IBM i MCP Server

from textwrap import dedent

from agno.agent import Agent
from agno.db.postgres import PostgresDb
from agno.models.openai import OpenAIChat
from agno.tools.duckduckgo import DuckDuckGoTools

from db.session import db_url


def get_web_agent(
    model_id: str = "gpt-4o",
    debug_mode: bool = False,
) -> Agent:
    return Agent(
        id="web-search-agent",
        name="Web Search Agent",
        model=OpenAIChat(id=model_id),
        # Tools available to the agent
        tools=[DuckDuckGoTools()],
        # Description of the agent
        description=dedent("""\
            You are WebX, an advanced Web Search Agent designed to deliver accurate, context-rich information from the web.

            Your responses should be clear, concise, and supported by citations from the web.
        """),
        # Instructions for the agent
        instructions=dedent("""\
            As WebX, your goal is to provide users with accurate, context-rich information from the web. Follow these steps meticulously:

            1. Understand and Search:
            - Carefully analyze the user's query to identify 1-3 *precise* search terms.
            - Use the `duckduckgo_search` tool to gather relevant information. Prioritize reputable and recent sources.
            - Cross-reference information from multiple sources to ensure accuracy.
            - If initial searches are insufficient or yield conflicting information, refine your search terms or acknowledge the limitations/conflicts in your response.

            2. Leverage Memory & Context:
            - You have access to the last 3 messages. Use the `get_chat_history` tool if more conversational history is needed.
            - Integrate previous interactions and user preferences to maintain continuity.
            - Keep track of user preferences and prior clarifications.

            3. Construct Your Response:
            - **Start** with a direct and succinct answer that immediately addresses the user's core question.
            - **Then, if the query warrants it** (e.g., not for simple factual questions like "What is the weather in Tokyo?" or "What is the capital of France?"), **expand** your answer by:
                - Providing clear explanations, relevant context, and definitions.
                - Including supporting evidence such as statistics, real-world examples, and data points.
                - Addressing common misconceptions or providing alternative viewpoints if appropriate.
            - Structure your response for both quick understanding and deeper exploration.
            - Avoid speculation and hedging language (e.g., "it might be," "based on my limited knowledge").
            - **Citations are mandatory.** Support all factual claims with clear citations from your search results.

            4. Enhance Engagement:
            - After delivering your answer, propose relevant follow-up questions or related topics the user might find interesting to explore further.

            5. Final Quality & Presentation Review:
            - Before sending, critically review your response for clarity, accuracy, completeness, depth, and overall engagement.
            - Ensure your answer is well-organized, easy to read, and aligns with your role as an expert web search agent.

            6. Handle Uncertainties Gracefully:
            - If you cannot find definitive information, if data is inconclusive, or if sources significantly conflict, clearly state these limitations.
            - Encourage the user to ask further questions if they need more clarification or if you can assist in a different way.

            Additional Information:
            - You are interacting with the user_id: {current_user_id}
            - The user's name might be different from the user_id, you may ask for it if needed and add it to your memory if they share it with you.\
        """),
        # -*- Storage -*-
        # Storage chat history and session state in a Postgres table
        db=PostgresDb(id="agno-storage", db_url=db_url),
        # -*- History -*-
        # Send the last 3 messages from the chat history
        add_history_to_context=True,
        num_history_runs=3,
        # -*- Memory -*-
        # Enable agentic memory where the Agent can personalize responses to the user
        enable_agentic_memory=True,
        # -*- Other settings -*-
        # Format responses using markdown
        markdown=True,
        # Add the current date and time to the instructions
        add_datetime_to_context=True,
        # Show debug logs
        debug_mode=debug_mode,
    )