netra-sdk 0.1.36__tar.gz → 0.1.38__tar.gz

{netra_sdk-0.1.36 → netra_sdk-0.1.38}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: netra-sdk
-Version: 0.1.36
+Version: 0.1.38
 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

{netra_sdk-0.1.36 → netra_sdk-0.1.38}/netra/__init__.py

@@ -222,16 +222,17 @@ class Netra:
     @classmethod
     def set_custom_attributes(cls, key: str, value: Any) -> None:
         """
-        Set custom attributes context in the current OpenTelemetry context.
+        Set a custom attribute on the currently active OpenTelemetry span only.
 
         Args:
             key: Custom attribute key
             value: Custom attribute value
         """
         if key and value:
-            SessionManager.set_session_context("custom_attributes", {key: value})
+            SessionManager.set_attribute_on_active_span(f"{Config.LIBRARY_NAME}.custom.{key}", value)
         else:
             logger.warning("Both key and value must be provided for custom attributes.")
+            return
 
     @classmethod
     def set_custom_event(cls, event_name: str, attributes: Any) -> None:

{netra_sdk-0.1.36 → netra_sdk-0.1.38}/netra/instrumentation/litellm/wrappers.py

@@ -47,115 +47,127 @@ def set_request_attributes(span: Span, kwargs: Dict[str, Any], operation_type: s
     """Set request attributes on span"""
     if not span.is_recording():
         return
-
-    # Set operation type
-    span.set_attribute(f"{SpanAttributes.LLM_REQUEST_TYPE}", operation_type)
-    span.set_attribute(f"{SpanAttributes.LLM_SYSTEM}", "LiteLLM")
-
-    # Common attributes
-    if kwargs.get("model"):
-        span.set_attribute(f"{SpanAttributes.LLM_REQUEST_MODEL}", kwargs["model"])
-
-    if kwargs.get("temperature") is not None:
-        span.set_attribute(f"{SpanAttributes.LLM_REQUEST_TEMPERATURE}", kwargs["temperature"])
-
-    if kwargs.get("max_tokens") is not None:
-        span.set_attribute(f"{SpanAttributes.LLM_REQUEST_MAX_TOKENS}", kwargs["max_tokens"])
-
-    if kwargs.get("stream") is not None:
-        span.set_attribute("gen_ai.stream", kwargs["stream"])
-
-    # Chat completion specific attributes
-    if operation_type == "chat" and kwargs.get("messages"):
-        messages = kwargs["messages"]
-        if isinstance(messages, list) and len(messages) > 0:
-            for index, message in enumerate(messages):
-                if isinstance(message, dict):
-                    span.set_attribute(f"{SpanAttributes.LLM_PROMPTS}.{index}.role", message.get("role", "user"))
-                    span.set_attribute(f"{SpanAttributes.LLM_PROMPTS}.{index}.content", str(message.get("content", "")))
-
-    # Embedding specific attributes
-    if operation_type == "embedding" and kwargs.get("input"):
-        input_data = kwargs["input"]
-        if isinstance(input_data, str):
-            span.set_attribute(f"{SpanAttributes.LLM_PROMPTS}.0.content", input_data)
-        elif isinstance(input_data, list):
-            for index, text in enumerate(input_data):
-                if isinstance(text, str):
-                    span.set_attribute(f"{SpanAttributes.LLM_PROMPTS}.{index}.content", text)
-
-    # Image generation specific attributes
-    if operation_type == "image_generation":
-        if kwargs.get("prompt"):
-            span.set_attribute("gen_ai.prompt", kwargs["prompt"])
-        if kwargs.get("n"):
-            span.set_attribute("gen_ai.request.n", kwargs["n"])
-        if kwargs.get("size"):
-            span.set_attribute("gen_ai.request.size", kwargs["size"])
-        if kwargs.get("quality"):
-            span.set_attribute("gen_ai.request.quality", kwargs["quality"])
-        if kwargs.get("style"):
-            span.set_attribute("gen_ai.request.style", kwargs["style"])
+    try:
+        # Set operation type
+        span.set_attribute(f"{SpanAttributes.LLM_REQUEST_TYPE}", operation_type)
+        span.set_attribute(f"{SpanAttributes.LLM_SYSTEM}", "LiteLLM")
+
+        # Common attributes
+        if kwargs.get("model"):
+            span.set_attribute(f"{SpanAttributes.LLM_REQUEST_MODEL}", kwargs["model"])
+
+        if kwargs.get("temperature") is not None:
+            span.set_attribute(f"{SpanAttributes.LLM_REQUEST_TEMPERATURE}", kwargs["temperature"])
+
+        if kwargs.get("max_tokens") is not None:
+            span.set_attribute(f"{SpanAttributes.LLM_REQUEST_MAX_TOKENS}", kwargs["max_tokens"])
+
+        if kwargs.get("stream") is not None:
+            span.set_attribute("gen_ai.stream", kwargs["stream"])
+
+        # Chat completion specific attributes
+        if operation_type == "chat" and kwargs.get("messages"):
+            messages = kwargs["messages"]
+            if isinstance(messages, list) and len(messages) > 0:
+                for index, message in enumerate(messages):
+                    if isinstance(message, dict):
+                        span.set_attribute(f"{SpanAttributes.LLM_PROMPTS}.{index}.role", message.get("role", "user"))
+                        span.set_attribute(
+                            f"{SpanAttributes.LLM_PROMPTS}.{index}.content", str(message.get("content", ""))
+                        )
+
+        # Embedding specific attributes
+        if operation_type == "embedding" and kwargs.get("input"):
+            input_data = kwargs["input"]
+            if isinstance(input_data, str):
+                span.set_attribute(f"{SpanAttributes.LLM_PROMPTS}.0.content", input_data)
+            elif isinstance(input_data, list):
+                for index, text in enumerate(input_data):
+                    if isinstance(text, str):
+                        span.set_attribute(f"{SpanAttributes.LLM_PROMPTS}.{index}.content", text)
+
+        # Image generation specific attributes
+        if operation_type == "image_generation":
+            if kwargs.get("prompt"):
+                span.set_attribute("gen_ai.prompt", kwargs["prompt"])
+            if kwargs.get("n"):
+                span.set_attribute("gen_ai.request.n", kwargs["n"])
+            if kwargs.get("size"):
+                span.set_attribute("gen_ai.request.size", kwargs["size"])
+            if kwargs.get("quality"):
+                span.set_attribute("gen_ai.request.quality", kwargs["quality"])
+            if kwargs.get("style"):
+                span.set_attribute("gen_ai.request.style", kwargs["style"])
+    except Exception as e:
+        logger.error(f"Failed to set attributes for LiteLLM span: {e}")
 
 
 def set_response_attributes(span: Span, response_dict: Dict[str, Any], operation_type: str) -> None:
     """Set response attributes on span"""
     if not span.is_recording():
         return
+    try:
+        if response_dict.get("model"):
+            span.set_attribute(f"{SpanAttributes.LLM_RESPONSE_MODEL}", response_dict["model"])
+
+        if response_dict.get("id"):
+            span.set_attribute("gen_ai.response.id", response_dict["id"])
+
+        # Usage information
+        usage = response_dict.get("usage", {})
+        if usage:
+            if usage.get("prompt_tokens"):
+                span.set_attribute(f"{SpanAttributes.LLM_USAGE_PROMPT_TOKENS}", usage["prompt_tokens"])
+            if usage.get("completion_tokens"):
+                span.set_attribute(f"{SpanAttributes.LLM_USAGE_COMPLETION_TOKENS}", usage["completion_tokens"])
+            if usage.get("cache_read_input_tokens"):
+                span.set_attribute(
+                    f"{SpanAttributes.LLM_USAGE_CACHE_READ_INPUT_TOKENS}", usage["cache_read_input_tokens"]
+                )
+            if usage.get("cache_creation_input_tokens"):
+                span.set_attribute("gen_ai.usage.cache_creation_input_tokens", usage["cache_creation_input_tokens"])
+            if usage.get("total_tokens"):
+                span.set_attribute(f"{SpanAttributes.LLM_USAGE_TOTAL_TOKENS}", usage["total_tokens"])
+
+        # Chat completion response content
+        if operation_type == "chat":
+            choices = response_dict.get("choices", [])
+            for index, choice in enumerate(choices):
+                if choice.get("message", {}).get("role"):
+                    span.set_attribute(f"{SpanAttributes.LLM_COMPLETIONS}.{index}.role", choice["message"]["role"])
+                if choice.get("message", {}).get("content"):
+                    span.set_attribute(
+                        f"{SpanAttributes.LLM_COMPLETIONS}.{index}.content", choice["message"]["content"]
+                    )
+                if choice.get("finish_reason"):
+                    span.set_attribute(
+                        f"{SpanAttributes.LLM_COMPLETIONS}.{index}.finish_reason", choice["finish_reason"]
+                    )
 
