Update open-telemetry/semantic-conventions-genai to latest

[github-actions]

Integrate unreleased gen-ai semantic conventions changes from upstream HEAD (e153ed9) in open-telemetry/semantic-conventions-genai.

This PR maintains continuous alignment with the OpenTelemetry GenAI semantic conventions (Development stability) by implementing actionable changes and documenting deferred items until they have emission sites in Microsoft.Extensions.AI.

Supersedes #11 (closed without merge).

What this PR implements

Change	Impact	Implementation
gen_ai.request.reasoning.level attribute added	🟡 Enhancement	Added constant to OpenTelemetryConsts.cs and emission in OpenTelemetryChatClient.cs. Maps ChatOptions.Reasoning.Effort enum to lowercase string values ("none", "low", "medium", "high", "extra_high").
document modality value added to message schemas	🟡 Enhancement	Updated DeriveModalityFromMediaType() to return "document" for application/* media types (e.g., PDF, DOCX).
moonshot_ai provider value added	🟢 Spec update	No code change required — gen_ai.provider.name already accepts any string value. This is a well-known value addition to the spec.

Deferred Changes (Documented Only)

Change	Type	Rationale
gen_ai.conversation.compacted attribute	🟢 Deferred	New attribute. No emission site — conversation compaction not yet instrumented.
gen_ai.prompt.version and gen_ai.prompt.variable attributes	🟢 Deferred	New attributes. No emission site — prompt management/templating not yet instrumented.
gen_ai.workflow.duration metric	🟢 Deferred	New metric. Workflow orchestration not yet instrumented.
gen_ai.invoke_agent.duration and gen_ai.execute_tool.duration metrics	🟢 Deferred	New metrics. Agent/tool duration tracking not yet instrumented beyond basic spans.
Memory operation spans and attributes (create/delete store, CRUD records)	🟢 Deferred	New capability. Memory store operations not yet instrumented.
plan operation name for agent planning spans	🟢 Deferred	New operation. Planning/task-decomposition not yet instrumented as a distinct operation.
Agent/tool sampling-relevance changes	🟢 Deferred	Attribute scope/requiredness changes for uninstrumented capabilities.
MCP context propagation clarification	🟢 Clarification	Documentation update for MCP params._meta keys per SEP-414. No code impact.
Token count guidance (billed vs. consumed)	🟢 Clarification	Documentation clarification. M.E.AI already emits what providers report.
GenAI span duration guidance (includes retries)	🟢 Clarification	Documentation clarification. M.E.AI spans already cover full operation duration.
gen_ai.conversation.id guidance (no fallback UUIDs)	🟢 Clarification	Documentation clarification. M.E.AI already emits only available conversation IDs.
gen_ai.provider.name generalization	🟢 Clarification	Documentation update removing inference-centric language. No code impact.
gen_ai.provider.name relaxed to Conditionally Required on duration metric	🟢 Clarification	Requiredness change for metric already emitted. No functional impact.
gen_ai.system_instructions part types limited to text	🟢 Clarification	Schema constraint. M.E.AI already only emits text system instructions.
gen_ai.request.top_k type change and scoping	🟢 Clarification	Type changed from double to int and scoped to decoding only (retrieval uses new gen_ai.retrieval.top_k). M.E.AI already uses int and only emits for chat (decoding context).

Testing

• Added test coverage for gen_ai.request.reasoning.level emission with ReasoningEffort.High
• Added test coverage for document modality with application/pdf content
• All existing tests pass

Upstream Tracking

Upstream-Repo: open-telemetry/semantic-conventions-genai
Upstream-Scan-Ref: e153ed94728993acd0ee6a958559032ca8b20afe
Upstream-Scan-Date: 2026-06-27T00:45:34Z
Upstream-Release: none
Core-Semconv-Dependency: v1.42.0
DotnetExtensions-Implemented-Version: v1.41

Generated by MEAI: Update Otel Gen-AI · ● 25M · ◷