Agent API Reference¶

The Agent class is the core of AgentiCraft, providing intelligent AI capabilities with tool usage, memory, and provider flexibility.

Agent¶

from agenticraft import Agent

agent = Agent(
    name="MyAgent",
    model="gpt-4",
    provider="openai",  # Optional, auto-detected from model
    **kwargs
)

Parameters¶

Parameter	Type	Default	Description
`name`	`str`	required	Unique name for the agent
`model`	`str`	`"gpt-4"`	Model to use
`provider`	`str`	`None`	LLM provider (auto-detected if None)
`tools`	`List[Tool]`	`[]`	Tools available to the agent
`memory_enabled`	`bool`	`False`	Enable conversation memory
`system_prompt`	`str`	`None`	System instructions
`temperature`	`float`	`0.7`	Sampling temperature
`max_tokens`	`int`	`None`	Maximum response tokens

Methods¶

run(prompt: str) -> Response¶

Execute the agent with a prompt.

response = agent.run("Hello, how are you?")
print(response.content)

set_provider(provider: str, model: str, **kwargs)¶

Switch to a different LLM provider at runtime.

agent.set_provider("anthropic", model="claude-3-opus-20240229")

get_provider_info() -> Dict[str, Any]¶

Get information about the current provider.

info = agent.get_provider_info()
# {'provider': 'openai', 'model': 'gpt-4', 'temperature': 0.7}

list_available_providers() -> List[str]¶

List all available providers.

providers = agent.list_available_providers()
# ['openai', 'anthropic', 'ollama']

ReasoningAgent¶

An agent that provides transparent reasoning traces.

from agenticraft import ReasoningAgent

agent = ReasoningAgent(
    name="Thinker",
    model="gpt-4",
    reasoning_style="chain_of_thought"
)

response = agent.run("Analyze this problem...")
print(response.reasoning)  # List of reasoning steps
print(response.confidence)  # Confidence score

Additional Parameters¶

Parameter	Type	Default	Description
`reasoning_style`	`str`	`"chain_of_thought"`	Reasoning approach
`explore_branches`	`int`	`1`	Branches for tree_of_thought
`enable_self_critique`	`bool`	`False`	Enable self-reflection

WorkflowAgent¶

An agent optimized for multi-step workflows.

from agenticraft import WorkflowAgent, Step

agent = WorkflowAgent(name="Processor", model="gpt-4")

workflow = [
    Step("analyze", "Analyze the data"),
    Step("process", "Process the results"),
    Step("report", "Generate report")
]

result = agent.run_workflow("Process sales data", workflow)

Response Objects¶

Response¶

Basic response from an agent.

@dataclass
class Response:
    content: str  # The response text
    metadata: Dict[str, Any]  # Additional metadata

ReasoningResponse¶

Response from a ReasoningAgent.

@dataclass 
class ReasoningResponse(Response):
    reasoning: List[str]  # Reasoning steps
    confidence: float  # Confidence score (0-1)
    assumptions: List[str]  # Assumptions made

WorkflowResponse¶

Response from a WorkflowAgent.

@dataclass
class WorkflowResponse(Response):
    steps: Dict[str, StepResult]  # Results by step name
    duration: float  # Total execution time

Examples¶

Basic Usage¶

from agenticraft import Agent

# Simple agent
agent = Agent(name="Assistant", model="gpt-4")
response = agent.run("Tell me a joke")
print(response.content)

With Tools¶

from agenticraft import Agent, tool

@tool
def calculate(expression: str) -> float:
    return eval(expression)

agent = Agent(name="MathBot", tools=[calculate])
response = agent.run("What's 15 * 23?")

Provider Switching¶

# Start with GPT-4
agent = Agent(name="Flex", model="gpt-4")
response = agent.run("Complex analysis...")

# Switch to cheaper model
agent.set_provider("ollama", model="llama2")
response = agent.run("Simple summary...")

With Memory¶

agent = Agent(name="MemBot", memory_enabled=True)

agent.run("My name is Alice")
response = agent.run("What's my name?")
# Agent remembers: "Your name is Alice"