Observability

Content-capture config primitives for GLLM Core.

ComponentIOCaptureConfig

Bases: ContentCaptureConfig

Configuration for capturing Component.run() input/output on trace spans.

Attributes:

Name	Type	Description
capture_input	bool	Whether to capture the component's input kwargs. Defaults to False.
capture_output	bool	Whether to capture the component's output. Defaults to False.

ContentCaptureConfig

Bases: BaseModel

Base for process-wide opt-in content-capture configs.

Attributes:

Name	Type	Description
max_attribute_bytes	int	Maximum number of bytes to capture for each attribute. Defaults to DEFAULT_MAX_ATTRIBUTE_BYTES.

SpanAttribute

Span attributes for GLLM Core component instrumentation.

build_content_capture_accessors(default, *, name)

Build an isolated (configure, get) pair for one content-capture config domain.

Parameters:

Name	Type	Description	Default
default	T	The config a fresh domain starts with, and the exact instance restored whenever the returned configure function is called with None.	required
name	str	Domain identifier stamped onto the returned functions' __name__ and __doc__ (e.g. "lm_trace_content"). Never used for lookup or cross-domain behavior.	required

Returns:

Type	Description
tuple[Callable[[T | dict[str, Any] | None], None], Callable[[], T]]	tuple[Callable[[T | dict[str, Any] | None], None], Callable[[], T]]: The configure_fn and get_fn for this domain, each with its own isolated lock and global state, independent of any other domain's accessor pair.

Raises:

Type	Description
ValueError	If name is empty or whitespace-only.

get_tracer()

Gets the GLLM Core tracer singleton.

Returns:

Type	Description
Tracer	trace.Tracer: The versioned tracer for GLLM Core instrumentation.