Final modifications before final run

mini_agents.py

@@ -113,13 +113,13 @@ context_search_model = InferenceClientModel(
 
 context_search_agent = CodeAgent(
     model=context_search_model,
-    tools=[*search_tools, tavily_search_tool],
+    tools=[*search_tools],
     max_steps=4,
     additional_authorized_imports=AUTHORIZED_IMPORTS,
     planning_interval=4,
     name="context_search_agent",
     prompt_templates=PROMPT_TEMPLATE["context_search_agent"],
-    description="This agent is responsible for searching the web for context using wikipedia for general information and arxiv for scientific information. It can also use Tavily to search the web for more relevantcontext."
+    description="This agent is responsible for searching the web for context using wikipedia for general information and arxiv for scientific information."
 )
 
 master_model = InferenceClientModel(
@@ -137,6 +137,7 @@ class MasterAgentWrapper:
             read_python_file_from_path, 
             PythonInterpreterTool(),
             DuckDuckGoSearchTool(),
+            tavily_search_tool,
             *community_tools,
         ]
         

prompts/prompts.yaml

@@ -157,7 +157,7 @@ system_prompt: |-
   ```
 
   {%- if managed_agents and managed_agents.values() | list %}
-  You should also give tasks to team members whenever possible. They are very useful to solve tasks that require multiple modalities.
+  You should also give tasks to team members whenever possible. They are very useful to solve tasks that require multiple modalities. One member is also adept at searching wikipedia and arxiv for information.
   Calling a team member works the same as for calling a tool: you can simply pass the task description as a string to the member without specifying any argument.
   Given that this team member is a real human, you should be very verbose in your task, it should be a long string providing informations as detailed as necessary.
   Here is a list of the team members that you can call: