679
|
+
elif action_type in ["click", "left_click"]:
|
|
680
|
+
coordinate = tool_input.get("coordinate", [0, 0])
|
|
681
|
+
responses_items.append(make_click_item(
|
|
682
|
+
x=coordinate[0] if len(coordinate) > 0 else 0,
|
|
683
|
+
y=coordinate[1] if len(coordinate) > 1 else 0,
|
|
684
|
+
call_id=call_id
|
|
685
|
+
))
|
|
686
|
+
elif action_type in ["type", "type_text"]:
|
|
687
|
+
responses_items.append(make_type_item(
|
|
688
|
+
text=tool_input.get("text", ""),
|
|
689
|
+
call_id=call_id
|
|
690
|
+
))
|
|
691
|
+
elif action_type in ["key", "keypress", "hotkey"]:
|
|
692
|
+
responses_items.append(make_keypress_item(
|
|
693
|
+
keys=tool_input.get("text", "").replace("+", "-").split("-"),
|
|
694
|
+
call_id=call_id
|
|
695
|
+
))
|
|
696
|
+
elif action_type in ["mouse_move", "move_cursor", "move"]:
|
|
697
|
+
# Mouse move - create a custom action item
|
|
698
|
+
coordinate = tool_input.get("coordinate", [0, 0])
|
|
699
|
+
responses_items.append(
|
|
700
|
+
make_move_item(
|
|
701
|
+
x=coordinate[0] if len(coordinate) > 0 else 0,
|
|
702
|
+
y=coordinate[1] if len(coordinate) > 1 else 0,
|
|
703
|
+
call_id=call_id
|
|
704
|
+
)
|
|
705
|
+
)
|
|
706
|
+
|
|
707
|
+
# Enhanced actions (computer_20250124) Available in Claude 4 and Claude Sonnet 3.7
|
|
708
|
+
elif action_type == "scroll":
|
|
709
|
+
coordinate = tool_input.get("coordinate", [0, 0])
|
|
710
|
+
scroll_amount = tool_input.get("scroll_amount", 3)
|
|
711
|
+
scroll_x = scroll_amount if tool_input.get("scroll_direction", "down") == "right" else \
|
|
712
|
+
-scroll_amount if tool_input.get("scroll_direction", "down") == "left" else 0
|
|
713
|
+
scroll_y = scroll_amount if tool_input.get("scroll_direction", "down") == "down" else \
|
|
714
|
+
-scroll_amount if tool_input.get("scroll_direction", "down") == "up" else 0
|
|
715
|
+
responses_items.append(make_scroll_item(
|
|
716
|
+
x=coordinate[0] if len(coordinate) > 0 else 0,
|
|
717
|
+
y=coordinate[1] if len(coordinate) > 1 else 0,
|
|
718
|
+
scroll_x=scroll_x,
|
|
719
|
+
scroll_y=scroll_y,
|
|
720
|
+
call_id=call_id
|
|
721
|
+
))
|
|
722
|
+
elif action_type in ["left_click_drag", "drag"]:
|
|
723
|
+
start_coord = tool_input.get("start_coordinate", [0, 0])
|
|
724
|
+
end_coord = tool_input.get("end_coordinate", [0, 0])
|
|
725
|
+
responses_items.append(make_drag_item(
|
|
726
|
+
path=[
|
|
727
|
+
{
|
|
728
|
+
"x": start_coord[0] if len(start_coord) > 0 else 0,
|
|
729
|
+
"y": start_coord[1] if len(start_coord) > 1 else 0
|
|
730
|
+
},
|
|
731
|
+
{
|
|
732
|
+
"x": end_coord[0] if len(end_coord) > 0 else 0,
|
|
733
|
+
"y": end_coord[1] if len(end_coord) > 1 else 0
|
|
734
|
+
}
|
|
735
|
+
],
|
|
736
|
+
call_id=call_id
|
|
737
|
+
))
|
|
738
|
+
elif action_type == "right_click":
|
|
739
|
+
coordinate = tool_input.get("coordinate", [0, 0])
|
|
740
|
+
responses_items.append(make_click_item(
|
|
741
|
+
x=coordinate[0] if len(coordinate) > 0 else 0,
|
|
742
|
+
y=coordinate[1] if len(coordinate) > 1 else 0,
|
|
743
|
+
button="right",
|
|
744
|
+
call_id=call_id
|
|
745
|
+
))
|
|
746
|
+
elif action_type == "middle_click":
|
|
747
|
+
coordinate = tool_input.get("coordinate", [0, 0])
|
|
748
|
+
responses_items.append(make_click_item(
|
|
749
|
+
x=coordinate[0] if len(coordinate) > 0 else 0,
|
|
750
|
+
y=coordinate[1] if len(coordinate) > 1 else 0,
|
|
751
|
+
button="wheel",
|
|
752
|
+
call_id=call_id
|
|
753
|
+
))
|
|
754
|
+
elif action_type == "double_click":
|
|
755
|
+
coordinate = tool_input.get("coordinate", [0, 0])
|
|
756
|
+
responses_items.append(make_double_click_item(
|
|
757
|
+
x=coordinate[0] if len(coordinate) > 0 else 0,
|
|
758
|
+
y=coordinate[1] if len(coordinate) > 1 else 0,
|
|
759
|
+
call_id=call_id
|
|
760
|
+
))
|
|
761
|
+
elif action_type == "triple_click":
|
|
762
|
+
# coordinate = tool_input.get("coordinate", [0, 0])
|
|
763
|
+
# responses_items.append({
|
|
764
|
+
# "type": "computer_call",
|
|
765
|
+
# "call_id": call_id,
|
|
766
|
+
# "action": {
|
|
767
|
+
# "type": "triple_click",
|
|
768
|
+
# "x": coordinate[0] if len(coordinate) > 0 else 0,
|
|
769
|
+
# "y": coordinate[1] if len(coordinate) > 1 else 0
|
|
770
|
+
# }
|
|
771
|
+
# })
|
|
772
|
+
raise NotImplementedError("triple_click")
|
|
773
|
+
elif action_type == "left_mouse_down":
|
|
774
|
+
# coordinate = tool_input.get("coordinate", [0, 0])
|
|
775
|
+
# responses_items.append({
|
|
776
|
+
# "type": "computer_call",
|
|
777
|
+
# "call_id": call_id,
|
