# Multi-Agent: Supervisor Pattern Ref: - https://docs.langchain.com/oss/python/langchain/multi-agent The supervisor pattern uses a central controller agent that calls other agents as tools. **Characteristics:** - Centralized control flow - Sub-agents don't interact directly with users - Good for structured workflows **Example:** Supervisor coordinates a calculator agent and a writer agent to solve math problems and explain solutions. ## Setup Configure `.env` before running. See `.env.sample`. ```python import rich from dotenv import load_dotenv load_dotenv() ``` True ## Define Sub-Agents Create specialized agents for different tasks. ```python from typing import Any from langchain.agents import create_agent from langchain_anthropic import ChatAnthropic from langchain_core.tools import tool from langgraph.graph.state import CompiledStateGraph model = ChatAnthropic(model="claude-sonnet-4-5-20250929") # Calculator agent tools @tool def add(a: float, b: float) -> float: """Add two numbers.""" return a + b @tool def subtract(a: float, b: float) -> float: """Subtract b from a.""" return a - b @tool def multiply(a: float, b: float) -> float: """Multiply two numbers.""" return a * b @tool def divide(a: float, b: float) -> float: """Divide a by b.""" if b == 0: raise ValueError("Cannot divide by zero") return a / b calculator_agent: CompiledStateGraph[Any] = create_agent( model=model, tools=[add, subtract, multiply, divide], system_prompt="You are a calculator. Use your tools to perform calculations. Return only the numeric result.", ) # Writer agent (uses LLM directly for writing) writer_agent: CompiledStateGraph[Any] = create_agent( model=model, tools=[], system_prompt="""You are a writing assistant. Create clear explanations based on the information provided. Write concisely and format appropriately.""", ) rich.print("Sub-agents created: calculator_agent, writer_agent") ``` Sub-agents created: calculator_agent, writer_agent ## Wrap Sub-Agents as Tools Create tool wrappers that the supervisor can call. ```python @tool("calculate", parse_docstring=True) def call_calculator_agent(expression: str) -> str: """Perform mathematical calculations. Use this to compute arithmetic expressions, solve equations, or do any math. Args: expression: The math problem to solve (e.g., "What is 25 * 4 + 10?"). """ result = calculator_agent.invoke({"messages": [{"role": "user", "content": expression}]}) return str(result["messages"][-1].content) @tool("write", parse_docstring=True) def call_writer_agent(request: str) -> str: """Write or explain content. Use this to create explanations, summaries, or written content. Args: request: What to write, including context and any specific requirements. """ result = writer_agent.invoke({"messages": [{"role": "user", "content": request}]}) return str(result["messages"][-1].content) rich.print("Tool wrappers created: calculate, write") ``` Tool wrappers created: calculate, write ## Create Supervisor Agent The supervisor coordinates the sub-agents to accomplish complex tasks. ```python supervisor: CompiledStateGraph[Any] = create_agent( model=model, tools=[call_calculator_agent, call_writer_agent], system_prompt="""You are a supervisor agent that coordinates calculation and writing tasks. When given a task: 1. Use the calculate tool to perform any necessary math 2. Use the write tool to explain or summarize results 3. Synthesize the results into a final response""", ) rich.print("Supervisor agent created") ``` Supervisor agent created ## Test: Coordinated Task Ask the supervisor to complete a task that requires both calculation and explanation. ```python response = supervisor.invoke( {"messages": [{"role": "user", "content": "Calculate 15% tip on a $85 dinner bill and explain the calculation."