Agent answers unreasonably

General
,
1

Here’s an answer from my agent. As you can see, the answer is irrelevant.

image
image654×690 64 KB

Below is the code I use to define the agent

agent = Agent(
    name="Test",
    agent_id = "test-agent",
    model=Ollama(id="llama3.2:3b"),
    # tools=[DuckDuckGo()],
    # show_tool_calls=True,
    markdown=True,
    # debug_mode = True,
    
    add_history_to_messages=True,
    read_chat_history = True,
    num_history_responses=5,
    memory = AgentMemory(
        db=PgMemoryDb(
            table_name="agent_memory",
            db_url="postgresql+psycopg://ai:ai@localhost:5532/ai"),
            create_user_memories=True,
            create_session_summary=True
            ),
    storage = SqlAgentStorage(table_name="agent_sessions", db_file="database.db"),
    session_id="unique_session_id",

    knowledge=combined_knowledge_base,
    search_knowledge=True,
    # add_context=True,
    add_datetime_to_instructions = True,
    prevent_hallucinations=True,
)

How can I fix this?

2

Hey @Seitoku, it’s a 3b model so the answer will not always be accurate. Maybe you can try to use an 8b model

1 Like
3

Thanks! I’ll do that.