|
778
|
+
# "action": {
|
|
779
|
+
# "type": "mouse_down",
|
|
780
|
+
# "button": "left",
|
|
781
|
+
# "x": coordinate[0] if len(coordinate) > 0 else 0,
|
|
782
|
+
# "y": coordinate[1] if len(coordinate) > 1 else 0
|
|
783
|
+
# }
|
|
784
|
+
# })
|
|
785
|
+
coordinate = tool_input.get("coordinate", [None, None])
|
|
786
|
+
responses_items.append(make_left_mouse_down_item(
|
|
787
|
+
x=coordinate[0] if len(coordinate) > 0 else None,
|
|
788
|
+
y=coordinate[1] if len(coordinate) > 1 else None,
|
|
789
|
+
call_id=call_id
|
|
790
|
+
))
|
|
791
|
+
elif action_type == "left_mouse_up":
|
|
792
|
+
# coordinate = tool_input.get("coordinate", [0, 0])
|
|
793
|
+
# responses_items.append({
|
|
794
|
+
# "type": "computer_call",
|
|
795
|
+
# "call_id": call_id,
|
|
796
|
+
# "action": {
|
|
797
|
+
# "type": "mouse_up",
|
|
798
|
+
# "button": "left",
|
|
799
|
+
# "x": coordinate[0] if len(coordinate) > 0 else 0,
|
|
800
|
+
# "y": coordinate[1] if len(coordinate) > 1 else 0
|
|
801
|
+
# }
|
|
802
|
+
# })
|
|
803
|
+
coordinate = tool_input.get("coordinate", [None, None])
|
|
804
|
+
responses_items.append(make_left_mouse_up_item(
|
|
805
|
+
x=coordinate[0] if len(coordinate) > 0 else None,
|
|
806
|
+
y=coordinate[1] if len(coordinate) > 1 else None,
|
|
807
|
+
call_id=call_id
|
|
808
|
+
))
|
|
809
|
+
elif action_type == "hold_key":
|
|
810
|
+
# responses_items.append({
|
|
811
|
+
# "type": "computer_call",
|
|
812
|
+
# "call_id": call_id,
|
|
813
|
+
# "action": {
|
|
814
|
+
# "type": "key_hold",
|
|
815
|
+
# "key": tool_input.get("key", "")
|
|
816
|
+
# }
|
|
817
|
+
# })
|
|
818
|
+
raise NotImplementedError("hold_key")
|
|
819
|
+
elif action_type == "wait":
|
|
820
|
+
responses_items.append(make_wait_item(
|
|
821
|
+
call_id=call_id
|
|
822
|
+
))
|
|
823
|
+
else:
|
|
824
|
+
raise ValueError(f"Unknown action type: {action_type}")
|
|
825
|
+
except Exception as e:
|
|
826
|
+
responses_items.extend(make_failed_tool_call_items(
|
|
827
|
+
tool_name="computer",
|
|
828
|
+
tool_kwargs=tool_input,
|
|
829
|
+
error_message=repr(e),
|
|
830
|
+
call_id=call_id
|
|
831
|
+
))
|
|
832
|
+
|
|
833
|
+
# Handle tool calls (alternative format)
|
|
834
|
+
if hasattr(message, 'tool_calls') and message.tool_calls:
|
|
835
|
+
for tool_call in message.tool_calls:
|
|
836
|
+
if tool_call.function.name == "computer":
|
|
837
|
+
try:
|
|
838
|
+
try:
|
|
839
|
+
args = json.loads(tool_call.function.arguments)
|
|
840
|
+
action_type = args.get("action")
|
|
841
|
+
call_id = tool_call.id
|
|
842
|
+
|
|
614
843
|
# Basic actions (all versions)
|
|
615
844
|
if action_type == "screenshot":
|
|
616
|
-
|
|
845
|
+
# Input:
|
|
846
|
+
# {
|
|
847
|
+
# "function": {
|
|
848
|
+
# "name": "computer",
|
|
849
|
+
# "arguments": json.dumps({
|
|
850
|
+
# "action": "screenshot"
|
|
851
|
+
# })
|
|
852
|
+
# },
|
|
853
|
+
# "id": "call_1",
|
|
854
|
+
# "type": "function"
|
|
855
|
+
# }
|
|
856
|
+
|
|
857
|
+
# Output:
|
|
858
|
+
# {
|
|
859
|
+
# "type": "computer_call",
|
|
860
|
+
# "call_id": "call_1",
|
|
861
|
+
# "action": {
|
|
862
|
+
# "type": "screenshot"
|
|
863
|
+
# }
|
|
864
|
+
# }
|
|
865
|
+
responses_items.append(make_screenshot_item(
|
|
866
|
+
call_id=call_id
|
|
867
|
+
))
|
|
617
868
|
elif action_type in ["click", "left_click"]:
|
|
618
|
-
|
|
869
|
+
# Input:
|
|
870
|
+
# {
|
|
871
|
+
# "function": {
|
|
872
|
+
# "name": "computer",
|
|
873
|
+
# "arguments": json.dumps({
|
|
874
|
+
# "action": "click",
|
|
875
|
+
# "coordinate": [100, 200]
|
|
876
|
+
# })
|
|
877
|
+
# },
|
|
878
|
+
# "id": "call_1",
|
|
879
|
+
# "type": "function"
|
|
880
|
+
# }
|
|
881
|
+
|
|
882
|
+
# Output:
|
|
883
|
+
# {
|
|
884
|
+
# "type": "computer_call",
|
|
885
|
+
# "call_id": "call_1",
|
|
886
|
+
# "action": {
|
|
887
|
+
# "type": "click",
|
|
888
|
+
# "x": 100,
|
|
889
|
+
# "y": 200
|
|
890
|
+
# }
|
|
891
|
+
# }
|
|
892
|
+
coordinate = args.get("coordinate", [0, 0])
|
|
619
893
|
responses_items.append(make_click_item(
|
|
620
894
|
x=coordinate[0] if len(coordinate) > 0 else 0,
|
|
621
895
|
y=coordinate[1] if len(coordinate) > 1 else 0,
|
|
622
896
|
call_id=call_id
|
|
623
897
|
))
|
|
624
898
|
elif action_type in ["type", "type_text"]:
|
|
899
|
+
# Input:
|
|
900
|
+
# {
|
|
901
|
+
# "function": {
|
|
902
|
+
# "name": "computer",
|
|
903
|
+
# "arguments": json.dumps({
|
|
904
|
+
# "action": "type",
|
|
905
|
+
# "text": "Hello World"
|
|
906
|
+
# })
|
|
907
|
+
# },
|