}]} ) rich.print("Final response:") rich.print(response["messages"][-1].content) ``` Final response: ## Answer For a $85 dinner bill with a 15% tip: **The tip amount is $12.75** ### Calculation Breakdown: 1. **Convert percentage to decimal**: 15% = 0.15 2. **Multiply**: $85 × 0.15 = $12.75 ### Quick Mental Math Alternative: - 10% of $85 = $8.50 - 5% of $85 = $4.25 - Add together: $8.50 + $4.25 = $12.75 ### Total Payment: - Bill: $85.00 - Tip: $12.75 - **Total: $97.75** The mental math method is especially handy when you're at the restaurant without a calculator! ## Stream Execution Stream to see the supervisor's decision-making process. ```python query = "If I invest $1000 at 5% annual interest, how much will I have after 3 years? Explain compound interest." for chunk in supervisor.stream({"messages": [{"role": "user", "content": query}]}): rich.print("chunk =", chunk) ``` chunk = { 'model': { 'messages': [ AIMessage( content=[ { 'text': "I'll help you calculate your investment return and explain compound interest.", 'type': 'text' }, { 'id': 'toolu_01TJSLzdaUhvTFxdasBNznTY', 'input': {'expression': '1000 * (1 + 0.05)^3'}, 'name': 'calculate', 'type': 'tool_use' }, { 'id': 'toolu_014FwM5KW8uAAG1sqTPp2Wuk', 'input': { 'request': "Explain what compound interest is, how it works, and why it's important for investments. Make it clear and accessible for someone learning about investing." }, 'name': 'write', 'type': 'tool_use' } ], additional_kwargs={}, response_metadata={ 'id': 'msg_015UKgtZgqSPJ3F2TCBagQoo', 'model': 'claude-sonnet-4-5-20250929', 'stop_reason': 'tool_use', 'stop_sequence': None, 'usage': { 'cache_creation': {'ephemeral_1h_input_tokens': 0, 'ephemeral_5m_input_tokens': 0}, 'cache_creation_input_tokens': 0, 'cache_read_input_tokens': 0, 'input_tokens': 763, 'output_tokens': 141, 'server_tool_use': None, 'service_tier': 'standard' }, 'model_name': 'claude-sonnet-4-5-20250929', 'model_provider': 'anthropic' }, id='lc_run--0d3bebe4-9b13-4b55-b413-6753e62e411a-0', tool_calls=[ { 'name': 'calculate', 'args': {'expression': '1000 * (1 + 0.05)^3'}, 'id': 'toolu_01TJSLzdaUhvTFxdasBNznTY', 'type': 'tool_call' }, { 'name': 'write', 'args': { 'request': "Explain what compound interest is, how it works, and why it's important for investments. Make it clear and accessible for someone learning about investing." }, 'id': 'toolu_014FwM5KW8uAAG1sqTPp2Wuk', 'type': 'tool_call' } ], usage_metadata={ 'input_tokens': 763, 'output_tokens': 141, 'total_tokens': 904, 'input_token_details': { 'cache_read': 0, 'cache_creation': 0, 'ephemeral_5m_input_tokens': 0, 'ephemeral_1h_input_tokens': 0 } } ) ] } } chunk = { 'tools': { 'messages': [ ToolMessage( content='# Compound Interest: Your Money\'s Growth Engine\n\n## What It Is\n\nCompound interest is when you earn returns not just on your original investment, but also on the returns you\'ve already earned. In simple terms: **you earn interest on your interest**.\n\n## How It Works\n\nLet\'s say you invest $1,000 at 10% annual interest:\n\n**Year 1:** \n- You earn $100 (10% of $1,000)\n- New total: $1,100\n\n**Year 2:**\n- You earn $110 (10% of $1,100) — *notice it\'s more than Year 1*\n- New total: $1,210\n\n**Year 3:**\n- You earn $121 (10% of $1,210)\n- New total: $1,331\n\nEach year, your earnings grow because you\'re calculating interest on an increasingly larger amount.\n\n## Why It Matters for Investing\n\n**1. Time is your best friend**\nThe longer your money compounds, the more dramatic the growth. After 30 years at 10%, that $1,000 becomes over $17,000.