Current Service Prefabs

Classification Agent with Simple Efficient Memory Handling

The Classification Agent is central to managing conversation flows and categorizing topics with smart memory handling. It integrates group chat management and smart prompt-based function calls for effective message classification and agent response generation. Key features include:

• Smart Memory Handling: This agent avoids redundant message loops by maintaining a streamlined, efficient memory.

• Group Chat Functionality: The agent uses a group chat manager to coordinate conversations across multiple specialized agents.

  classification_agent = autogen.AssistantAgent(
      name="classification_agent",
      system_message="I am a classification agent responsible for analyzing the topic of conversations and routing them accordingly. "
                     "My responses are based strictly on the user input and I do not retain context from previous responses.",
      description="A core agent that classifies conversation topics and forwards them to specialized topic agents.",
      llm_config=llm_config,
  )
  

• Agent Topic Assignment: Multiple agents handle different conversation topics, such as soccer, tennis, or basketball. These agents are added dynamically to the classification flow:

  tennis_agent = autogen.AssistantAgent(
      name="tennis",
      system_message="You are a topic agent responsible for answering queries about tennis. "
                     "Your responses should be accurate, concise, and formatted for easy readability. "
                     "Do not retain or provide memory.",
      description="A topic agent specialized in tennis-related information.",
      llm_config=llm_config,
  )
  
  classification_agent.add_topic_agent(tennis_agent)
  

• Smart Routing: The classification agent routes messages based on detected topics to relevant topic agents, ensuring efficient conversation flow.

  soccer_agent = autogen.AssistantAgent(
      name="soccer",
      system_message="You are a topic agent responsible for answering queries about soccer. "
                     "Provide accurate and concise responses, without retaining memory or context.",
      description="A topic agent focused on soccer.",
      llm_config=llm_config,
  )
  
  classification_agent.add_topic_agent(soccer_agent)
  

Knowledge Retrieval Agent

The Knowledge Retrieval Agent excels in extracting specific information from a predefined knowledge base with smart function calls and routing. It employs dynamic index selection based on the query context and supports advanced retrieval configurations:

• Smart Index Selection: The agent selects the relevant index for retrieving information based on the query type. For instance, health-related queries will trigger a search in the “vector-health” index, while safety-related queries will use “vector-safety.”

  search_agent = autogen.AssistantAgent(
      name="search_agent",
      system_message="I am a search agent responsible for retrieving results from searches and passing them to the researcher. "
                     "My responses must be strictly based on the search results or guidelines, with no additional information. "
                     "If a query is ambiguous, I will search across all indices.",
      description="A search agent focused on accurate retrieval from the correct indices.",
      llm_config=llm_config,
  )
  

• Chunk-Based Retrieval: The agent uses token-based chunking for efficient retrieval from large documents.

• Smart Function Call and Termination: This agent is also capable of calling functions only when needed, and can terminate based on specific message content to avoid redundant queries.

Education Agent

The Education Agent is in early development but aims to support structured learning experiences by leveraging knowledge retrieval capabilities and personalized learning path generation.