|
908
|
+
# "id": "call_1",
|
|
909
|
+
# "type": "function"
|
|
910
|
+
# }
|
|
911
|
+
|
|
912
|
+
# Output:
|
|
913
|
+
# {
|
|
914
|
+
# "type": "computer_call",
|
|
915
|
+
# "call_id": "call_1",
|
|
916
|
+
# "action": {
|
|
917
|
+
# "type": "type",
|
|
918
|
+
# "text": "Hello World"
|
|
919
|
+
# }
|
|
920
|
+
# }
|
|
625
921
|
responses_items.append(make_type_item(
|
|
626
|
-
text=
|
|
922
|
+
text=args.get("text", ""),
|
|
627
923
|
call_id=call_id
|
|
628
924
|
))
|
|
629
925
|
elif action_type in ["key", "keypress", "hotkey"]:
|
|
926
|
+
# Input:
|
|
927
|
+
# {
|
|
928
|
+
# "function": {
|
|
929
|
+
# "name": "computer",
|
|
930
|
+
# "arguments": json.dumps({
|
|
931
|
+
# "action": "key",
|
|
932
|
+
# "text": "ctrl+c"
|
|
933
|
+
# })
|
|
934
|
+
# },
|
|
935
|
+
# "id": "call_1",
|
|
936
|
+
# "type": "function"
|
|
937
|
+
# }
|
|
938
|
+
|
|
939
|
+
# Output:
|
|
940
|
+
# {
|
|
941
|
+
# "type": "computer_call",
|
|
942
|
+
# "call_id": "call_1",
|
|
943
|
+
# "action": {
|
|
944
|
+
# "type": "keypress",
|
|
945
|
+
# "keys": ["ctrl", "c"]
|
|
946
|
+
# }
|
|
947
|
+
# }
|
|
630
948
|
responses_items.append(make_keypress_item(
|
|
631
|
-
keys=
|
|
949
|
+
keys=args.get("text", "").replace("+", "-").split("-"),
|
|
632
950
|
call_id=call_id
|
|
633
951
|
))
|
|
634
952
|
elif action_type in ["mouse_move", "move_cursor", "move"]:
|
|
635
|
-
#
|
|
636
|
-
|
|
637
|
-
|
|
638
|
-
|
|
639
|
-
|
|
640
|
-
|
|
641
|
-
|
|
642
|
-
|
|
643
|
-
|
|
953
|
+
# Input:
|
|
954
|
+
# {
|
|
955
|
+
# "function": {
|
|
956
|
+
# "name": "computer",
|
|
957
|
+
# "arguments": json.dumps({
|
|
958
|
+
# "action": "mouse_move",
|
|
959
|
+
# "coordinate": [150, 250]
|
|
960
|
+
# })
|
|
961
|
+
# },
|
|
962
|
+
# "id": "call_1",
|
|
963
|
+
# "type": "function"
|
|
964
|
+
# }
|
|
965
|
+
|
|
966
|
+
# Output:
|
|
967
|
+
# {
|
|
968
|
+
# "type": "computer_call",
|
|
969
|
+
# "call_id": "call_1",
|
|
970
|
+
# "action": {
|
|
971
|
+
# "type": "mouse_move",
|
|
972
|
+
# "x": 150,
|
|
973
|
+
# "y": 250
|
|
974
|
+
# }
|
|
975
|
+
# }
|
|
976
|
+
coordinate = args.get("coordinate", [0, 0])
|
|
977
|
+
responses_items.append(make_move_item(
|
|
978
|
+
x=coordinate[0] if len(coordinate) > 0 else 0,
|
|
979
|
+
y=coordinate[1] if len(coordinate) > 1 else 0,
|
|
980
|
+
call_id=call_id
|
|
981
|
+
))
|
|
644
982
|
|
|
645
983
|
# Enhanced actions (computer_20250124) Available in Claude 4 and Claude Sonnet 3.7
|
|
646
984
|
elif action_type == "scroll":
|
|
647
|
-
|
|
648
|
-
|
|
649
|
-
|
|
650
|
-
|
|
651
|
-
|
|
652
|
-
|
|
985
|
+
# Input:
|
|
986
|
+
# {
|
|
987
|
+
# "function": {
|
|
988
|
+
# "name": "computer",
|
|
989
|
+
# "arguments": json.dumps({
|
|
990
|
+
# "action": "scroll",
|
|
991
|
+
# "coordinate": [300, 400],
|
|
992
|
+
# "scroll_direction": "down",
|
|
993
|
+
# "scroll_amount": 5
|
|
994
|
+
# })
|
|
995
|
+
# },
|
|
996
|
+
# "id": "call_1",
|
|
997
|
+
# "type": "function"
|
|
998
|
+
# }
|
|
999
|
+
|
|
1000
|
+
# Output:
|
|
1001
|
+
# {
|
|
1002
|
+
# "type": "computer_call",
|
|
1003
|
+
# "call_id": "call_1",
|
|
1004
|
+
# "action": {
|
|
1005
|
+
# "type": "scroll",
|
|
1006
|
+
# "x": 300,
|
|
1007
|
+
# "y": 400,
|
|
1008
|
+
# "scroll_x": 0,
|
|
1009
|
+
# "scroll_y": -5
|
|
1010
|
+
# }
|
|
1011
|
+
# }
|
|
1012
|
+
coordinate = args.get("coordinate", [0, 0])
|
|
1013
|
+
direction = args.get("scroll_direction", "down")
|
|
1014
|
+
amount = args.get("scroll_amount", 3)
|
|
1015
|
+
scroll_x = amount if direction == "left" else \
|
|
1016
|
+
-amount if direction == "right" else 0
|
|
1017
|
+
scroll_y = amount if direction == "up" else \
|
|
1018
|
+
-amount if direction == "down" else 0
|
|
653
1019
|
responses_items.append(make_scroll_item(
|
|
654
1020
|
x=coordinate[0] if len(coordinate) > 0 else 0,
|
|
655
1021
|
y=coordinate[1] if len(coordinate) > 1 else 0,
|
|
@@ -658,8 +1024,34 @@ def _convert_completion_to_responses_items(response: Any) -> List[Dict[str, Any]
|
|
|
658
1024
|
call_id=call_id
|
|
659
1025
|
))
|
|
660
1026
|
elif action_type in ["left_click_drag", "drag"]:
|
|
661
|
-
|
|
662
|
-
|
|
1027
|
+
# Input:
|
|
1028
|
+
# {
|
|
1029
|
+
# "function": {
|
|
1030
|
+
# "name": "computer",
|
|
1031
|
+
# "arguments": json.dumps({
|
|
1032
|
+
# "action": "left_click_drag",
|
|
1033
|
+
# "start_coordinate": [100, 150],
|
|
1034
|
+
# "end_coordinate": [200, 250]
|
|
1035
|
+
# })
|
|
1036
|
+
# },
|
|
1037
|
+
# "id": "call_1",
|
|
1038
|
+
# "type": "function"
|
