refactor: revise workflow for react agent (agiresearch#197)

aios/llm_core/llm_classes/gemini_llm.py

@@ -98,7 +98,7 @@ def process(self,
                     )
             else:
                 if message_return_type == "json":
-                    result = self.json_parse_format(result)
+                    result = self.parse_json_format(result)
                 agent_process.set_response(
                     Response(
                         response_message=result,

aios/llm_core/llm_classes/vllm.py

@@ -80,6 +80,8 @@ def process(self,
             # print(response)
             result = response[0].outputs[0].text
 
+            print(f"***** Result: {result} *****")
+
             tool_calls = self.parse_tool_calls(
                 result
             )

pyopenagi/agents/base_agent.py

@@ -39,6 +39,7 @@ def __init__(self,
                  agent_process_factory,
                  log_mode: str
         ):
+
         self.agent_name = agent_name
         self.config = self.load_config()
         self.tool_names = self.config["tools"]
@@ -47,6 +48,8 @@ def __init__(self,
 
         self.tool_list = dict()
         self.tools = []
+        self.tool_info = [] # simplified information of the tool: {"name": "xxx", "description": "xxx"}
+
         self.load_tools(self.tool_names)
 
         self.start_time = None
@@ -130,16 +133,19 @@ def snake_to_camel(self, snake_str):
         return ''.join(x.title() for x in components)
 
     def load_tools(self, tool_names):
+
         for tool_name in tool_names:
             org, name = tool_name.split("/")
             module_name = ".".join(["pyopenagi", "tools", org, name])
             class_name = self.snake_to_camel(name)
-
             tool_module = importlib.import_module(module_name)
             tool_class = getattr(tool_module, class_name)
-
             self.tool_list[name] = tool_class()
-            self.tools.append(tool_class().get_tool_call_format())
+            tool_format = tool_class().get_tool_call_format()
+            self.tools.append(tool_format)
+            self.tool_info.append(
+                {"name": tool_format["function"]["name"], "description": tool_format["function"]["description"]}
+            )
 
     def pre_select_tools(self, tool_names):
         pre_selected_tools = []

pyopenagi/agents/react_agent.py

@@ -22,8 +22,6 @@ def __init__(self,
             log_mode
         )
 
-        # self.tool_list = {}
-
         self.plan_max_fail_times = 3
         self.tool_call_max_fail_times = 3
 
@@ -33,10 +31,11 @@ def build_system_instruction(self):
                 "".join(self.config["description"])
             ]
         )
+
         plan_instruction = "".join(
             [
-                f'You are given the available tools from the tool list: {json.dumps(self.tools)} to help you solve problems.',
-                'Generate a plan of steps you need to take.',
+                f'You are given the available tools from the tool list: {json.dumps(self.tool_info)} to help you solve problems. ',
+                'Generate a plan of steps you need to take. ',
                 'The plan must follow the json format as: ',
                 '[',
                 '{"message": "message_value1","tool_use": [tool_name1, tool_name2,...]}',
@@ -46,45 +45,27 @@ def build_system_instruction(self):
                 'In each step of the planned workflow, you must select the most related tool to use',
                 'Followings are some plan examples:',
                 '[',
-                '{"message": "Gather information from arxiv", "tool_use": ["arxiv"]},',
-                '{"message", "Based on the gathered information, write a summarization", "tool_use": []}',
-                '];',
-                '[',
-                '{"message": "identify the tool that you need to call to obtain information.", "tool_use": ["imdb_top_movies", "imdb_top_series"]},',
-                '{"message", "based on the information, give recommendations for the user based on the constrains.", "tool_use": []}',
+                '{"message": "gather information from arxiv. ", "tool_use": ["arxiv"]},',
+                '{"message", "write a summarization based on the gathered information. ", "tool_use": []}',
                 '];',
                 '[',
-                '{"message": "identify the tool that you need to call to obtain information.", "tool_use": ["imdb_top_movies", "imdb_top_series"]},',
-                '{"message", "based on the information, give recommendations for the user based on the constrains.", "tool_use": []}',
+                '{"message": "identify the tool that you need to call to obtain information. ", "tool_use": ["imdb_top_movies", "imdb_top_series"]},',
+                '{"message", "give recommendations for the user based on the information. ", "tool_use": []}',
                 '];',
-                '[',
-                '{"message": "identify the tool that you need to call to obtain information.", "tool_use": ["imdb_top_movies", "imdb_top_series"]},'
-                '{"message", "based on the information, give recommendations for the user based on the constrains.", "tool_use": []}',
-                ']'
             ]
         )
-        # exection_instruction = "".join(
-        #     [
-        #         'To execute each step in the workflow, you need to output as the following json format:',
-        #         '{"[Action]": "Your action that is indended to take",',
-        #         '"[Observation]": "What will you do? If you will call an external tool, give a valid tool call of the tool name and tool parameters"}'
-        #     ]
-        # )
+
         if self.workflow_mode == "manual":
             self.messages.append(
                 {"role": "system", "content": prefix}
             )
-            # self.messages.append(
-            #     {"role": "user", "content": exection_instruction}
-            # )
+
         else:
             assert self.workflow_mode == "automatic"
             self.messages.append(
                 {"role": "system", "content": prefix + plan_instruction}
             )
-            # self.messages.append(
-            #     {"role": "user", "content": plan_instruction}
-            # )
+
 
     def automatic_workflow(self):
         return super().automatic_workflow()
@@ -136,6 +117,10 @@ def run(self):
             {"role": "assistant", "content": f"[Thinking]: The workflow generated for the problem is {json.dumps(workflow)}"}
         )
 
+        self.messages.append(
+            {"role": "user", "content": "[Thinking]: Follow the workflow to solve the problem step by step. "}
+        )
+
         self.logger.log(f"Generated workflow is: {workflow}\n", level="info")
 
         if workflow:

pyopenagi/tools/stability-ai/sdxl_turbo.py

@@ -19,7 +19,7 @@ def get_tool_call_format(self):
             "type": "function",
             "function": {
                 "name": "sdxl_turbo",
-                "description": "generate images by calling sdxl-turbo model",
+                "description": "generate images with the given texts",
                 "parameters": {
                     "type": "object",
                     "properties": {