\n\n**2. Small differences compound too**\nAn extra 2% annual return doesn\'t sound like much, but over decades, it can mean tens of thousands of dollars more.\n\n**3. Starting early beats investing more later**\nSomeone who invests $5,000 at age 25 can end up with more than someone who invests $10,000 at age 40, even though they put in less money.\n\n## The Key Takeaway\n\nAlbert Einstein allegedly called compound interest "the eighth wonder of the world." The sooner you start investing, the harder your money works for you—automatically.', name='write', id='bcd1feaa-1f25-485f-995b-ed3e0a20cfbf', tool_call_id='toolu_014FwM5KW8uAAG1sqTPp2Wuk' ) ] } } chunk = { 'tools': { 'messages': [ ToolMessage( content='1157.625', name='calculate', id='ca341ba0-ef91-4b4d-a851-60c1a2c7f94f', tool_call_id='toolu_01TJSLzdaUhvTFxdasBNznTY' ) ] } } chunk = { 'model': { 'messages': [ AIMessage( content='## Your Investment Results\n\n**After 3 years, your $1,000 investment at 5% annual compound interest will grow to $1,157.63**\n\nHere\'s the breakdown:\n- **Initial investment:** $1,000\n- **Interest rate:** 5% annually\n- **Time period:** 3 years\n- **Final amount:** $1,157.63\n- **Total interest earned:** $157.63\n\n### Year-by-Year Growth:\n- **Year 1:** $1,000 → $1,050.00 (earned $50)\n- **Year 2:** $1,050 → $1,102.50 (earned $52.50)\n- **Year 3:** $1,102.50 → $1,157.63 (earned $55.13)\n\nNotice how each year you earn slightly more interest than the previous year? That\'s compound interest at work!\n\n---\n\n# Compound Interest: Your Money\'s Growth Engine\n\n## What It Is\n\nCompound interest is when you earn returns not just on your original investment, but also on the returns you\'ve already earned. In simple terms: **you earn interest on your interest**.\n\n## How It Works\n\nLet\'s say you invest $1,000 at 10% annual interest:\n\n**Year 1:** \n- You earn $100 (10% of $1,000)\n- New total: $1,100\n\n**Year 2:**\n- You earn $110 (10% of $1,100) — *notice it\'s more than Year 1*\n- New total: $1,210\n\n**Year 3:**\n- You earn $121 (10% of $1,210)\n- New total: $1,331\n\nEach year, your earnings grow because you\'re calculating interest on an increasingly larger amount.\n\n## Why It Matters for Investing\n\n**1. Time is your best friend**\nThe longer your money compounds, the more dramatic the growth. After 30 years at 10%, that $1,000 becomes over $17,000.\n\n**2. Small differences compound too**\nAn extra 2% annual return doesn\'t sound like much, but over decades, it can mean tens of thousands of dollars more.\n\n**3. Starting early beats investing more later**\nSomeone who invests $5,000 at age 25 can end up with more than someone who invests $10,000 at age 40, even though they put in less money.\n\n## The Key Takeaway\n\nAlbert Einstein allegedly called compound interest "the eighth wonder of the world." The sooner you start investing, the harder your money works for you—automatically.', additional_kwargs={}, response_metadata={ 'id': 'msg_01DXYC52BQ4MBChytyjF4WT7', 'model': 'claude-sonnet-4-5-20250929', 'stop_reason': 'end_turn', 'stop_sequence': None, 'usage': { 'cache_creation': {'ephemeral_1h_input_tokens': 0, 'ephemeral_5m_input_tokens': 0}, 'cache_creation_input_tokens': 0, 'cache_read_input_tokens': 0, 'input_tokens': 1348, 'output_tokens': 596, 'server_tool_use': None, 'service_tier': 'standard' }, 'model_name': 'claude-sonnet-4-5-20250929', 'model_provider': 'anthropic' }, id='lc_run--cf76c322-a342-454e-9ae5-2404d8f54e5d-0', usage_metadata={ 'input_tokens': 1348, 'output_tokens': 596, 'total_tokens': 1944, 'input_token_details': { 'cache_read': 0, 'cache_creation': 0, 'ephemeral_5m_input_tokens': 0, 'ephemeral_1h_input_tokens': 0 } } ) ] } }