|
1039
|
+
# }
|
|
1040
|
+
|
|
1041
|
+
# Output:
|
|
1042
|
+
# {
|
|
1043
|
+
# "type": "computer_call",
|
|
1044
|
+
# "call_id": "call_1",
|
|
1045
|
+
# "action": {
|
|
1046
|
+
# "type": "drag",
|
|
1047
|
+
# "path": [
|
|
1048
|
+
# {"x": 100, "y": 150},
|
|
1049
|
+
# {"x": 200, "y": 250}
|
|
1050
|
+
# ]
|
|
1051
|
+
# }
|
|
1052
|
+
# }
|
|
1053
|
+
start_coord = args.get("start_coordinate", [0, 0])
|
|
1054
|
+
end_coord = args.get("end_coordinate", [0, 0])
|
|
663
1055
|
responses_items.append(make_drag_item(
|
|
664
1056
|
path=[
|
|
665
1057
|
{
|
|
@@ -674,7 +1066,31 @@ def _convert_completion_to_responses_items(response: Any) -> List[Dict[str, Any]
|
|
|
674
1066
|
call_id=call_id
|
|
675
1067
|
))
|
|
676
1068
|
elif action_type == "right_click":
|
|
677
|
-
|
|
1069
|
+
# Input:
|
|
1070
|
+
# {
|
|
1071
|
+
# "function": {
|
|
1072
|
+
# "name": "computer",
|
|
1073
|
+
# "arguments": json.dumps({
|
|
1074
|
+
# "action": "right_click",
|
|
1075
|
+
# "coordinate": [120, 180]
|
|
1076
|
+
# })
|
|
1077
|
+
# },
|
|
1078
|
+
# "id": "call_1",
|
|
1079
|
+
# "type": "function"
|
|
1080
|
+
# }
|
|
1081
|
+
|
|
1082
|
+
# Output:
|
|
1083
|
+
# {
|
|
1084
|
+
# "type": "computer_call",
|
|
1085
|
+
# "call_id": "call_1",
|
|
1086
|
+
# "action": {
|
|
1087
|
+
# "type": "click",
|
|
1088
|
+
# "x": 120,
|
|
1089
|
+
# "y": 180,
|
|
1090
|
+
# "button": "right"
|
|
1091
|
+
# }
|
|
1092
|
+
# }
|
|
1093
|
+
coordinate = args.get("coordinate", [0, 0])
|
|
678
1094
|
responses_items.append(make_click_item(
|
|
679
1095
|
x=coordinate[0] if len(coordinate) > 0 else 0,
|
|
680
1096
|
y=coordinate[1] if len(coordinate) > 1 else 0,
|
|
@@ -682,7 +1098,31 @@ def _convert_completion_to_responses_items(response: Any) -> List[Dict[str, Any]
|
|
|
682
1098
|
call_id=call_id
|
|
683
1099
|
))
|
|
684
1100
|
elif action_type == "middle_click":
|
|
685
|
-
|
|
1101
|
+
# Input:
|
|
1102
|
+
# {
|
|
1103
|
+
# "function": {
|
|
1104
|
+
# "name": "computer",
|
|
1105
|
+
# "arguments": json.dumps({
|
|
1106
|
+
# "action": "middle_click",
|
|
1107
|
+
# "coordinate": [140, 220]
|
|
1108
|
+
# })
|
|
1109
|
+
# },
|
|
1110
|
+
# "id": "call_1",
|
|
1111
|
+
# "type": "function"
|
|
1112
|
+
# }
|
|
1113
|
+
|
|
1114
|
+
# Output:
|
|
1115
|
+
# {
|
|
1116
|
+
# "type": "computer_call",
|
|
1117
|
+
# "call_id": "call_1",
|
|
1118
|
+
# "action": {
|
|
1119
|
+
# "type": "click",
|
|
1120
|
+
# "x": 140,
|
|
1121
|
+
# "y": 220,
|
|
1122
|
+
# "button": "wheel"
|
|
1123
|
+
# }
|
|
1124
|
+
# }
|
|
1125
|
+
coordinate = args.get("coordinate", [0, 0])
|
|
686
1126
|
responses_items.append(make_click_item(
|
|
687
1127
|
x=coordinate[0] if len(coordinate) > 0 else 0,
|
|
688
1128
|
y=coordinate[1] if len(coordinate) > 1 else 0,
|
|
@@ -690,518 +1130,175 @@ def _convert_completion_to_responses_items(response: Any) -> List[Dict[str, Any]
|
|
|
690
1130
|
call_id=call_id
|
|
691
1131
|
))
|
|
692
1132
|
elif action_type == "double_click":
|
|
693
|
-
|
|
1133
|
+
# Input:
|
|
1134
|
+
# {
|
|
1135
|
+
# "function": {
|
|
1136
|
+
# "name": "computer",
|
|
1137
|
+
# "arguments": json.dumps({
|
|
1138
|
+
# "action": "double_click",
|
|
1139
|
+
# "coordinate": [160, 240]
|
|
1140
|
+
# })
|
|
1141
|
+
# },
|
|
1142
|
+
# "id": "call_1",
|
|
1143
|
+
# "type": "function"
|
|
1144
|
+
# }
|
|
1145
|
+
|
|
1146
|
+
# Output:
|
|
1147
|
+
# {
|
|
1148
|
+
# "type": "computer_call",
|
|
1149
|
+
# "call_id": "call_1",
|
|
1150
|
+
# "action": {
|
|
1151
|
+
# "type": "double_click",
|
|
1152
|
+
# "x": 160,
|
|
1153
|
+
# "y": 240
|
|
1154
|
+
# }
|
|
1155
|
+
# }
|
|
1156
|
+
coordinate = args.get("coordinate", [0, 0])
|
|
694
1157
|
responses_items.append(make_double_click_item(
|
|
695
1158
|
x=coordinate[0] if len(coordinate) > 0 else 0,
|
|
696
1159
|
y=coordinate[1] if len(coordinate) > 1 else 0,
|
|
697
1160
|
call_id=call_id
|
|
698
1161
|
))
|
|
699
1162
|
elif action_type == "triple_click":
|
|
700
|
-
#
|
|
701
|
-
#
|
|
1163
|
+
# Input:
|
|
1164
|
+
# {
|
|
1165
|
+
# "function": {
|
|
1166
|
+
# "name": "computer",
|
|
1167
|
+
# "arguments": json.dumps({
|
|
1168
|
+
# "action": "triple_click",
|
|
1169
|
+
# "coordinate": [180, 260]
|
|
1170
|
+
# })
|
|
1171
|
+
# },
|
|
1172
|
+
# "id": "call_1",
|
|
1173
|
+
# "type": "function"
|
|
1174
|
+
# }
|
|
1175
|
+
|
|
1176
|
+
# Output:
|
|
1177
|
+
# {
|
|
702
1178
|
# "type": "computer_call",
|
|
703
|
-
# "call_id":
|
|
1179
|
+
# "call_id": "call_1",
|
|
704
1180
|
# "action": {
|
|
705
1181
|
# "type": "triple_click",
|
|
706
|
-
# "x":
|
|
707
|
-
# "y":
|
|
1182
|
+
# "x": 180,
|
|
1183
|
+
# "y": 260
|
|
708
1184
|
# }
|
|
709
|
-
# }
|
|
1185
|
+
# }
|
|
710
1186
|
raise NotImplementedError("triple_click")
|
|
711
1187
|
elif action_type == "left_mouse_down":
|
|
712
|
-
#
|
|
713
|
-
#
|
|
1188
|
+
# Input:
|
|
1189
|
+
# {
|
|
1190
|
+
# "function": {
|
|
1191
|
+
# "name": "computer",
|
|
1192
|
+
# "arguments": json.dumps({
|
|
1193
|
+
# "action": "left_mouse_down",
|
|
1194
|
+
# "coordinate": [200, 280]
|
|
1195
|
+
# })
|
|
1196
|
+
# },
|
|
1197
|
+
# "id": "call_1",
|
|
1198
|
+
# "type": "function"
|
|
1199
|
+
# }
|
|
1200
|
+
|
|
1201
|
+
# Output:
|
|
1202
|
+
# {
|
|
714
1203
|
# "type": "computer_call",
|
|
715
|
-
# "call_id":
|
|
1204
|
+
# "call_id": "call_1",
|
|
716
1205
|
# "action": {
|
|
717
1206
|
# "type": "mouse_down",
|
|
718
1207
|
# "button": "left",
|
|
719
|
-
# "x":
|
|
720
|
-
# "y":
|
|
1208
|
+
# "x": 200,
|
|
1209
|
+
# "y": 280
|
|
721
1210
|
# }
|
|
722
|
-
# }
|
|
723
|
-
|
|
1211
|
+
# }
|
|
1212
|
+
coordinate = args.get("coordinate", [None, None])
|
|
1213
|
+
responses_items.append(make_left_mouse_down_item(
|
|
1214
|
+
x=coordinate[0] if len(coordinate) > 0 else None,
|
|
1215
|
+
y=coordinate[1] if len(coordinate) > 1 else None,
|
|
1216
|
+
call_id=call_id
|
|
1217
|
+
))
|
|
724
1218
|
elif action_type == "left_mouse_up":
|
|
725
|
-
#
|
|
726
|
-
#
|
|
1219
|
+
# Input:
|
|
1220
|
+
# {
|
|
1221
|
+
# "function": {
|
|
1222
|
+
# "name": "computer",
|
|
1223
|
+
# "arguments": json.dumps({
|
|
1224
|
+
# "action": "left_mouse_up",
|
|
1225
|
+
# "coordinate": [220, 300]
|
|
1226
|
+
# })
|
|
1227
|
+
# },
|
|
1228
|
+
# "id": "call_1",
|
|
1229
|
+
# "type": "function"
|
|
1230
|
+
# }
|
|
1231
|
+
|
|
1232
|
+
# Output:
|
|
1233
|
+
# {
|
|
727
1234
|
# "type": "computer_call",
|
|
728
|
-
# "call_id":
|
|
1235
|
+
# "call_id": "call_1",
|
|
729
1236
|
# "action": {
|
|
730
1237
|
# "type": "mouse_up",
|
|
731
1238
|
# "button": "left",
|
|
732
|
-
# "x":
|
|
733
|
-
# "y":
|
|
1239
|
+
# "x": 220,
|
|
1240
|
+
# "y": 300
|
|
734
1241
|
# }
|
|
735
|
-
# }
|
|
736
|
-
|
|
1242
|
+
# }
|
|
1243
|
+
coordinate = args.get("coordinate", [None, None])
|
|
1244
|
+
responses_items.append(make_left_mouse_up_item(
|
|
1245
|
+
x=coordinate[0] if len(coordinate) > 0 else None,
|
|
1246
|
+
y=coordinate[1] if len(coordinate) > 1 else None,
|
|
1247
|
+
call_id=call_id
|
|
1248
|
+
))
|
|
737
1249
|
elif action_type == "hold_key":
|
|
738
|
-
#
|
|
1250
|
+
# Input:
|
|
1251
|
+
# {
|
|
1252
|
+
# "function": {
|
|
1253
|
+
# "name": "computer",
|
|
1254
|
+
# "arguments": json.dumps({
|
|
1255
|
+
# "action": "hold_key",
|
|
1256
|
+
# "key": "shift"
|
|
1257
|
+
# })
|
|
1258
|
+
# },
|
|
1259
|
+
# "id": "call_1",
|
|
1260
|
+
# "type": "function"
|
|
1261
|
+
# }
|
|
1262
|
+
|
|
1263
|
+
# Output:
|
|
1264
|
+
# {
|
|
739
1265
|
# "type": "computer_call",
|
|
740
|
-
# "call_id":
|
|
1266
|
+
# "call_id": "call_1",
|
|
741
1267
|
# "action": {
|
|
742
1268
|
# "type": "key_hold",
|
|
743
|
-
# "key":
|
|
1269
|
+
# "key": "shift"
|
|
744
1270
|
# }
|
|
745
|
-
# }
|
|
1271
|
+
# }
|
|
746
1272
|
raise NotImplementedError("hold_key")
|
|
747
1273
|
elif action_type == "wait":
|
|
1274
|
+
# Input:
|
|
1275
|
+
# {
|
|
1276
|
+
# "function": {
|
|
1277
|
+
# "name": "computer",
|
|
1278
|
+
# "arguments": json.dumps({
|
|
1279
|
+
# "action": "wait"
|
|
1280
|
+
# })
|
|
1281
|
+
# },
|
|
1282
|
+
# "id": "call_1",
|
|
1283
|
+
# "type": "function"
|
|
1284
|
+
# }
|
|
1285
|
+
|
|
1286
|
+
# Output:
|
|
1287
|
+
# {
|
|
1288
|
+
# "type": "computer_call",
|
|
1289
|
+
# "call_id": "call_1",
|
|
1290
|
+
# "action": {
|
|
1291
|
+
# "type": "wait"
|
|
1292
|
+
# }
|
|
1293
|
+
# }
|
|
748
1294
|
responses_items.append(make_wait_item(
|
|
749
1295
|
call_id=call_id
|
|
750
1296
|
))
|
|
751
|
-
|
|
752
|
-
|
|
753
|
-
|
|
754
|
-
|
|
755
|
-
|
|
756
|
-
for tool_call in message.tool_calls:
|
|
757
|
-
if tool_call.function.name == "computer":
|
|
758
|
-
try:
|
|
759
|
-
args = json.loads(tool_call.function.arguments)
|
|
760
|
-
action_type = args.get("action")
|
|
761
|
-
call_id = tool_call.id
|
|
762
|
-
|
|
763
|
-
# Basic actions (all versions)
|
|
764
|
-
if action_type == "screenshot":
|
|
765
|
-
# Input:
|
|
766
|
-
# {
|
|
767
|
-
# "function": {
|
|
768
|
-
# "name": "computer",
|
|
769
|
-
