update react demo

cookbook/simple_agent/experience.jsonl

@@ -0,0 +1,30 @@
+[
+  {
+    "workspace_id": "test_workspace1",
+    "experience_id": "e249d97f56dd452badd764a81dc22709",
+    "experience_type": "text",
+    "when_to_use": "When analyzing a large, multifaceted corporation across various dimensions such as financials, market position, and technological innovations.",
+    "content": "Breaking down the problem into smaller subtasks (e.g., financial performance, market standing, recent developments) allows for a more systematic and comprehensive analysis. Using multiple `web_search` calls with specific queries ensures that each dimension is thoroughly covered from different angles.",
+    "score": null,
+    "metadata": {
+      "author": "qwen-max-2025-01-25",
+      "created_time": "2025-07-21 16:35:35",
+      "modified_time": "2025-07-21 16:35:35",
+      "extra_info": null
+    }
+  },
+  {
+    "workspace_id": "test_workspace1",
+    "experience_id": "c13abd248d1a48e89582f0177a7a0f84",
+    "experience_type": "text",
+    "when_to_use": "When information retrieved seems sufficient but fragmented across several sources and needs to be synthesized into a coherent response.",
+    "content": "After gathering data through various tool calls, it is crucial to integrate all relevant pieces of information into a unified narrative. This involves cross-referencing the content, ensuring consistency, and presenting the findings in a structured manner so that the final output directly addresses the user’s question without leaving gaps or introducing redundancies.",
+    "score": null,
+    "metadata": {
+      "author": "qwen-max-2025-01-25",
+      "created_time": "2025-07-21 16:35:35",
+      "modified_time": "2025-07-21 16:35:35",
+      "extra_info": null
+    }
+  }
+]

cookbook/simple_agent/react_demo.py

@@ -5,11 +5,10 @@ from dotenv import load_dotenv
 
 load_dotenv()
 base_url = "http://0.0.0.0:8001/"
-workspace_id = "test_workspace"
+workspace_id = "test_workspace1"
 
 
 def run_agent(query: str, dump_messages: bool = False):
-    query = "Analyze Xiaomi Corporation"
 
     response = requests.post(url=base_url + "agent", json={"query": query})
     if response.status_code != 200:
@@ -30,7 +29,7 @@ def run_agent(query: str, dump_messages: bool = False):
 
 
 def run_summary(messages: list, dump_experience: bool = True):
-    response = requests.post(url=base_url + "summary", json={
+    response = requests.post(url=base_url + "summarizer", json={
         "workspace_id": workspace_id,
         "traj_list": [
             {"messages": messages, "score": 1.0}
@@ -65,8 +64,8 @@ def run_retriever(query: str):
 
 
 def run_agent_with_experience(query_first: str, query_second: str, dump_experience: bool = True):
-    messages = run_agent(query=query_second)
-    run_summary(messages, dump_experience)
+    # messages = run_agent(query=query_second)
+    # run_summary(messages, dump_experience)
     experience_merged = run_retriever(query_first)
     messages = run_agent(query=f"{experience_merged}\n\nUser Question:\n{query_first}")
     return messages

