Merge pull request #202 from XeonHis/develop/v0.2.3

Develop/v0.2.3: LLM payload fix, update DnC operator

Author: panregedit

examples/general_dnc/agent/conclude/conclude.py

@@ -50,9 +50,7 @@ def _run(self, dnc_structure: dict, last_output: str, *args, **kwargs):
         chat_complete_res = self.simple_infer(
             task=task.get_root().task,
             result=str(last_output),
-            img_placeholders="".join(
-                list(self.stm(self.workflow_instance_id).get("image_cache", {}).keys())
-            ),
+            img_placeholders=self.stm(self.workflow_instance_id).get("image_cache"),
         )
         if isinstance(chat_complete_res, Iterator):
             last_output = "Answer: "

examples/general_dnc/agent/input_interface/input_interface.py

@@ -35,5 +35,5 @@ def _run(self, *args, **kwargs):
             elif each_content["type"] == "text":
                 text = each_content["data"]
         if image is not None:
-            self.stm(self.workflow_instance_id)["image_cache"] = {f"<image_0>": image}
+            self.stm(self.workflow_instance_id)["image_cache"] = image
         return {"query": text}

examples/video_understanding/docs/images/local-ai.png

@@ -18,6 +18,7 @@
 from pydantic import Field
 from tenacity import (retry, retry_if_exception_message, stop_after_attempt,
                       stop_after_delay)
+from omagent_core.utils.logger import logging
 
 CURRENT_PATH = Path(__file__).parents[0]
 
@@ -96,13 +97,12 @@ def _run(self, dnc_structure: dict, last_output: str, *args, **kwargs):
             ),
             "former_results": self.stm(self.workflow_instance_id)["former_results"],
             "extra_info": self.stm(self.workflow_instance_id).get("extra"),
-            "img_placeholders": "".join(
-                list(self.stm(self.workflow_instance_id).get("image_cache", {}).keys())
-            ),
+            "img_placeholders": self.stm(self.workflow_instance_id).get("image_cache"),
         }
 
         # Call LLM to get next actions or task completion results
         chat_complete_res = self.infer(input_list=[payload])
+        logging.info(f"Conqueror chat_complete_res: {chat_complete_res}")
         content = chat_complete_res[0]["choices"][0]["message"].get("content")
         content = json_repair.loads(content)
 

omagent-core/src/omagent_core/advanced_components/workflow/dnc/agent/conqueror/conqueror.py

@@ -14,6 +14,7 @@
 from pydantic import Field
 from tenacity import (retry, retry_if_exception_message, stop_after_attempt,
                       stop_after_delay)
+from omagent_core.utils.logger import logging
 
 CURRENT_PATH = Path(__file__).parents[0]
 
@@ -85,6 +86,7 @@ def _run(self, dnc_structure: dict, last_output: str, *args, **kwargs):
             former_results=last_output,
             tools=self.tool_manager.generate_prompt(),
         )
+        logging.info(f"Divider chat_complete_res: {chat_complete_res}")
         chat_complete_res = json_repair.loads(
             chat_complete_res["choices"][0]["message"]["content"]
         )

omagent-core/src/omagent_core/advanced_components/workflow/dnc/agent/divider/divider.py

@@ -151,7 +151,7 @@ def content_validator(
             raise ValueError(
                 "Content must be a string, a list of Content objects or list of dicts."
             )
-        return formatted
+        return formatted[0] if len(formatted) == 1 else formatted
 
     @classmethod
     def system(cls, content: str | List[str | Dict | Content]) -> "Message":

omagent-core/src/omagent_core/models/llms/schemas.py

@@ -247,6 +247,7 @@ def execute_task(self, task, related_info="", function=None):
             [{"task": task, "related_info": related_info}],
             tools=self.generate_schema(),
         )[0]
+        logging.info(f"ToolManager execute_task chat_complete_res: {chat_complete_res}")
         content = chat_complete_res["choices"][0]["message"].get("content")
         tool_calls = chat_complete_res["choices"][0]["message"].get("tool_calls")
         if not tool_calls:
@@ -318,6 +319,7 @@ async def aexecute_task(self, task, related_info=None, function=None):
             [{"task": task, "related_info": list(related_info.keys())}],
             tools=self.generate_schema(),
         )[0]
+        logging.info(f"ToolManager aexecute_task chat_complete_res: {chat_complete_res}")
         content = chat_complete_res["choices"][0]["message"].get("content")
         tool_calls = chat_complete_res["choices"][0]["message"].get("tool_calls")
         if not tool_calls: