netra-sdk 0.1.23__tar.gz → 0.1.25__tar.gz

{netra_sdk-0.1.23 → netra_sdk-0.1.25}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: netra-sdk
-Version: 0.1.23
+Version: 0.1.25
 Summary: A Python SDK for AI application observability that provides OpenTelemetry-based monitoring, tracing, and PII protection for LLM and vector database applications. Enables easy instrumentation, session tracking, and privacy-focused data collection for AI systems in production environments.
 License: Apache-2.0
 Keywords: netra,tracing,observability,sdk,ai,llm,vector,database
@@ -72,6 +72,7 @@ Requires-Dist: presidio-anonymizer (==2.2.358) ; extra == "presidio"
 Requires-Dist: stanza (>=1.10.1,<2.0.0) ; extra == "presidio"
 Requires-Dist: traceloop-sdk (>=0.40.7,<0.43.0)
 Requires-Dist: transformers (==4.51.3) ; extra == "presidio"
+Requires-Dist: twine (>=6.1.0,<7.0.0)
 Project-URL: Bug Tracker, https://github.com/KeyValueSoftwareSystems/netra-sdk-py/issues
 Project-URL: Documentation, https://github.com/KeyValueSoftwareSystems/netra-sdk-py/blob/main/README.md
 Project-URL: Homepage, https://github.com/KeyValueSoftwareSystems/netra-sdk-py
@@ -286,6 +287,7 @@ class CustomerSupportAgent:
 - **LlamaIndex** - Data framework for LLM applications
 - **Haystack** - End-to-end NLP framework
 - **CrewAI** - Multi-agent AI systems
+- **Pydantic AI** - AI model communication standard
 - **MCP (Model Context Protocol)** - AI model communication standard
 
 ## 🛡️ Privacy Protection & Security

{netra_sdk-0.1.23 → netra_sdk-0.1.25}/README.md

@@ -206,6 +206,7 @@ class CustomerSupportAgent:
 - **LlamaIndex** - Data framework for LLM applications
 - **Haystack** - End-to-end NLP framework
 - **CrewAI** - Multi-agent AI systems
+- **Pydantic AI** - AI model communication standard
 - **MCP (Model Context Protocol)** - AI model communication standard
 
 ## 🛡️ Privacy Protection & Security

{netra_sdk-0.1.23 → netra_sdk-0.1.25}/netra/instrumentation/__init__.py

@@ -34,16 +34,11 @@ def init_instrumentations(
             else:
                 traceloop_block_instruments.add(getattr(Instruments, instrument.name))
 
-    # If no instruments are provided for instrumentation
-    if instruments is None:
-        traceloop_block_instruments = set(Instruments)
-        netra_custom_block_instruments = set(CustomInstruments)
-
-    # If only custom instruments from netra are provided for instrumentation
+    # If no instruments in traceloop are provided for instrumentation
     if instruments is not None and not traceloop_instruments and not traceloop_block_instruments:
         traceloop_block_instruments = set(Instruments)
 
-    # If only traceloop instruments are provided for instrumentation
+    # If no custom instruments in netra are provided for instrumentation
     if instruments is not None and not netra_custom_instruments and not netra_custom_block_instruments:
         netra_custom_block_instruments = set(CustomInstruments)
 
@@ -102,6 +97,10 @@ def init_instrumentations(
     if CustomInstruments.OPENAI in netra_custom_instruments:
         init_openai_instrumentation()
 
