Agent Generate Structured Output Examples

!pip install "camel-ai[all]==0.2.70"

import os
from getpass import getpass

# Prompt for the AgentOps API key securely
openai_api_key = getpass('Enter your API key: ')
os.environ["OPENAI_API_KEY"] = openai_api_key

from pydantic import BaseModel, Field
from camel.agents import ChatAgent
from camel.models import ModelFactory
from camel.types import ModelPlatformType, ModelType

# Define system message
assistant_sys_msg = "You are a helpful assistant."

# Initialize model and agent
model = ModelFactory.create(
    model_platform=ModelPlatformType.DEFAULT,
    model_type=ModelType.DEFAULT,
)
agent = ChatAgent(assistant_sys_msg, model=model)

# Define the expected response structure using Pydantic
class JokeResponse(BaseModel):
    joke: str = Field(description="A funny joke")
    funny_level: int = Field(description="How funny the joke is, from 1 to 10")

response = agent.step("Tell me a joke.", response_format=JokeResponse)


# Display the results
print("=== Raw Response ===")
print(response.msgs[0].content)

print("\n=== Parsed Object ===")
print(response.msgs[0].parsed)

print("\n=== Type of Parsed Object ===")
print(type(response.msgs[0].parsed))

from pydantic import BaseModel, Field
from camel.agents import ChatAgent
from camel.configs.openai_config import ChatGPTConfig
from camel.models import ModelFactory
from camel.toolkits import MathToolkit, SearchToolkit
from camel.types import ModelPlatformType, ModelType

# Use wiki tool to avoid additional api config
search_tools = SearchToolkit().get_tools()
wiki_tool = search_tools[0]

tools_list = [
    *MathToolkit().get_tools(),  # Adds math calculation capabilities
    wiki_tool,  # Adds web search capabilities
]

# Configure model with specific settings
assistant_model_config = ChatGPTConfig(
    temperature=0.0,  # Use low temperature for more deterministic outputs
)

# Define system message
assistant_sys_msg = "You are a helpful assistant that's good at calculations and research."

# Cell 3: Initialize the model
model = ModelFactory.create(
    model_platform=ModelPlatformType.DEFAULT,
    model_type=ModelType.DEFAULT,
    model_config_dict=assistant_model_config.as_dict(),
)

# Initialize agent with tools
camel_agent = ChatAgent(
    assistant_sys_msg,
    model=model,
    tools=tools_list,  # Pass the tools to the agent
)

# Define the structured response format
class CalculationResult(BaseModel):
    current_age: str = Field(description="The current age being calculated")
    calculated_age: str = Field(description="The age after adding years")
    calculation_steps: str = Field(description="Detailed steps of the calculation")

# Define the user's question
user_msg = """Assume now is 2024 in the Gregorian calendar, 
estimate the current age of University of Oxford 
and then add 10 more years to this age."""

# Get the structured response
response = camel_agent.step(
    user_msg, 
    response_format=CalculationResult
)

# Display the results
print("=== Raw Response ===")
print(response.msgs[0].content)

print("\n=== Parsed Object ===")
print(response.msgs[0].parsed)

print("\n=== Accessing Fields ===")
print(f"Current age: {response.msgs[0].parsed.current_age}")
print(f"Calculated age: {response.msgs[0].parsed.calculated_age}")
print(f"\nCalculation steps:\n{response.msgs[0].parsed.calculation_steps}")

from pydantic import BaseModel, Field
from typing import List, Optional
from camel.agents import ChatAgent
from camel.models import ModelFactory
from camel.types import ModelPlatformType, ModelType
import json

# Define Pydantic models for our data structure
# Define our data models
class Ingredient(BaseModel):
    name: str
    amount: str
    unit: str

class RecipeStep(BaseModel):
    step_number: int
    instruction: str
    duration: str

class Recipe(BaseModel):
    name: str
    description: str
    prep_time: str
    cook_time: str
    servings: int
    ingredients: List[Ingredient]
    instructions: List[RecipeStep]
    dietary_info: List[str] = Field(description="List of dietary categories")

# Cell 2: Initialize the default model
model = ModelFactory.create(
    model_platform=ModelPlatformType.DEFAULT,
    model_type=ModelType.GPT_3_5_TURBO,
    api_key=openai_api_key,
)

agent = ChatAgent("You are a professional chef assistant.")

# Function to demonstrate manual parsing
def generate_recipe(dish: str) -> Recipe:
    # First, get a structured response
    response = agent.step(
        f"Provide a detailed recipe for {dish} in JSON format with the following structure: "
        "{\"name\": \"...\", \"description\": \"...\", \"prep_time\": \"...\", "
        "\"cook_time\": \"...\", \"servings\": 0, \"ingredients\": [{\"name\": \"...\", "
        "\"amount\": \"...\", \"unit\": \"...\"}], \"instructions\": [{\"step_number\": 1, "
        "\"instruction\": \"...\", \"duration\": \"...\"}], \"dietary_info\": [\"...\"]}\n\n"
        "Return ONLY the JSON object, without any additional text or markdown formatting."
    )
    
    try:
        # Extract JSON from the response
        content = response.msgs[0].content.strip()
        if content.startswith("```json"):
            content = content[7:-3].strip()  # Remove markdown code block if present
        
        # Parse and validate the response
        recipe_data = json.loads(content)
        return Recipe(**recipe_data)
    except Exception as e:
        print(f"Error parsing response: {e}")
        print("Raw response:", response.msgs[0].content)
        raise

# Cell 4: Generate and display a recipe
try:
    recipe = generate_recipe("vegetable lasagna")
    
    print(f"=== {recipe.name.upper()} ===")
    print(recipe.description)
    print(f"\nPreparation: {recipe.prep_time} | Cooking: {recipe.cook_time} | Servings: {recipe.servings}")
    
    print("\nINGREDIENTS:")
    for ing in recipe.ingredients:
        print(f"- {ing.amount} {ing.unit} {ing.name}")
    
    print("\nINSTRUCTIONS:")
    for step in recipe.instructions:
        print(f"{step.step_number}. {step.instruction} ({step.duration})")
    
    print("\nDIETARY INFO:", ", ".join(recipe.dietary_info))

except Exception as e:
    print(f"Failed to generate recipe: {e}")

# Cell 5: Alternative approach - Using response_format with the default model
# This shows how it would work with a model that supports structured output
try:
    response = agent.step(
        "Give me a recipe for vegetable lasagna",
        response_format=Recipe
    )
    
    print("\n=== Using response_format ===")
    print("Recipe name:", response.msgs[0].parsed.name)
    print("First ingredient:", response.msgs[0].parsed.ingredients[0].name)
    
except Exception as e:
    print("\nNote: The default model might not support structured output natively.")
    print("Error:", e)