core Package

Packages

exporters
middleware
models
spans_scopes

Modules

agent_details
apply_guardrail_scope	OpenTelemetry tracing scope for guardrail (security guardian) evaluations.
channel
constants
enricher_utils	Shared utilities for A365 span enrichers. Provides content extraction helpers used by framework-specific enrichers (Agent Framework, Semantic Kernel, LangChain) to convert structured OTel messages to plain content arrays before A365 export.
execute_tool_scope
guardrail_decision_type	String constants for guardrail decision types.
guardrail_details	Data class for guardrail evaluation details.
guardrail_finding	Data class for a single guardrail security finding.
guardrail_risk_severity	String constants for guardrail risk severity levels.
guardrail_target_type	String constants for guardrail target types.
inference_call_details
inference_operation_type
inference_scope
invoke_agent_details
invoke_agent_scope
message_utils	Conversion and serialization helpers for OTEL gen-ai message format. Provides normalization from plain `list[str]` (backward compat) to the structured array format, and a non-throwing `serialize_messages` function.
opentelemetry_scope
request
span_details
tool_call_details
tool_type
utils

Classes

AgentDetails	Details about an AI agent in the system.
ApplyGuardrailScope	Provides OpenTelemetry tracing scope for security guardrail evaluations. Guardian spans SHOULD be children of the operation span they are protecting (e.g., inference or execute_tool spans). Multiple guardian spans MAY exist under a single operation span if multiple guardians are chained. Example usage: from microsoft.opentelemetry.a365.core import ( ApplyGuardrailScope, GuardrailDetails, AgentDetails, GuardrailDecisionType, GuardrailTargetType, GuardrailRiskSeverity, GuardrailFinding, ) details = GuardrailDetails( target_type=GuardrailTargetType.LLM_INPUT, decision_type=GuardrailDecisionType.ALLOW, guardian_name="Azure Content Safety", ) with ApplyGuardrailScope.start(details, agent_details) as scope: # ... run guardrail evaluation ... scope.record_finding(GuardrailFinding( risk_category="hate_speech", risk_severity=GuardrailRiskSeverity.HIGH, risk_score=0.95, )) scope.record_decision(GuardrailDecisionType.DENY, "Blocked by policy") Initialize the guardrail scope.
BaggageBuilder	Per request baggage builder. This class provides a fluent API for setting baggage values that will be propagated in the OpenTelemetry context. Initialize the baggage builder.
BlobPart	Inline binary data (base64-encoded).
CallerDetails	Composite caller details for agent-to-agent (A2A) scenarios. Groups the human caller identity and the calling agent identity together.
Channel	Channel information for agent execution context.
ChatMessage	An input message sent to a model (OTEL gen-ai semantic conventions).
ExecuteToolScope	Provides OpenTelemetry tracing scope for AI tool execution operations. Initialize the tool execution scope.
FilePart	Reference to a pre-uploaded file.
GenericPart	Extensible part for custom / future types.
GuardrailDecisionType	Well-known guardrail decision type values.
GuardrailDetails	Immutable input contract describing a guardrail evaluation.
GuardrailFinding	A single security finding from a guardrail evaluation.
GuardrailRiskSeverity	Well-known risk severity level values.
GuardrailTargetType	Well-known guardrail target type values. Users may also supply custom string values not listed here.
InferenceCallDetails	Details of an inference call for generative AI operations.
InferenceScope	Provides OpenTelemetry tracing scope for generative AI inference operations. Initialize the inference scope.
InputMessages	Container for input messages. Serializes as a plain JSON array per OTel spec.
InvokeAgentScope	Provides OpenTelemetry tracing scope for AI agent invocation operations. Initialize the agent invocation scope.
InvokeAgentScopeDetails	Scope-level configuration for agent invocation tracing.
OpenTelemetryScope	Base class for OpenTelemetry tracing scopes in the SDK. Initialize the OpenTelemetry scope.
OutputMessage	An output message produced by a model (OTEL gen-ai semantic conventions). `finish_reason` defaults to `"stop"` per OTel spec when not provided.
OutputMessages	Container for output messages. Serializes as a plain JSON array per OTel spec.
OutputScope	Provides OpenTelemetry tracing scope for output messages. Output messages are set once (via the constructor or `record_output_messages`) rather than accumulated. For streaming scenarios, the agent developer should collect all output (e.g. via a list or string builder) and pass the final result to `OutputScope`. Initialize the output scope.
ReasoningPart	Model reasoning / chain-of-thought content.
Request	Request details for agent execution.
Response	Response details from agent execution. Accepts plain strings (backward compat), structured `OutputMessages`, or a `dict` for tool call results (per OTEL spec).
ServerToolCallPart	Server-side tool invocation.
ServerToolCallResponsePart	Server-side tool response.
ServiceEndpoint	Represents a service endpoint with hostname and optional port.
SpanDetails	Groups span configuration for scope construction.
TextPart	Plain text content.
ToolCallDetails	Details of a tool call made by an agent in the system.
ToolCallRequestPart	A tool call requested by the model.
ToolCallResponsePart	Result of a tool call.
UriPart	External URI reference.
UserDetails	Details about the human user that invoked an agent.

Enums

FinishReason	Reason a model stopped generating per OTEL gen-ai semantic conventions.
InferenceOperationType	Supported inference operation types for generative AI.
MessageRole	Role of a message participant per OTEL gen-ai semantic conventions.
Modality	Media modality for blob, file, and URI parts.
ToolType	Enumeration for different tool types for execute tool contexts.