+    # Initialize Pydantic AI instrumentation.
+    if CustomInstruments.PYDANTIC_AI in netra_custom_instruments:
+        init_pydantic_ai_instrumentation()
+
     # Initialize aio_pika instrumentation.
     if CustomInstruments.AIO_PIKA in netra_custom_instruments:
         init_aio_pika_instrumentation()
@@ -1124,3 +1123,19 @@ def init_urllib3_instrumentation() -> bool:
         logging.error(f"Error initializing urllib3 instrumentor: {e}")
         Telemetry().log_exception(e)
         return False
+
+
+def init_pydantic_ai_instrumentation() -> bool:
+    """Initialize pydantic-ai instrumentation."""
+    try:
+        if is_package_installed("pydantic-ai"):
+            from netra.instrumentation.pydantic_ai import NetraPydanticAIInstrumentor
+
+            instrumentor = NetraPydanticAIInstrumentor()
+            if not instrumentor.is_instrumented_by_opentelemetry:
+                instrumentor.instrument()
+        return True
+    except Exception as e:
+        logging.error(f"Error initializing pydantic-ai instrumentation: {e}")
+        Telemetry().log_exception(e)
+        return False

{netra_sdk-0.1.23 → netra_sdk-0.1.25}/netra/instrumentation/instruments.py

@@ -10,6 +10,7 @@ class CustomInstruments(Enum):
     HTTPX = "httpx"
     MISTRALAI = "mistral_ai"
     OPENAI = "openai"
+    PYDANTIC_AI = "pydantic_ai"
     QDRANTDB = "qdrant_db"
     WEAVIATEDB = "weaviate_db"
     GOOGLE_GENERATIVEAI = "google_genai"

netra_sdk-0.1.25/netra/instrumentation/pydantic_ai/__init__.py

@@ -0,0 +1,200 @@
+import logging
+from typing import Collection
+
+from opentelemetry.instrumentation.instrumentor import BaseInstrumentor
+from opentelemetry.instrumentation.utils import unwrap
+from opentelemetry.trace import get_tracer
+from wrapt import wrap_function_wrapper
+
+from netra.instrumentation.pydantic_ai.version import __version__
+from netra.instrumentation.pydantic_ai.wrappers import (
+    agent_iter_wrapper,
+    agent_run_stream_wrapper,
+    agent_run_sync_wrapper,
+    agent_run_wrapper,
+    tool_function_wrapper,
+)
+
+logger = logging.getLogger(__name__)
+
+_instruments = ("pydantic-ai >= 0.0.1",)
+
+
+class NetraPydanticAIInstrumentor(BaseInstrumentor):  # type: ignore[misc]
+    """
+    Custom Pydantic AI instrumentor for Netra SDK with enhanced support for:
+    - Agent.run, Agent.run_sync, Agent.iter, Agent.run_stream methods
+    - Tool function execution tracing
+    - OpenTelemetry semantic conventions for Generative AI
+    - Integration with Netra tracing and monitoring
+    """
+
+    def instrumentation_dependencies(self) -> Collection[str]:
+        return _instruments
+
+    def _instrument(self, **kwargs):  # type: ignore[no-untyped-def]
+        """Instrument Pydantic AI Agent methods and tool functions"""
+        tracer_provider = kwargs.get("tracer_provider")
+        tracer = get_tracer(__name__, __version__, tracer_provider)
+
+        # Instrument Agent.run method
+        try:
+            wrap_function_wrapper(
+                "pydantic_ai.agent",
+                "Agent.run",
+                agent_run_wrapper(tracer),
+            )
+        except (AttributeError, ModuleNotFoundError):
+            logger.debug("Agent.run method not available in this pydantic-ai version")
+
+        # Instrument Agent.run_sync method
+        try:
+            wrap_function_wrapper(
+                "pydantic_ai.agent",
+                "Agent.run_sync",
+                agent_run_sync_wrapper(tracer),
+            )
+        except (AttributeError, ModuleNotFoundError):
+            logger.debug("Agent.run_sync method not available in this pydantic-ai version")
+
+        # Instrument Agent.iter method
+        try:
+            wrap_function_wrapper(
+                "pydantic_ai.agent",
+                "Agent.iter",
+                agent_iter_wrapper(tracer),
+            )
+        except (AttributeError, ModuleNotFoundError):
+            logger.debug("Agent.iter method not available in this pydantic-ai version")
+
+        # Instrument Agent.run_stream method (if available)
+        try:
+            wrap_function_wrapper(
+                "pydantic_ai.agent",
+                "Agent.run_stream",
+                agent_run_stream_wrapper(tracer),
+            )
+        except (AttributeError, ModuleNotFoundError):
+            logger.debug("Agent.run_stream method not available in this pydantic-ai version")
+
+        # Instrument AgentRun methods
+        try:
+            wrap_function_wrapper(
+                "pydantic_ai.agent",
+                "AgentRun.run",
+                agent_run_wrapper(tracer),
+            )
+        except (AttributeError, ModuleNotFoundError):
+            logger.debug("AgentRun.run method not available in this pydantic-ai version")
+
+        try:
+            wrap_function_wrapper(
+                "pydantic_ai.agent",
+                "AgentRun.run_sync",
+                agent_run_sync_wrapper(tracer),
+            )
+        except (AttributeError, ModuleNotFoundError):
+            logger.debug("AgentRun.run_sync method not available in this pydantic-ai version")
+
+        # Instrument tool execution (if tools module exists)
+        try:
+            wrap_function_wrapper(
+                "pydantic_ai.tools",
+                "Tool.__call__",
+                tool_function_wrapper(tracer),
+            )
+        except (AttributeError, ModuleNotFoundError):
+            logger.debug("Tool.__call__ method not available in this pydantic-ai version")
+
+        # Instrument function tools (if function_tools module exists)
+        try:
+            wrap_function_wrapper(
+                "pydantic_ai.tools",
+                "FunctionTool.__call__",
+                tool_function_wrapper(tracer),
+            )
+        except (AttributeError, ModuleNotFoundError):
+            logger.debug("FunctionTool.__call__ method not available in this pydantic-ai version")
+
+        # Instrument model calls (if models module exists)
+        try:
+            wrap_function_wrapper(
+                "pydantic_ai.models.base",
+                "Model.request",
+                agent_run_wrapper(tracer),
+            )
+        except (AttributeError, ModuleNotFoundError):
+            logger.debug("Model.request method not available in this pydantic-ai version")
+
+        try:
+            wrap_function_wrapper(
+                "pydantic_ai.models.base",
+                "Model.request_stream",
+                agent_run_stream_wrapper(tracer),
+            )
+        except (AttributeError, ModuleNotFoundError):
+            logger.debug("Model.request_stream method not available in this pydantic-ai version")
+
+    def _uninstrument(self, **kwargs):  # type: ignore[no-untyped-def]
+        """Uninstrument Pydantic AI Agent methods and tool functions"""
+        # Uninstrument Agent methods
+        try:
+            unwrap("pydantic_ai.agent", "Agent.run")
+        except (AttributeError, ModuleNotFoundError):
+            pass
+
+        try:
+            unwrap("pydantic_ai.agent", "Agent.run_sync")
+        except (AttributeError, ModuleNotFoundError):
+            pass
+
+        try:
+            unwrap("pydantic_ai.agent", "Agent.iter")
+        except (AttributeError, ModuleNotFoundError):
+            pass
+
+        try:
+            unwrap("pydantic_ai.agent", "Agent.run_stream")
+        except (AttributeError, ModuleNotFoundError):
+            pass
+
+        # Uninstrument AgentRun methods
+        try:
+            unwrap("pydantic_ai.agent", "AgentRun.run")
+        except (AttributeError, ModuleNotFoundError):
+            pass
+
+        try:
+            unwrap("pydantic_ai.agent", "AgentRun.run_sync")
+        except (AttributeError, ModuleNotFoundError):
+            pass
+
+        # Uninstrument tool methods
+        try:
+            unwrap("pydantic_ai.tools", "Tool.__call__")
+        except (AttributeError, ModuleNotFoundError):
+            pass
+
+        try:
+            unwrap("pydantic_ai.tools", "FunctionTool.__call__")
+        except (AttributeError, ModuleNotFoundError):
+            pass
+
+        # Uninstrument model methods
+        try:
+            unwrap("pydantic_ai.models.base", "Model.request")
+        except (AttributeError, ModuleNotFoundError):
+            pass
+
+        try:
+            unwrap("pydantic_ai.models.base", "Model.request_stream")
+        except (AttributeError, ModuleNotFoundError):
+            pass
+
+
+def should_suppress_instrumentation() -> bool:
+    """Check if instrumentation should be suppressed"""
+    from opentelemetry import context as context_api
+    from opentelemetry.instrumentation.utils import _SUPPRESS_INSTRUMENTATION_KEY
+
+    return context_api.get_value(_SUPPRESS_INSTRUMENTATION_KEY) is True

netra_sdk-0.1.25/netra/instrumentation/pydantic_ai/utils.py

@@ -0,0 +1,385 @@
+from typing import Any, Dict, Optional
+
+from opentelemetry import context as context_api
+from opentelemetry.instrumentation.utils import _SUPPRESS_INSTRUMENTATION_KEY
+from opentelemetry.semconv_ai import SpanAttributes
+from opentelemetry.trace.status import Status, StatusCode
+
+# Constants for consistent truncation and limits
+MAX_CONTENT_LENGTH = 1000
+MAX_ARGS_LENGTH = 500
+MAX_ITEMS_TO_PROCESS = 5
+MAX_FUNCTIONS_TO_PROCESS = 3
+
+
+def _safe_set_attribute(span: Any, key: str, value: Any, max_length: Optional[int] = None) -> None:
+    """Safely set span attribute with optional truncation and null checks."""
+    if not span.is_recording() or value is None:
+        return
+
+    str_value = str(value)
+    if max_length and len(str_value) > max_length:
+        str_value = str_value[:max_length]
+
+    span.set_attribute(key, str_value)
+
+
+def _safe_get_attribute(obj: Any, attr_name: str, default: Any = None) -> Any:
+    """Safely get attribute from object with default fallback."""
+    return getattr(obj, attr_name, default) if hasattr(obj, attr_name) else default
+
+
+def _handle_span_error(span: Any, exception: Exception) -> None:
+    """Common error handling for spans."""
+    span.set_status(Status(StatusCode.ERROR, str(exception)))
+    _safe_set_attribute(span, "error.type", type(exception).__name__)
+    _safe_set_attribute(span, "error.message", str(exception))
+
+
+def _set_timing_attributes(span: Any, start_time: float, end_time: float) -> None:
+    """Set timing attributes on span."""
+    duration_ms = (end_time - start_time) * 1000
+    _safe_set_attribute(span, "llm.response.duration", duration_ms)
+
+
+def _set_assistant_response_content(span: Any, result: Any, finish_reason: str = "completed") -> None:
+    """Set assistant response content in OpenAI wrapper format."""
+    if not span.is_recording():
+        return
+
+    # Set the assistant response in the same format as OpenAI wrapper
+    index = 0  # Always use index 0 for pydantic_ai responses
+    _safe_set_attribute(span, f"{SpanAttributes.LLM_COMPLETIONS}.{index}.role", "assistant")
+    _safe_set_attribute(span, f"{SpanAttributes.LLM_COMPLETIONS}.{index}.finish_reason", finish_reason)
+
+    # Get the output content from the result
+    output = _safe_get_attribute(result, "output")
+    if output is not None:
+        _safe_set_attribute(span, f"{SpanAttributes.LLM_COMPLETIONS}.{index}.content", output, MAX_CONTENT_LENGTH)
+
+
+def set_pydantic_request_attributes(
+    span: Any,
+    kwargs: Dict[str, Any],
+    operation_type: str,
+    model_name: Optional[str] = None,
+    include_model: bool = False,
+) -> None:
+    """Set request attributes on span for pydantic_ai."""
+    if not span.is_recording():
+        return
+
+    # Set operation type
+    _safe_set_attribute(span, f"{SpanAttributes.LLM_REQUEST_TYPE}", operation_type)
+
+    # Set model only if explicitly requested (for CallToolsNode spans)
+    if include_model and model_name:
+        _safe_set_attribute(span, f"{SpanAttributes.LLM_REQUEST_MODEL}", model_name)
+
+    # Set temperature and max_tokens if available
+    _safe_set_attribute(span, f"{SpanAttributes.LLM_REQUEST_TEMPERATURE}", kwargs.get("temperature"))
+    _safe_set_attribute(span, f"{SpanAttributes.LLM_REQUEST_MAX_TOKENS}", kwargs.get("max_tokens"))
+
+
+def set_pydantic_response_attributes(span: Any, result: Any) -> None:
+    """Set response attributes on span for pydantic_ai."""
+    if not span.is_recording():
+        return
+
+    # Set response model if available
+    model_name = _safe_get_attribute(result, "model_name")
+    _safe_set_attribute(span, f"{SpanAttributes.LLM_RESPONSE_MODEL}", model_name)
+
+    # Set usage information
+    if hasattr(result, "usage"):
+        usage = result.usage()
+        if usage:
+            _safe_set_attribute(
+                span, f"{SpanAttributes.LLM_USAGE_PROMPT_TOKENS}", _safe_get_attribute(usage, "request_tokens")
+            )
+            _safe_set_attribute(
+                span, f"{SpanAttributes.LLM_USAGE_COMPLETION_TOKENS}", _safe_get_attribute(usage, "response_tokens")
+            )
+            _safe_set_attribute(
+                span, f"{SpanAttributes.LLM_USAGE_TOTAL_TOKENS}", _safe_get_attribute(usage, "total_tokens")
+            )
+
+            # Set any additional details from usage
+            details = _safe_get_attribute(usage, "details")
+            if details:
+                for key, value in details.items():
+                    if value:
+                        _safe_set_attribute(span, f"gen_ai.usage.details.{key}", value)
+
+    # Set output content if available
+    output = _safe_get_attribute(result, "output")
+    if output is not None:
+        _safe_set_attribute(span, f"{SpanAttributes.LLM_COMPLETIONS}.0.role", "assistant")
+        _safe_set_attribute(span, f"{SpanAttributes.LLM_COMPLETIONS}.0.content", output, MAX_CONTENT_LENGTH)
+
+
+def should_suppress_instrumentation() -> bool:
+    """Check if instrumentation should be suppressed"""
+    return context_api.get_value(_SUPPRESS_INSTRUMENTATION_KEY) is True
+
+
+def get_node_span_name(node: Any) -> str:
+    """Get appropriate span name for a node type"""
+    node_type = type(node).__name__
+    if "UserPrompt" in node_type:
+        return "pydantic_ai.node.user_prompt"
+    elif "ModelRequest" in node_type:
+        return "pydantic_ai.node.model_request"
+    elif "CallTools" in node_type:
+        return "pydantic_ai.node.call_tools"
+    elif "End" in node_type:
+        return "pydantic_ai.node.end"
+    else:
+        return f"pydantic_ai.node.{node_type.lower()}"
+
+
+def set_node_attributes(span: Any, node: Any) -> None:
+    """Set attributes on span based on node type and content"""
+    if not span.is_recording():
+        return
+
+    node_type = type(node).__name__
+    _safe_set_attribute(span, "pydantic_ai.node.type", node_type)
+
+    # UserPromptNode attributes
+    if "UserPrompt" in node_type:
+        _set_user_prompt_node_attributes(span, node)
+
+    # ModelRequestNode attributes
+    elif "ModelRequest" in node_type:
+        _set_model_request_node_attributes(span, node)
+
+    # CallToolsNode attributes
+    elif "CallTools" in node_type:
+        _set_call_tools_node_attributes(span, node)
+
+    # End node attributes
+    elif "End" in node_type:
+        _set_end_node_attributes(span, node)
+
+    # Generic node attributes for any other node types
+    else:
+        _set_generic_node_attributes(span, node)
+
+
+def _set_user_prompt_node_attributes(span: Any, node: Any) -> None:
+    """Set attributes specific to UserPromptNode."""
+    # User prompt content
+    user_prompt = _safe_get_attribute(node, "user_prompt")
+    if user_prompt:
+        _safe_set_attribute(span, f"{SpanAttributes.LLM_PROMPTS}.0.role", "user")
+        _safe_set_attribute(span, f"{SpanAttributes.LLM_PROMPTS}.0.content", user_prompt, MAX_CONTENT_LENGTH)
+        _safe_set_attribute(span, "pydantic_ai.user_prompt", user_prompt, MAX_CONTENT_LENGTH)
+
+    # Instructions
+    instructions = _safe_get_attribute(node, "instructions")
+    _safe_set_attribute(span, "pydantic_ai.instructions", instructions, MAX_CONTENT_LENGTH)
+
+    # Instructions functions
+    instructions_functions = _safe_get_attribute(node, "instructions_functions")
+    if instructions_functions:
+        _safe_set_attribute(span, "pydantic_ai.instructions_functions_count", len(instructions_functions))
+        for i, func in enumerate(instructions_functions[:MAX_FUNCTIONS_TO_PROCESS]):
+            func_name = _safe_get_attribute(func, "__name__")
+            _safe_set_attribute(span, f"pydantic_ai.instructions_functions.{i}.name", func_name)
+
+    # System prompts
+    system_prompts = _safe_get_attribute(node, "system_prompts")
+    if system_prompts:
+        _safe_set_attribute(span, "pydantic_ai.system_prompts_count", len(system_prompts))
+        for i, prompt in enumerate(system_prompts[:MAX_FUNCTIONS_TO_PROCESS]):
+            _safe_set_attribute(span, f"pydantic_ai.system_prompts.{i}", prompt, MAX_ARGS_LENGTH)
+
+    # System prompt functions
+    system_prompt_functions = _safe_get_attribute(node, "system_prompt_functions")
+    if system_prompt_functions:
+        _safe_set_attribute(span, "pydantic_ai.system_prompt_functions_count", len(system_prompt_functions))
+
+    # System prompt dynamic functions
+    system_prompt_dynamic_functions = _safe_get_attribute(node, "system_prompt_dynamic_functions")
+    if system_prompt_dynamic_functions:
+        _safe_set_attribute(
+            span, "pydantic_ai.system_prompt_dynamic_functions_count", len(system_prompt_dynamic_functions)
+        )
+        for key in list(system_prompt_dynamic_functions.keys())[:MAX_FUNCTIONS_TO_PROCESS]:
+            func_type = type(system_prompt_dynamic_functions[key]).__name__
+            _safe_set_attribute(span, f"pydantic_ai.system_prompt_dynamic_functions.{key}", func_type)
+
+
+def _set_model_request_node_attributes(span: Any, node: Any) -> None:
+    """Set attributes specific to ModelRequestNode."""
+    request = _safe_get_attribute(node, "request")
+    if not request:
+        return
+
+    # Request parts
+    parts = _safe_get_attribute(request, "parts")
+    if parts:
+        _safe_set_attribute(span, "pydantic_ai.request.parts_count", len(parts))
+
+        for i, part in enumerate(parts[:MAX_ITEMS_TO_PROCESS]):
+            part_type = type(part).__name__
+            _safe_set_attribute(span, f"pydantic_ai.request.parts.{i}.type", part_type)
+
+            # Content for text parts
+            content = _safe_get_attribute(part, "content")
+            if content:
+                _safe_set_attribute(span, f"{SpanAttributes.LLM_PROMPTS}.{i}.content", content, MAX_CONTENT_LENGTH)
+                _safe_set_attribute(span, f"pydantic_ai.request.parts.{i}.content", content, MAX_CONTENT_LENGTH)
+
+            # Role for message parts
+            role = _safe_get_attribute(part, "role")
+            if role:
+                _safe_set_attribute(span, f"{SpanAttributes.LLM_PROMPTS}.{i}.role", role)
+                _safe_set_attribute(span, f"pydantic_ai.request.parts.{i}.role", role)
+
+            # Timestamp, tool call information
+            _safe_set_attribute(
+                span, f"pydantic_ai.request.parts.{i}.timestamp", _safe_get_attribute(part, "timestamp")