# "arguments": json.dumps({
|
|
770
|
-
# "action": "screenshot"
|
|
771
|
-
# })
|
|
772
|
-
# },
|
|
773
|
-
# "id": "call_1",
|
|
774
|
-
# "type": "function"
|
|
775
|
-
# }
|
|
776
|
-
|
|
777
|
-
# Output:
|
|
778
|
-
# {
|
|
779
|
-
# "type": "computer_call",
|
|
780
|
-
# "call_id": "call_1",
|
|
781
|
-
# "action": {
|
|
782
|
-
# "type": "screenshot"
|
|
783
|
-
# }
|
|
784
|
-
# }
|
|
785
|
-
responses_items.append(make_screenshot_item(
|
|
786
|
-
call_id=call_id
|
|
787
|
-
))
|
|
788
|
-
elif action_type in ["click", "left_click"]:
|
|
789
|
-
# Input:
|
|
790
|
-
# {
|
|
791
|
-
# "function": {
|
|
792
|
-
# "name": "computer",
|
|
793
|
-
# "arguments": json.dumps({
|
|
794
|
-
# "action": "click",
|
|
795
|
-
# "coordinate": [100, 200]
|
|
796
|
-
# })
|
|
797
|
-
# },
|
|
798
|
-
# "id": "call_1",
|
|
799
|
-
# "type": "function"
|
|
800
|
-
# }
|
|
801
|
-
|
|
802
|
-
# Output:
|
|
803
|
-
# {
|
|
804
|
-
# "type": "computer_call",
|
|
805
|
-
# "call_id": "call_1",
|
|
806
|
-
# "action": {
|
|
807
|
-
# "type": "click",
|
|
808
|
-
# "x": 100,
|
|
809
|
-
# "y": 200
|
|
810
|
-
# }
|
|
811
|
-
# }
|
|
812
|
-
coordinate = args.get("coordinate", [0, 0])
|
|
813
|
-
responses_items.append(make_click_item(
|
|
814
|
-
x=coordinate[0] if len(coordinate) > 0 else 0,
|
|
815
|
-
y=coordinate[1] if len(coordinate) > 1 else 0,
|
|
816
|
-
call_id=call_id
|
|
817
|
-
))
|
|
818
|
-
elif action_type in ["type", "type_text"]:
|
|
819
|
-
# Input:
|
|
820
|
-
# {
|
|
821
|
-
# "function": {
|
|
822
|
-
# "name": "computer",
|
|
823
|
-
# "arguments": json.dumps({
|
|
824
|
-
# "action": "type",
|
|
825
|
-
# "text": "Hello World"
|
|
826
|
-
# })
|
|
827
|
-
# },
|
|
828
|
-
# "id": "call_1",
|
|
829
|
-
# "type": "function"
|
|
830
|
-
# }
|
|
831
|
-
|
|
832
|
-
# Output:
|
|
833
|
-
# {
|
|
834
|
-
# "type": "computer_call",
|
|
835
|
-
# "call_id": "call_1",
|
|
836
|
-
# "action": {
|
|
837
|
-
# "type": "type",
|
|
838
|
-
# "text": "Hello World"
|
|
839
|
-
# }
|
|
840
|
-
# }
|
|
841
|
-
responses_items.append(make_type_item(
|
|
842
|
-
text=args.get("text", ""),
|
|
843
|
-
call_id=call_id
|
|
844
|
-
))
|
|
845
|
-
elif action_type in ["key", "keypress", "hotkey"]:
|
|
846
|
-
# Input:
|
|
847
|
-
# {
|
|
848
|
-
# "function": {
|
|
849
|
-
# "name": "computer",
|
|
850
|
-
# "arguments": json.dumps({
|
|
851
|
-
# "action": "key",
|
|
852
|
-
# "text": "ctrl+c"
|
|
853
|
-
# })
|
|
854
|
-
# },
|
|
855
|
-
# "id": "call_1",
|
|
856
|
-
# "type": "function"
|
|
857
|
-
# }
|
|
858
|
-
|
|
859
|
-
# Output:
|
|
860
|
-
# {
|
|
861
|
-
# "type": "computer_call",
|
|
862
|
-
# "call_id": "call_1",
|
|
863
|
-
# "action": {
|
|
864
|
-
# "type": "keypress",
|
|
865
|
-
# "keys": ["ctrl", "c"]
|
|
866
|
-
# }
|
|
867
|
-
# }
|
|
868
|
-
responses_items.append(make_keypress_item(
|
|
869
|
-
keys=args.get("text", "").replace("+", "-").split("-"),
|
|
870
|
-
call_id=call_id
|
|
871
|
-
))
|
|
872
|
-
elif action_type in ["mouse_move", "move_cursor", "move"]:
|
|
873
|
-
# Input:
|
|
874
|
-
# {
|
|
875
|
-
# "function": {
|
|
876
|
-
# "name": "computer",
|
|
877
|
-
# "arguments": json.dumps({
|
|
878
|
-
# "action": "mouse_move",
|
|
879
|
-
# "coordinate": [150, 250]
|
|
880
|
-
# })
|
|
881
|
-
# },
|
|
882
|
-
# "id": "call_1",
|
|
883
|
-
# "type": "function"
|
|
884
|
-
# }
|
|
885
|
-
|
|
886
|
-
# Output:
|
|
887
|
-
# {
|
|
888
|
-
# "type": "computer_call",
|
|
889
|
-
# "call_id": "call_1",
|
|
890
|
-
# "action": {
|
|
891
|
-
# "type": "mouse_move",
|
|
892
|
-
# "x": 150,
|
|
893
|
-
# "y": 250
|
|
894
|
-
# }
|
|
895
|
-
# }
|
|
896
|
-
coordinate = args.get("coordinate", [0, 0])
|
|
897
|
-
responses_items.append(make_move_item(
|
|
898
|
-
x=coordinate[0] if len(coordinate) > 0 else 0,
|
|
899
|
-
y=coordinate[1] if len(coordinate) > 1 else 0,
|
|
900
|
-
call_id=call_id
|
|
901
|
-
))
|
|
902
|
-
|
|
903
|
-
# Enhanced actions (computer_20250124) Available in Claude 4 and Claude Sonnet 3.7
|
|
904
|
-
elif action_type == "scroll":
|
|
905
|
-
# Input:
|
|
906
|
-
# {
|
|
907
|
-
# "function": {
|
|
908
|
-
# "name": "computer",
|
|
909
|
-
# "arguments": json.dumps({
|
|
910
|
-
# "action": "scroll",
|
|
911
|
-
# "coordinate": [300, 400],
|
|
912
|
-
# "scroll_direction": "down",
|
|
913
|
-
# "scroll_amount": 5
|
|
914
|
-
# })
|
|
915
|
-
# },
|
|
916
|
-
# "id": "call_1",
|
|
917
|
-
# "type": "function"
|
|
918
|
-
# }
|
|
919
|
-
|
|
920
|
-
# Output:
|
|
921
|
-
# {
|
|
922
|
-
