Smolagents instrumentation (#1070)

* Add SmoLAgents instrumentation

* Remove deprecated instrumentor definitions from `__init__.py` to streamline the instrumentation module. This change enhances code clarity and prepares for future updates.

* Add SmoLAgents instrumentation

* Refactor multi_smolagents_system notebook to streamline agent management. Removed ManagedAgent and adjusted CodeAgent to directly manage web_agent, enhancing clarity and simplifying the agent structure.

* Add a blank line for improved readability in `__init__.py`

* Refactor SmoLAgents instrumentation for improved clarity and consistency. Updated class names and removed unused dynamic span naming functions. Simplified attribute extraction for agents and models, enhancing observability in telemetry data.

* feat: Add multi-agent system notebook example

* fix: Update execution counts and correct session ending method in multi-agent notebook

* fix: Correct markdown formatting and remove deprecated multi-agent example notebook

* feat: Add Smolagents integration documentation and examples

* fix: Add missing newline before InstrumentorLoader class definition

* feat: Add smolagents integration card to examples documentation

* feat: Update examples and introduction documentation to include Smolagents integration

---------

Co-authored-by: Braelyn Boynton <[email protected]>

Author: Dwij1704

agentops/instrumentation/common/attributes.py

@@ -252,3 +252,27 @@ def get_base_span_attributes(span: Any) -> AttributeMap:
         attributes[CoreAttributes.PARENT_ID] = parent_id
 
     return attributes
+
+
+def extract_token_usage(response: Any) -> Dict[str, int]:
+    """Extract token usage information from a response.
+
+    Args:
+        response: The response object to extract token usage from
+
+    Returns:
+        Dictionary containing token usage information
+    """
+    usage = {}
+
+    # Try to extract token counts from response
+    if hasattr(response, "usage"):
+        usage_data = response.usage
+        if hasattr(usage_data, "prompt_tokens"):
+            usage["prompt_tokens"] = usage_data.prompt_tokens
+        if hasattr(usage_data, "completion_tokens"):
+            usage["completion_tokens"] = usage_data.completion_tokens
+        if hasattr(usage_data, "total_tokens"):
+            usage["total_tokens"] = usage_data.total_tokens
+
+    return usage

agentops/instrumentation/smolagents/README.md

@@ -0,0 +1,88 @@
+# SmoLAgents Instrumentation
+
+This module provides OpenTelemetry instrumentation for the SmoLAgents framework. It captures telemetry data from model operations, agent executions, and tool usage.
+
+## Features
+
+- Model operation tracking
+  - Text generation
+  - Token usage
+  - Streaming responses
+  - Latency metrics
+
+- Agent execution monitoring
+  - Step-by-step execution
+  - Planning phases
+  - Tool usage
+  - Execution time
+
+- Tool usage analytics
+  - Tool call patterns
+  - Success/failure rates
+  - Execution time
+  - Error tracking
+
+## Usage
+
+```python
+from agentops import init
+from agentops.instrumentation.smolagents import SmoLAgentsInstrumentor
+
+# Initialize AgentOps with your API key
+init(api_key="your-api-key")
+
+# The instrumentation will be automatically activated
+# All SmoLAgents operations will now be tracked
+```
+
+## Metrics Collected
+
+1. Token Usage
+   - Input tokens
+   - Output tokens
+   - Total tokens per operation
+
+2. Timing Metrics
+   - Operation duration
+   - Time to first token (streaming)
+   - Tool execution time
+   - Planning phase duration
+
+3. Agent Metrics
+   - Step counts
+   - Planning steps
+   - Tools used
+   - Success/failure rates
+
+4. Error Tracking
+   - Generation errors
+   - Tool execution errors
+   - Parsing errors
+
+## Architecture
+
+The instrumentation is built on OpenTelemetry and follows the same pattern as other AgentOps instrumentors:
+
+1. Attribute Extractors
+   - Model attributes
+   - Agent attributes
+   - Tool call attributes
+
+2. Wrappers
+   - Method wrappers for sync operations
+   - Stream wrappers for async operations
+   - Context propagation handling
+
+3. Metrics
+   - Histograms for distributions
+   - Counters for events
+   - Custom attributes for filtering
+
+## Contributing
+
+When adding new features or modifying existing ones:
+
+1. Follow the established pattern for attribute extraction
+2. Maintain context propagation
+3. Add appropriate error handling
+4. Update tests and documentation 

agentops/instrumentation/smolagents/__init__.py

@@ -0,0 +1,8 @@
+"""SmoLAgents instrumentation for AgentOps."""
+
+LIBRARY_NAME = "smolagents"
+LIBRARY_VERSION = "1.16.0"
+
+from agentops.instrumentation.smolagents.instrumentor import SmolAgentsInstrumentor  # noqa: E402
+
+__all__ = ["SmolAgentsInstrumentor"]