+            )
+            _safe_set_attribute(
+                span, f"pydantic_ai.request.parts.{i}.tool_name", _safe_get_attribute(part, "tool_name")
+            )
+            _safe_set_attribute(
+                span, f"pydantic_ai.request.parts.{i}.tool_call_id", _safe_get_attribute(part, "tool_call_id")
+            )
+            _safe_set_attribute(
+                span, f"pydantic_ai.request.parts.{i}.args", _safe_get_attribute(part, "args"), MAX_ARGS_LENGTH
+            )
+
+    # Request metadata
+    _safe_set_attribute(span, "pydantic_ai.request.model_name", _safe_get_attribute(request, "model_name"))
+    _safe_set_attribute(span, f"{SpanAttributes.LLM_REQUEST_TEMPERATURE}", _safe_get_attribute(request, "temperature"))
+    _safe_set_attribute(span, f"{SpanAttributes.LLM_REQUEST_MAX_TOKENS}", _safe_get_attribute(request, "max_tokens"))
+
+
+def _set_call_tools_node_attributes(span: Any, node: Any) -> None:
+    """Set attributes specific to CallToolsNode."""
+    response = _safe_get_attribute(node, "model_response")
+    if not response:
+        return
+
+    # Response parts
+    parts = _safe_get_attribute(response, "parts")
+    if parts:
+        _safe_set_attribute(span, "pydantic_ai.response.parts_count", len(parts))
+
+        for i, part in enumerate(parts[:MAX_ITEMS_TO_PROCESS]):
+            part_type = type(part).__name__
+            _safe_set_attribute(span, f"pydantic_ai.response.parts.{i}.type", part_type)
+
+            # Content for text parts
+            content = _safe_get_attribute(part, "content")
+            if content:
+                _safe_set_attribute(span, f"{SpanAttributes.LLM_COMPLETIONS}.{i}.content", content, MAX_CONTENT_LENGTH)
+                _safe_set_attribute(span, f"pydantic_ai.response.parts.{i}.content", content, MAX_CONTENT_LENGTH)
+                _safe_set_attribute(span, f"{SpanAttributes.LLM_COMPLETIONS}.{i}.role", "assistant")
+
+            # Tool call information
+            _safe_set_attribute(
+                span, f"pydantic_ai.response.parts.{i}.tool_name", _safe_get_attribute(part, "tool_name")
+            )
+            _safe_set_attribute(
+                span, f"pydantic_ai.response.parts.{i}.tool_call_id", _safe_get_attribute(part, "tool_call_id")
+            )
+            _safe_set_attribute(
+                span, f"pydantic_ai.response.parts.{i}.args", _safe_get_attribute(part, "args"), MAX_ARGS_LENGTH
+            )
+
+    # Usage information
+    usage = _safe_get_attribute(response, "usage")
+    if usage:
+        _safe_set_attribute(span, "pydantic_ai.usage.requests", _safe_get_attribute(usage, "requests"))
+
+        # Token usage with dual attributes for compatibility
+        request_tokens = _safe_get_attribute(usage, "request_tokens")
+        if request_tokens is not None:
+            _safe_set_attribute(span, f"{SpanAttributes.LLM_USAGE_PROMPT_TOKENS}", request_tokens)
+            _safe_set_attribute(span, "pydantic_ai.usage.request_tokens", request_tokens)
+
+        response_tokens = _safe_get_attribute(usage, "response_tokens")
+        if response_tokens is not None:
+            _safe_set_attribute(span, f"{SpanAttributes.LLM_USAGE_COMPLETION_TOKENS}", response_tokens)
+            _safe_set_attribute(span, "pydantic_ai.usage.response_tokens", response_tokens)
+
+        total_tokens = _safe_get_attribute(usage, "total_tokens")
+        if total_tokens is not None:
