Access these through the tracing module:
Message
A single message in a conversation.
Two construction styles are supported:
-
Flat (back-compat, ergonomic for plain text):
Message(role="assistant", content="Hi there")
-
Explicit parts (richer; supports tool calls, mixed reasoning+text,
inline media):
Message(role="assistant", parts=[TextPart(content="Let me check"), ToolCallPart(id="c1", name="get_weather", arguments='{...}')])
When parts is non-empty it is the canonical representation. When
empty, the serializer synthesizes a single TextPart (or
ToolCallResponsePart for role="tool") from the flat fields.
Message is a Pydantic model. Set any field below as a keyword argument when you construct it.
Fields
Message.user
Build a user message from plain text.
Message.system
Build a system message from plain text.
Message.assistant
Build an assistant message with optional text and tool calls.
Use plain text for simple replies; pass tool_calls when the
assistant requests one or more tools. When both are present the
text is emitted as a leading TextPart followed by each
ToolCallPart so the chat view renders them inline.
Build a tool-result message for a previously-requested tool call.
output may be a string, dict, list, scalar, or None; the
underlying ToolCallResponsePart JSON-encodes non-strings.
MessagePart
TextPart
Plain-text content part.
TextPart is a Pydantic model. Set any field below as a keyword argument when you construct it.
Fields
ReasoningPart
Reasoning / chain-of-thought content part.
ReasoningPart is a Pydantic model. Set any field below as a keyword argument when you construct it.
Fields
An assistant’s request to invoke a tool.
arguments is a string on the wire (per GenAI semconv). The
JSONString annotation lets callers pass a dict / list / scalar
at construction or assignment and the SDK JSON-encodes it; an
already-encoded string passes through unchanged.
ToolCallPart is a Pydantic model. Set any field below as a keyword argument when you construct it.
Fields
The result of executing a previously-requested tool call.
response is a string on the wire. Same JSONString ergonomics
as ToolCallPart.arguments.
ToolCallResponsePart is a Pydantic model. Set any field below as a keyword argument when you construct it.
Fields
BlobPart
Inline binary content (for example base64-encoded image / audio).
BlobPart is a Pydantic model. Set any field below as a keyword argument when you construct it.
Fields
UriPart
Reference to media content by URI.
UriPart is a Pydantic model. Set any field below as a keyword argument when you construct it.
Fields
FilePart
Reference to media content by provider-side file id.
FilePart is a Pydantic model. Set any field below as a keyword argument when you construct it.
Fields
Usage
Token usage for an LLM call.
Usage is a Pydantic model. Set any field below as a keyword argument when you construct it.
Fields
Reasoning
Reasoning/chain-of-thought content from an LLM call.
Reasoning is a Pydantic model. Set any field below as a keyword argument when you construct it.
Fields
Media emitted as an OTLP GenAI blob, URI, or provider file part.
MediaAttachment is a Pydantic model. Set any field below as a keyword argument when you construct it.
Fields
Last modified on September 30, 2026