fix: add warning when decorators used without tracer initialization

[wingding12]

Summary

Fixes #1104

When @agent, @tool, and other decorators are used without calling agentops.init() first, they now provide clear feedback instead of silently failing.

Root Cause

The decorator factory checked if not tracer.initialized: and silently returned the function result without any indication that tracing wasn't happening. This caused confusion because:

• LLM instrumentation (via providers) worked because it has its own initialization
• Decorators silently passed through without creating spans
• Users had no way to know their decorators weren't working

Solution

Added _ensure_tracer_initialized() helper that:

1. Shows a clear warning message explaining the issue
2. Suggests solutions: call agentops.init() or set AGENTOPS_AUTO_INIT=true
3. Only shows warning once to avoid log spam
4. Still executes the decorated function normally (graceful degradation)

New Environment Variable

AGENTOPS_AUTO_INIT=true - When set, decorators will auto-initialize AgentOps with sensible defaults:

• auto_start_session=False
• instrument_llm_calls=True

Files Changed

• agentops/sdk/decorators/factory.py - Add initialization check with warning
• tests/unit/sdk/test_decorators.py - Add tests for warning behavior

Test Plan

• Test that warning is shown when tracer not initialized
• Test that warning is only shown once (no spam)
• Test that decorated functions still execute without tracer