+            _safe_set_attribute(span, f"{SpanAttributes.LLM_USAGE_TOTAL_TOKENS}", total_tokens)
+            _safe_set_attribute(span, "pydantic_ai.usage.total_tokens", total_tokens)
+
+        # Additional usage details
+        details = _safe_get_attribute(usage, "details")
+        if details:
+            for key, value in details.items():
+                if value is not None:
+                    _safe_set_attribute(span, f"pydantic_ai.usage.details.{key}", value)
+
+    # Model information (only for CallToolsNode)
+    model_name = _safe_get_attribute(response, "model_name")
+    if model_name:
+        _safe_set_attribute(span, f"{SpanAttributes.LLM_RESPONSE_MODEL}", model_name)
+        _safe_set_attribute(span, f"{SpanAttributes.LLM_REQUEST_MODEL}", model_name)
+        _safe_set_attribute(span, "pydantic_ai.response.model_name", model_name)
+
+    # Timestamp
+    _safe_set_attribute(span, "pydantic_ai.response.timestamp", _safe_get_attribute(response, "timestamp"))
+
+    # Tool execution results (if available)
+    tool_results = _safe_get_attribute(node, "tool_results")
+    if tool_results:
+        _safe_set_attribute(span, "pydantic_ai.tool_results_count", len(tool_results))
+        for i, result in enumerate(tool_results[:MAX_FUNCTIONS_TO_PROCESS]):
+            _safe_set_attribute(span, f"pydantic_ai.tool_results.{i}", result, MAX_ARGS_LENGTH)
+
+
+def _set_end_node_attributes(span: Any, node: Any) -> None:
+    """Set attributes specific to End node."""
+    data = _safe_get_attribute(node, "data")
+    if not data:
+        return
+
+    # Final output
+    output = _safe_get_attribute(data, "output")
+    if output is not None:
+        _safe_set_attribute(span, f"{SpanAttributes.LLM_COMPLETIONS}.0.role", "assistant")
+        _safe_set_attribute(span, f"{SpanAttributes.LLM_COMPLETIONS}.0.content", output, MAX_CONTENT_LENGTH)
+        _safe_set_attribute(span, "pydantic_ai.final_output", output, MAX_CONTENT_LENGTH)
+
+    # Cost information
+    cost = _safe_get_attribute(data, "cost")
+    if cost is not None:
+        _safe_set_attribute(span, "pydantic_ai.cost", float(cost))
+
+    # Usage summary
+    usage = _safe_get_attribute(data, "usage")
+    if usage:
+        _safe_set_attribute(span, "pydantic_ai.final_usage.total_tokens", _safe_get_attribute(usage, "total_tokens"))
+        _safe_set_attribute(
+            span, "pydantic_ai.final_usage.request_tokens", _safe_get_attribute(usage, "request_tokens")
+        )
+        _safe_set_attribute(
+            span, "pydantic_ai.final_usage.response_tokens", _safe_get_attribute(usage, "response_tokens")
+        )
+
+    # Messages history
+    messages = _safe_get_attribute(data, "messages")
+    if messages:
+        _safe_set_attribute(span, "pydantic_ai.messages_count", len(messages))
+
+    # New messages
+    new_messages = _safe_get_attribute(data, "new_messages")
+    if new_messages:
+        _safe_set_attribute(span, "pydantic_ai.new_messages_count", len(new_messages))
+
+
+def _set_generic_node_attributes(span: Any, node: Any) -> None:
+    """Set attributes for any other node types."""
+    # Try to extract common attributes that might be available
+    for attr_name in ["content", "message", "value", "result", "error"]:
+        attr_value = _safe_get_attribute(node, attr_name)
+        _safe_set_attribute(span, f"pydantic_ai.{attr_name}", attr_value, MAX_CONTENT_LENGTH)

netra_sdk-0.1.25/netra/instrumentation/pydantic_ai/version.py

@@ -0,0 +1 @@
+__version__ = "1.0.0"