Agent API

Class: Agent

Module: djangosdk.agents.base

Base class for all AI agents. Inherits from Promptable, ReasoningMixin, HasTools, HasStructuredOutput, and Conversational.

Class Attributes

Attribute	Type	Default	Description
provider	str	""	Provider name from AI_SDK.PROVIDERS. Falls back to DEFAULT_PROVIDER.
model	str	""	Model identifier. Falls back to DEFAULT_MODEL.
system_prompt	str	""	System prompt prepended to every request.
temperature	float	0.7	Sampling temperature.
max_tokens	int	2048	Maximum tokens in the response.
reasoning	ReasoningConfig | None	None	Reasoning config for o3, Claude 3.7, DeepSeek R1.
enable_cache	bool	True	Enable prompt caching.
max_tool_iterations	int	10	Maximum tool-call rounds per handle() call.
tools	list	[]	List of @tool functions or BaseTool instances.
output_schema	type[BaseModel] | None	None	Pydantic model for structured output.
mcp_servers	list[dict]	[]	MCP server connection configs.

Instance Methods

handle(prompt: str, **kwargs) -> AgentResponse

Synchronous completion with tool-call loop.

• Fires agent_started and agent_completed (or agent_failed) signals

• Runs the tool loop until no more tool calls are returned (up to max_tool_iterations)

• Persists messages if bound to a conversation

ahandle(prompt: str, **kwargs) -> Coroutine[AgentResponse]

Async version of handle().

stream(prompt: str, **kwargs) -> StreamingHttpResponse

Synchronous streaming. Returns a Django StreamingHttpResponse with SSE content.

astream(prompt: str, **kwargs) -> AsyncGenerator[StreamChunk, None]

Async streaming generator.

start_conversation(metadata=None) -> Conversation

Creates a new Conversation ORM record and binds this agent to it.

with_conversation(conversation_id) -> Agent

Returns a copy of this agent bound to the given conversation ID.

Example

from djangosdk.agents.base import Agent
from djangosdk.providers.schemas import ReasoningConfig
from djangosdk.tools.decorator import tool

@tool
def search(query: str) -> str:
    """Search the knowledge base."""
    return f"Results for: {query}"

class MyAgent(Agent):
    provider = "anthropic"
    model = "claude-3-5-haiku-20241022"
    system_prompt = "You are a research assistant."
    temperature = 0.3
    max_tokens = 4096
    tools = [search]

agent = MyAgent()
response = agent.handle("Find information about Django migrations.")
print(response.text)