# "type": "computer_call",
|
|
923
|
-
# "call_id": "call_1",
|
|
924
|
-
# "action": {
|
|
925
|
-
# "type": "scroll",
|
|
926
|
-
# "x": 300,
|
|
927
|
-
# "y": 400,
|
|
928
|
-
# "scroll_x": 0,
|
|
929
|
-
# "scroll_y": -5
|
|
930
|
-
# }
|
|
931
|
-
# }
|
|
932
|
-
coordinate = args.get("coordinate", [0, 0])
|
|
933
|
-
direction = args.get("scroll_direction", "down")
|
|
934
|
-
amount = args.get("scroll_amount", 3)
|
|
935
|
-
scroll_x = amount if direction == "left" else \
|
|
936
|
-
-amount if direction == "right" else 0
|
|
937
|
-
scroll_y = amount if direction == "up" else \
|
|
938
|
-
-amount if direction == "down" else 0
|
|
939
|
-
responses_items.append(make_scroll_item(
|
|
940
|
-
x=coordinate[0] if len(coordinate) > 0 else 0,
|
|
941
|
-
y=coordinate[1] if len(coordinate) > 1 else 0,
|
|
942
|
-
scroll_x=scroll_x,
|
|
943
|
-
scroll_y=scroll_y,
|
|
944
|
-
call_id=call_id
|
|
945
|
-
))
|
|
946
|
-
elif action_type in ["left_click_drag", "drag"]:
|
|
947
|
-
# Input:
|
|
948
|
-
# {
|
|
949
|
-
# "function": {
|
|
950
|
-
# "name": "computer",
|
|
951
|
-
# "arguments": json.dumps({
|
|
952
|
-
# "action": "left_click_drag",
|
|
953
|
-
# "start_coordinate": [100, 150],
|
|
954
|
-
# "end_coordinate": [200, 250]
|
|
955
|
-
# })
|
|
956
|
-
# },
|
|
957
|
-
# "id": "call_1",
|
|
958
|
-
# "type": "function"
|
|
959
|
-
# }
|
|
960
|
-
|
|
961
|
-
# Output:
|
|
962
|
-
# {
|
|
963
|
-
# "type": "computer_call",
|
|
964
|
-
# "call_id": "call_1",
|
|
965
|
-
# "action": {
|
|
966
|
-
# "type": "drag",
|
|
967
|
-
# "path": [
|
|
968
|
-
# {"x": 100, "y": 150},
|
|
969
|
-
# {"x": 200, "y": 250}
|
|
970
|
-
# ]
|
|
971
|
-
# }
|
|
972
|
-
# }
|
|
973
|
-
start_coord = args.get("start_coordinate", [0, 0])
|
|
974
|
-
end_coord = args.get("end_coordinate", [0, 0])
|
|
975
|
-
responses_items.append(make_drag_item(
|
|
976
|
-
path=[
|
|
977
|
-
{
|
|
978
|
-
"x": start_coord[0] if len(start_coord) > 0 else 0,
|
|
979
|
-
"y": start_coord[1] if len(start_coord) > 1 else 0
|
|
980
|
-
},
|
|
981
|
-
{
|
|
982
|
-
"x": end_coord[0] if len(end_coord) > 0 else 0,
|
|
983
|
-
"y": end_coord[1] if len(end_coord) > 1 else 0
|
|
984
|
-
}
|
|
985
|
-
],
|
|
986
|
-
call_id=call_id
|
|
987
|
-
))
|
|
988
|
-
elif action_type == "right_click":
|
|
989
|
-
# Input:
|
|
990
|
-
# {
|
|
991
|
-
# "function": {
|
|
992
|
-
# "name": "computer",
|
|
993
|
-
# "arguments": json.dumps({
|
|
994
|
-
# "action": "right_click",
|
|
995
|
-
# "coordinate": [120, 180]
|
|
996
|
-
# })
|
|
997
|
-
# },
|
|
998
|
-
# "id": "call_1",
|
|
999
|
-
# "type": "function"
|
|
1000
|
-
# }
|
|
1001
|
-
|
|
1002
|
-
# Output:
|
|
1003
|
-
# {
|
|
1004
|
-
# "type": "computer_call",
|
|
1005
|
-
# "call_id": "call_1",
|
|
1006
|
-
# "action": {
|
|
1007
|
-
# "type": "click",
|
|
1008
|
-
# "x": 120,
|
|
1009
|
-
# "y": 180,
|
|
1010
|
-
# "button": "right"
|
|
1011
|
-
# }
|
|
1012
|
-
# }
|
|
1013
|
-
coordinate = args.get("coordinate", [0, 0])
|
|
1014
|
-
responses_items.append(make_click_item(
|
|
1015
|
-
x=coordinate[0] if len(coordinate) > 0 else 0,
|
|
1016
|
-
y=coordinate[1] if len(coordinate) > 1 else 0,
|
|
1017
|
-
button="right",
|
|
1018
|
-
call_id=call_id
|
|
1019
|
-
))
|
|
1020
|
-
elif action_type == "middle_click":
|
|
1021
|
-
# Input:
|
|
1022
|
-
# {
|
|
1023
|
-
# "function": {
|
|
1024
|
-
# "name": "computer",
|
|
1025
|
-
# "arguments": json.dumps({
|
|
1026
|
-
# "action": "middle_click",
|
|
1027
|
-
# "coordinate": [140, 220]
|
|
1028
|
-
# })
|
|
1029
|
-
# },
|
|
1030
|
-
# "id": "call_1",
|
|
1031
|
-
# "type": "function"
|
|
1032
|
-
# }
|
|
1033
|
-
|
|
1034
|
-
# Output:
|
|
1035
|
-
# {
|
|
1036
|
-
# "type": "computer_call",
|
|
1037
|
-
# "call_id": "call_1",
|
|
1038
|
-
# "action": {
|
|
1039
|
-
# "type": "click",
|
|
1040
|
-
# "x": 140,
|
|
1041
|
-
# "y": 220,
|
|
1042
|
-
# "button": "wheel"
|
|
1043
|
-
# }
|
|
1044
|
-
# }
|
|
1045
|
-
coordinate = args.get("coordinate", [0, 0])
|
|
1046
|
-
responses_items.append(make_click_item(
|
|
1047
|
-
x=coordinate[0] if len(coordinate) > 0 else 0,
|
|
1048
|
-
y=coordinate[1] if len(coordinate) > 1 else 0,
|
|
1049
|
-
button="wheel",
|
|
1050
|
-
call_id=call_id
|
|
1051
|
-
))
|
|
1052
|
-
elif action_type == "double_click":
|
|
1053
|
-
# Input:
|
|
1054
|
-
# {
|
|
1055
|
-
# "function": {
|
|
1056
|
-
# "name": "computer",
|
|
1057
|
-
# "arguments": json.dumps({
|
|
1058
|
-
# "action": "double_click",
|
|
1059
|
-
# "coordinate": [160, 240]
|
|
1060
|