cookbook/simple_agent/your_own_agent_enhanced.py

@@ -1,70 +0,0 @@
-from loguru import logger
-from pydantic import Field, model_validator
-
-from cookbook.simple_agent.your_own_agent import YourOwnAgent
-from experiencemaker.em_client import EMClient
-from experiencemaker.schema.request import ContextGeneratorRequest, SummarizerRequest
-from experiencemaker.schema.response import ContextGeneratorResponse, SummarizerResponse
-from experiencemaker.schema.trajectory import Trajectory
-
-
-class YourOwnAgentEnhanced(YourOwnAgent):
-    em_client: EMClient | None = Field(default=None)
-    workspace_id: str = Field(default=...)
-
-    @model_validator(mode="after")
-    def init_client(self):
-        self.em_client = EMClient(base_url="http://0.0.0.0:8001")
-        return self
-
-    def summary_experience(self, query: str):
-        messages = self.run(query)
-        trajectory: Trajectory = Trajectory(query=query, steps=messages, answer=messages[-1].content, is_terminated=True)
-
-        request: SummarizerRequest = SummarizerRequest(trajectories=[trajectory], workspace_id=self.workspace_id)
-        response: SummarizerResponse = self.em_client.call_summarizer(request)
-        for experience in response.experiences:
-            logger.info(experience.model_dump_json())
-        return response.experiences
-
-    def run_with_experience(self, query: str):
-        trajectory: Trajectory = Trajectory(query=query)
-        request: ContextGeneratorRequest = ContextGeneratorRequest(trajectory=trajectory,
-                                                                   retrieve_top_k=1,
-                                                                   workspace_id=self.workspace_id)
-        response: ContextGeneratorResponse = self.em_client.call_context_generator(request)
-        new_query = f"{response.context_msg.content}\n\nUser Question\n{query}"
-        logger.info(f"new query={new_query}")
-        messages = self.run(new_query)
-
-        trajectory.steps = messages
-        trajectory.answer = messages[-1].content
-        trajectory.is_terminated = True
-        trajectory.metadata["experience"] = response.context_msg.content
-        return trajectory
-
-    def execute(self):
-        self.summary_experience(query="Analyze the company Tesla.")
-        # self.summary_experience(query="Analyze the company Apple.")
-
-        return self.run_with_experience(query="Analyze Xiaomi Corporation.")
-
-
-if __name__ == "__main__":
-    # agent = YourOwnAgentEnhanced(workspace_id="w_agent_enhanced",
-    #                              llm=OpenAICompatibleBaseLLM(model_name="qwen3-32b", temperature=0.00001))
-    # traj = agent.execute()
-    # logger.info(traj.model_dump_json(indent=2))
-
-    em_client = EMClient(base_url="http://0.0.0.0:8001")
-
-
-    request: ContextGeneratorRequest = ContextGeneratorRequest(trajectory=Trajectory(query="hello"),
-                                                               workspace_id="w123")
-    response = em_client.call_summarizer()
-    print(response)
-
-
-    request: ContextGeneratorRequest = ContextGeneratorRequest(trajectory=Trajectory(query="hello"), workspace_id="w123")
-    response = em_client.call_context_generator(request=request)
-    print(response)

cookbook/simple_agent/your_own_agent_prompt.yaml

@@ -1,27 +0,0 @@
-role_prompt: |
-  You are a helpful assistant named BeyondAgent.
-  The current time is {time}.
-  
-  Please proactively choose the most suitable tool or combination of tools based on the user's question, including {tools} etc.
-  For complex tasks, you can break down the problem step by step and use different tools to solve it incrementally.
-  Please determine the response language based on the language of the user's question.
-  
-  {query}
-
-next_prompt: |
-  User question
-  {query}
-  
-  Plan the most suitable tool or combination of tools based on the context and the user's question. 
-  For complex tasks, break down the problem and use different tools step by step to solve it, don't give up easily.  
-  Try calling the same tool multiple times with different parameters to obtain information from various perspectives.  
-  If the task is completed and the user's question can now be answered, use the **terminate** tool.
-  Please determine the response language based on the language of the user's question.
-
-final_prompt: |
-  Please integrate the context and provide a complete answer to the user's question.
-  Please determine the response language based on the language of the user's question.
-
-  User question
-  {query}
-

experiencemaker/op/react/react_v1_prompt.yaml

@@ -6,7 +6,6 @@ role_prompt: |
   Try calling the same tool multiple times with different parameters to obtain information from various perspectives.
   Please determine the response language based on the language of the user's question.
   
-  # User's Question
   {query}
 
 # write a complete and rigorous report to answer user's questions based on the context.
@@ -19,11 +18,10 @@ next_prompt: |
     Please first provide the reasoning process, then give the tool call name and parameters.
   
   - If the current context is sufficient to answer the user's question, use the **terminate** tool.
-    Please determine the response language based on the language of the user's question.
 
+#   Please determine the response language based on the language of the user's question.
 final_prompt: |
   Please integrate the context and provide a complete answer to the user's question.
-  Please determine the response language based on the language of the user's question.
 
   # User's Question
   {query}