fix: instrument httpx client at module level for MCP context

The lifespan-based instrumentation only runs when FastAPI app starts.
In MCP context, the app never starts but the httpx client is still used.

Solution: Instrument the client immediately after creation at module level.
This works in both contexts:
- MCP: client is instrumented when module is imported
- API: client is instrumented before lifespan runs (lifespan still safe)

This enables distributed tracing from MCP -> Cloud -> API.

Signed-off-by: phernandez <paul@basicmachines.co>

Author: phernandez

src/basic_memory/mcp/async_client.py

@@ -38,3 +38,15 @@ def create_client() -> AsyncClient:
 
 # Create shared async client
 client = create_client()
+
+# Instrument client for distributed tracing when in cloud mode
+# This must happen AFTER client creation and works in both MCP and API contexts
+config = ConfigManager().config
+if config.cloud_mode_enabled:
+    try:
+        import logfire  # pyright: ignore[reportMissingImports]
+
+        logger.info("Cloud mode: instrumenting httpx client for distributed tracing")
+        logfire.instrument_httpx(client=client)
+    except ImportError:
+        logger.warning("logfire not available - skipping httpx instrumentation")