Skip to main content
This page lists all public functions exported by the AgentBasis SDK.

​
agentbasis

Core module functions imported directly from agentbasis. 
import agentbasis

​
Initialization

FunctionDescription
init(api_key: str = None, agent_id: str = None) -> AgentBasisInitialize the SDK. Reads from AGENTBASIS_API_KEY and AGENTBASIS_AGENT_ID env vars if not provided.
flush(timeout_millis: int = 30000) -> boolForce flush all pending telemetry data. Returns True if successful.
shutdown() -> NoneShut down the SDK and flush pending data. Auto-called on exit.

​
Context Management

FunctionDescription
set_user(user_id: str | None) -> NoneSet the current user ID globally.
set_session(session_id: str | None) -> NoneSet the current session ID globally.
set_conversation(conversation_id: str | None) -> NoneSet the current conversation ID globally.
set_metadata(metadata: dict | None) -> NoneSet custom metadata globally (e.g., {"plan": "pro"}).
get_user() -> str | NoneGet the current user ID.
get_session() -> str | NoneGet the current session ID.
get_conversation() -> str | NoneGet the current conversation ID.
get_metadata() -> dict | NoneGet the current metadata.

​
Context Managers & Decorators

FunctionDescription
context(user_id, session_id, conversation_id, metadata)Context manager for scoped context.
with_context(user_id, session_id, conversation_id, metadata)Decorator for setting context on a function.
traceDecorator to track function execution as a span. Works with sync and async.

​
agentbasis.llms.openai

from agentbasis.llms.openai import instrument
FunctionDescription
instrument() -> NoneAuto-instrument OpenAI SDK. Traces both OpenAI and AsyncOpenAI clients.

​
agentbasis.llms.anthropic

from agentbasis.llms.anthropic import instrument
FunctionDescription
instrument() -> NoneAuto-instrument Anthropic SDK. Traces both Anthropic and AsyncAnthropic clients.

​
agentbasis.llms.gemini

from agentbasis.llms.gemini import instrument
FunctionDescription
instrument() -> NoneAuto-instrument Google Gemini SDK. Traces GenerativeModel.generate_content() sync and async.

​
agentbasis.frameworks.langchain

from agentbasis.frameworks.langchain import get_callback_handler, instrument, get_callback_config
FunctionDescription
get_callback_handler() -> AgentBasisCallbackHandlerGet a new callback handler instance.
instrument() -> AgentBasisCallbackHandlerGet the global singleton callback handler.
get_callback_config() -> dictGet a RunnableConfig dict with callbacks pre-configured.

​
AgentBasisCallbackHandler

Callback handler that traces:
  • LLM calls (on_llm_start/end/error)
  • Chain execution (on_chain_start/end/error)
  • Tool invocations (on_tool_start/end/error)
  • Retriever operations (on_retriever_start/end/error)

​
agentbasis.frameworks.pydanticai

from agentbasis.frameworks.pydanticai import instrument, get_instrumentation_settings, get_metadata_callback, create_traced_agent
FunctionDescription
instrument(include_content: bool = True, include_binary_content: bool = False) -> NoneEnable global instrumentation for all Pydantic AI agents.
get_instrumentation_settings(include_content, include_binary_content) -> InstrumentationSettingsGet settings for per-agent instrumentation.
get_metadata_callback() -> CallableGet a callback that injects AgentBasis context into traces.
create_traced_agent(model: str, include_content: bool = True, **kwargs) -> AgentCreate a Pydantic AI agent pre-configured with tracing.

​
Parameters

ParameterDescription
include_contentWhether to log prompts and completions. Set False for privacy. Default: True
include_binary_contentWhether to include binary content like images. Default: False