OTel enhancements for deployed server

[pamelafox]

OTel improvements to deployed MCP:

• Adds an explicit option for AppInsights vs Logfire vs None for OTel export
• Adds an env var for Logfire token that can proprogate to prod
• Migrates the middleware to use new MCP semantic conventions
• New README sections

[Copilot]

Pull request overview

This PR enhances OpenTelemetry (OTel) configuration for the deployed MCP server by introducing a unified approach to telemetry platform selection and implementing MCP semantic conventions for better observability.

Key Changes:

• Introduces OPENTELEMETRY_PLATFORM environment variable to explicitly choose between AppInsights, Logfire, or no telemetry export
• Migrates OpenTelemetry middleware to follow MCP semantic conventions (PR #2083) with standardized span naming and GenAI attributes
• Adds infrastructure support for Logfire token as a secure environment variable

Reviewed changes

Copilot reviewed 13 out of 13 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
servers/opentelemetry_middleware.py	Implements MCP semantic conventions with new span naming pattern "{method} {target}", adds GenAI attributes (gen_ai.tool.name, gen_ai.prompt.name), and includes safe JSON serialization helper
servers/deployed_mcp.py	Updates telemetry configuration to use OPENTELEMETRY_PLATFORM env var, changes get_expenses_data from @mcp.resource to @mcp.tool, removes enable_cross_partition_query parameter, and simplifies import
servers/auth_mcp.py	Updates to use OPENTELEMETRY_PLATFORM for consistent telemetry configuration and renames csv_content to expense_summary
servers/basic_mcp_http.py	Changes logging level from INFO to WARNING and updates transport from "http" to "streamable-http"
infra/write_env.sh	Adds LOGFIRE_TOKEN (conditional) and OPENTELEMETRY_PLATFORM to environment file generation
infra/write_env.ps1	PowerShell equivalent of write_env.sh changes for Windows environments
infra/server.bicep	Adds openTelemetryPlatform parameter with validation, logfireToken secure parameter, and conditional environment variables/secrets
infra/main.parameters.json	Replaces useMonitoring with openTelemetryPlatform parameter and adds logfireToken
infra/main.bicep	Replaces useMonitoring boolean with openTelemetryPlatform enum, derives useAppInsights, and propagates new parameters throughout infrastructure
agents/langchainv1_github.py	Updates example query to be more specific (requests 5 repositories with bulleted list)
agents/agentframework_learn.py	Removes unnecessary Authorization header from MCPStreamableHTTPTool initialization
agents/agentframework_http.py	Restructures environment loading to only call load_dotenv when not in production
README.md	Adds new sections for using deployed MCP server with GitHub Copilot, running server locally, viewing traces in Application Insights, and viewing traces in Logfire

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The decorator was changed from @mcp.resource to @mcp.tool for this function in deployed_mcp.py, but remains as @mcp.resource in basic_mcp_http.py. This creates an inconsistency between the two server implementations. Consider whether get_expenses_data should be a resource or a tool, and ensure both files use the same decorator type for consistency.

Suggested change

	@mcp.tool
	@mcp.resource

[pamelafox]

Thats because we're explicitly showing resources in the first session, when we use basic. But my experience with the deployed servers is that resources dont get called often, so I changed them to tools to be more useful.

[Copilot]

The removal of 'enable_cross_partition_query=True' may cause the query to fail if the Cosmos DB container is partitioned. Without this parameter, Cosmos DB queries are limited to a single partition. If the container uses a partition key (which is common in production), this query should either specify the partition key or include enable_cross_partition_query=True to query across all partitions.

Suggested change

	async for item in cosmos_container.query_items(query=query):
	async for item in cosmos_container.query_items(query=query, enable_cross_partition_query=True):

[pamelafox]

Hm I think this comment is based off outdated knowledge. This is what Opus says:
"Looking at the Azure Cosmos DB Python SDK documentation, enable_cross_partition_query is no longer needed in the modern SDK. The parameter was deprecated and is now the default behavior - cross-partition queries are automatically enabled. So you removed it because it's unnecessary with the current version of the azure-cosmos SDK.

The SDK automatically handles cross-partition queries without requiring this flag, making the code cleaner."

I believe I verified this change.