-
# })
|
|
1061
|
-
# },
|
|
1062
|
-
# "id": "call_1",
|
|
1063
|
-
# "type": "function"
|
|
1064
|
-
# }
|
|
1065
|
-
|
|
1066
|
-
# Output:
|
|
1067
|
-
# {
|
|
1068
|
-
# "type": "computer_call",
|
|
1069
|
-
# "call_id": "call_1",
|
|
1070
|
-
# "action": {
|
|
1071
|
-
# "type": "double_click",
|
|
1072
|
-
# "x": 160,
|
|
1073
|
-
# "y": 240
|
|
1074
|
-
# }
|
|
1075
|
-
# }
|
|
1076
|
-
coordinate = args.get("coordinate", [0, 0])
|
|
1077
|
-
responses_items.append(make_double_click_item(
|
|
1078
|
-
x=coordinate[0] if len(coordinate) > 0 else 0,
|
|
1079
|
-
y=coordinate[1] if len(coordinate) > 1 else 0,
|
|
1080
|
-
call_id=call_id
|
|
1081
|
-
))
|
|
1082
|
-
elif action_type == "triple_click":
|
|
1083
|
-
# Input:
|
|
1084
|
-
# {
|
|
1085
|
-
# "function": {
|
|
1086
|
-
# "name": "computer",
|
|
1087
|
-
# "arguments": json.dumps({
|
|
1088
|
-
# "action": "triple_click",
|
|
1089
|
-
# "coordinate": [180, 260]
|
|
1090
|
-
# })
|
|
1091
|
-
# },
|
|
1092
|
-
# "id": "call_1",
|
|
1093
|
-
# "type": "function"
|
|
1094
|
-
# }
|
|
1095
|
-
|
|
1096
|
-
# Output:
|
|
1097
|
-
# {
|
|
1098
|
-
# "type": "computer_call",
|
|
1099
|
-
# "call_id": "call_1",
|
|
1100
|
-
# "action": {
|
|
1101
|
-
# "type": "triple_click",
|
|
1102
|
-
# "x": 180,
|
|
1103
|
-
# "y": 260
|
|
1104
|
-
# }
|
|
1105
|
-
# }
|
|
1106
|
-
raise NotImplementedError("triple_click")
|
|
1107
|
-
elif action_type == "left_mouse_down":
|
|
1108
|
-
# Input:
|
|
1109
|
-
# {
|
|
1110
|
-
# "function": {
|
|
1111
|
-
# "name": "computer",
|
|
1112
|
-
# "arguments": json.dumps({
|
|
1113
|
-
# "action": "left_mouse_down",
|
|
1114
|
-
# "coordinate": [200, 280]
|
|
1115
|
-
# })
|
|
1116
|
-
# },
|
|
1117
|
-
# "id": "call_1",
|
|
1118
|
-
# "type": "function"
|
|
1119
|
-
# }
|
|
1120
|
-
|
|
1121
|
-
# Output:
|
|
1122
|
-
# {
|
|
1123
|
-
# "type": "computer_call",
|
|
1124
|
-
# "call_id": "call_1",
|
|
1125
|
-
# "action": {
|
|
1126
|
-
# "type": "mouse_down",
|
|
1127
|
-
# "button": "left",
|
|
1128
|
-
# "x": 200,
|
|
1129
|
-
# "y": 280
|
|
1130
|
-
# }
|
|
1131
|
-
# }
|
|
1132
|
-
raise NotImplementedError("left_mouse_down")
|
|
1133
|
-
elif action_type == "left_mouse_up":
|
|
1134
|
-
# Input:
|
|
1135
|
-
# {
|
|
1136
|
-
# "function": {
|
|
1137
|
-
# "name": "computer",
|
|
1138
|
-
# "arguments": json.dumps({
|
|
1139
|
-
# "action": "left_mouse_up",
|
|
1140
|
-
# "coordinate": [220, 300]
|
|
1141
|
-
# })
|
|
1142
|
-
# },
|
|
1143
|
-
# "id": "call_1",
|
|
1144
|
-
# "type": "function"
|
|
1145
|
-
# }
|
|
1146
|
-
|
|
1147
|
-
# Output:
|
|
1148
|
-
# {
|
|
1149
|
-
# "type": "computer_call",
|
|
1150
|
-
# "call_id": "call_1",
|
|
1151
|
-
# "action": {
|
|
1152
|
-
# "type": "mouse_up",
|
|
1153
|
-
# "button": "left",
|
|
1154
|
-
# "x": 220,
|
|
1155
|
-
# "y": 300
|
|
1156
|
-
# }
|
|
1157
|
-
# }
|
|
1158
|
-
raise NotImplementedError("left_mouse_up")
|
|
1159
|
-
elif action_type == "hold_key":
|
|
1160
|
-
# Input:
|
|
1161
|
-
# {
|
|
1162
|
-
# "function": {
|
|
1163
|
-
# "name": "computer",
|
|
1164
|
-
# "arguments": json.dumps({
|
|
1165
|
-
# "action": "hold_key",
|
|
1166
|
-
# "key": "shift"
|
|
1167
|
-
# })
|
|
1168
|
-
# },
|
|
1169
|
-
# "id": "call_1",
|
|
1170
|
-
# "type": "function"
|
|
1171
|
-
# }
|
|
1172
|
-
|
|
1173
|
-
# Output:
|
|
1174
|
-
# {
|
|
1175
|
-
# "type": "computer_call",
|
|
1176
|
-
# "call_id": "call_1",
|
|
1177
|
-
# "action": {
|
|
1178
|
-
# "type": "key_hold",
|
|
1179
|
-
# "key": "shift"
|
|
1180
|
-
# }
|
|
1181
|
-
# }
|
|
1182
|
-
raise NotImplementedError("hold_key")
|
|
1183
|
-
elif action_type == "wait":
|
|
1184
|
-
# Input:
|
|
1185
|
-
# {
|
|
1186
|
-
# "function": {
|
|
1187
|
-
# "name": "computer",
|
|
1188
|
-
# "arguments": json.dumps({
|
|
1189
|
-
# "action": "wait"
|
|
1190
|
-
# })
|
|
1191
|
-
# },
|
|
1192
|
-
# "id": "call_1",
|
|
1193
|
-
# "type": "function"
|
|
1194
|
-
# }
|
|
1195
|
-
|
|
1196
|
-
# Output:
|
|
1197
|
-
# {
|
|
1198
|
-
# "type": "computer_call",
|
|
1199
|
-
# "call_id": "call_1",
|
|
1200
|
-
# "action": {
|
|
1201
|
-
# "type": "wait"
|
|
1202
|
-
# }
|
|
1203
|
-
# }
|
|
1204
|
-
responses_items.append(make_wait_item(
|
|
1297
|
+
except Exception as e:
|
|
1298
|
+
responses_items.extend(make_failed_tool_call_items(
|
|
1299
|
+
tool_name="computer",
|
|
1300
|
+
tool_kwargs=args,
|
|
1301
|
+
error_message=repr(e),
|
|
1205
1302
|
call_id=call_id
|
|
1206
1303
|
))
|
|
1207
1304
|
except json.JSONDecodeError:
|