-    if response_dict.get("model"):
-        span.set_attribute(f"{SpanAttributes.LLM_RESPONSE_MODEL}", response_dict["model"])
-
-    if response_dict.get("id"):
-        span.set_attribute("gen_ai.response.id", response_dict["id"])
-
-    # Usage information
-    usage = response_dict.get("usage", {})
-    if usage:
-        if usage.get("prompt_tokens"):
-            span.set_attribute(f"{SpanAttributes.LLM_USAGE_PROMPT_TOKENS}", usage["prompt_tokens"])
-        if usage.get("completion_tokens"):
-            span.set_attribute(f"{SpanAttributes.LLM_USAGE_COMPLETION_TOKENS}", usage["completion_tokens"])
-        if usage.get("cache_read_input_tokens"):
-            span.set_attribute(f"{SpanAttributes.LLM_USAGE_CACHE_READ_INPUT_TOKENS}", usage["cache_read_input_tokens"])
-        if usage.get("cache_creation_input_tokens"):
-            span.set_attribute("gen_ai.usage.cache_creation_input_tokens", usage["cache_creation_input_tokens"])
-        if usage.get("total_tokens"):
-            span.set_attribute(f"{SpanAttributes.LLM_USAGE_TOTAL_TOKENS}", usage["total_tokens"])
-
-    # Chat completion response content
-    if operation_type == "chat":
-        choices = response_dict.get("choices", [])
-        for index, choice in enumerate(choices):
-            if choice.get("message", {}).get("role"):
-                span.set_attribute(f"{SpanAttributes.LLM_COMPLETIONS}.{index}.role", choice["message"]["role"])
-            if choice.get("message", {}).get("content"):
-                span.set_attribute(f"{SpanAttributes.LLM_COMPLETIONS}.{index}.content", choice["message"]["content"])
-            if choice.get("finish_reason"):
-                span.set_attribute(f"{SpanAttributes.LLM_COMPLETIONS}.{index}.finish_reason", choice["finish_reason"])
-
-    # Embedding response content
-    elif operation_type == "embedding":
-        data = response_dict.get("data", [])
-        for index, embedding_data in enumerate(data):
-            if embedding_data.get("index") is not None:
-                span.set_attribute(f"gen_ai.response.embeddings.{index}.index", embedding_data["index"])
-            if embedding_data.get("embedding"):
-                # Don't log the actual embedding vector, just its dimensions
-                embedding_vector = embedding_data["embedding"]
-                if isinstance(embedding_vector, list):
-                    span.set_attribute(f"gen_ai.response.embeddings.{index}.dimensions", len(embedding_vector))
-
-    # Image generation response content
-    elif operation_type == "image_generation":
-        data = response_dict.get("data", [])
-        for index, image_data in enumerate(data):
-            if image_data.get("url"):
-                span.set_attribute(f"gen_ai.response.images.{index}.url", image_data["url"])
-            if image_data.get("b64_json"):
-                span.set_attribute(f"gen_ai.response.images.{index}.has_b64_json", True)
-            if image_data.get("revised_prompt"):
-                span.set_attribute(f"gen_ai.response.images.{index}.revised_prompt", image_data["revised_prompt"])
+        # Embedding response content
+        elif operation_type == "embedding":
+            data = response_dict.get("data", [])
+            for index, embedding_data in enumerate(data):
+                if embedding_data.get("index") is not None:
+                    span.set_attribute(f"gen_ai.response.embeddings.{index}.index", embedding_data["index"])
+                if embedding_data.get("embedding"):
+                    # Don't log the actual embedding vector, just its dimensions
+                    embedding_vector = embedding_data["embedding"]
+                    if isinstance(embedding_vector, list):
+                        span.set_attribute(f"gen_ai.response.embeddings.{index}.dimensions", len(embedding_vector))
+
+        # Image generation response content
+        elif operation_type == "image_generation":
+            data = response_dict.get("data", [])
+            for index, image_data in enumerate(data):
+                if image_data.get("url"):
+                    span.set_attribute(f"gen_ai.response.images.{index}.url", image_data["url"])
+                if image_data.get("b64_json"):
+                    span.set_attribute(f"gen_ai.response.images.{index}.has_b64_json", True)
+                if image_data.get("revised_prompt"):
+                    span.set_attribute(f"gen_ai.response.images.{index}.revised_prompt", image_data["revised_prompt"])
+    except Exception as e:
+        logger.error(f"Failed to set attributes for LiteLLM span: {e}")
 
 
 def completion_wrapper(tracer: Tracer) -> Callable[..., Any]:

{netra_sdk-0.1.36 → netra_sdk-0.1.38}/netra/processors/instrumentation_span_processor.py

@@ -30,13 +30,15 @@ class InstrumentationSpanProcessor(SpanProcessor):  # type: ignore[misc]
         if scope is not None:
             name = getattr(scope, "name", None)
             if isinstance(name, str) and name:
-                # Normalize common pattern like 'opentelemetry.instrumentation.httpx' -> 'httpx'
-                try:
-                    base = name.rsplit(".", 1)[-1].strip()
-                    if base:
-                        return base
-                except Exception:
-                    pass
+                # Only truncate when coming from known namespaces
+                if name.startswith("opentelemetry.instrumentation.") or name.startswith("netra.instrumentation."):
+                    try:
+                        base = name.rsplit(".", 1)[-1].strip()
+                        if base:
+                            return base
+                    except Exception:
+                        pass
+                # Otherwise, return as-is
                 return name
         return None
 
@@ -87,7 +89,8 @@ class InstrumentationSpanProcessor(SpanProcessor):  # type: ignore[misc]
 
             # Set this span's instrumentation name
             name = self._detect_raw_instrumentation_name(span)
-            if name and any(allowed in name for allowed in ALLOWED_INSTRUMENTATION_NAMES):
+            print(name)
+            if name in ALLOWED_INSTRUMENTATION_NAMES:
                 span.set_attribute(f"{Config.LIBRARY_NAME}.instrumentation.name", name)
         except Exception:
             pass

{netra_sdk-0.1.36 → netra_sdk-0.1.38}/netra/session_manager.py

@@ -196,9 +196,21 @@ class SessionManager:
         }
 
     @staticmethod
-    def set_session_context(session_key: str, value: Union[str, Dict[str, str]]) -> None:
+    def set_session_context(
+        session_key: str,
+        value: Union[str, Dict[str, str]],
+        attach_globally: bool = False,
+    ) -> None:
         """
-        Set session context attributes in the current OpenTelemetry baggage.
+        Set session context attributes in OpenTelemetry baggage.
+
+        Behavior:
+        - Adds values to baggage on the current context.
+        - To avoid context token mismatch errors, this method only attaches the
+          modified context when it is safe (no active recording span), unless
+          forced via attach_globally=True.
+        - When not attaching (e.g., inside an active span), it annotates the
+          current span with equivalent attributes for visibility.
 
         Args:
             session_key: Key to set in baggage (session_id, user_id, tenant_id, or custom_attributes)
@@ -219,7 +231,20 @@ class SessionManager:
                     ctx = baggage.set_baggage("custom_keys", ",".join(custom_keys), ctx)
                     for key, val in value.items():
                         ctx = baggage.set_baggage(f"custom.{key}", str(val), ctx)
-            otel_context.attach(ctx)
+
+            # Decide whether to attach globally. We only attach if there is no
+            # active recording span (safe point) or if the caller forces it.
+            current_span = trace.get_current_span()
+            has_active_span = bool(current_span and getattr(current_span, "is_recording", lambda: False)())
+            if attach_globally or not has_active_span:
+                otel_context.attach(ctx)
+            else:
+                # Best-effort: annotate the current span for observability
+                if isinstance(value, str) and value:
+                    current_span.set_attribute(f"{Config.LIBRARY_NAME}.session.{session_key}", value)
+                elif isinstance(value, dict) and value and session_key == "custom_attributes":
+                    for key, val in value.items():
+                        current_span.set_attribute(f"{Config.LIBRARY_NAME}.session.custom.{key}", str(val))
         except Exception as e:
             logger.exception(f"Failed to set session context for key={session_key}: {e}")
 
@@ -321,3 +346,31 @@ class SessionManager:
             candidate.set_attribute(attr_key, attr_str)
         except Exception as e:
             logger.exception("Failed setting attribute %s: %s", attr_key, e)
+
+    @staticmethod
+    def set_attribute_on_active_span(attr_key: str, attr_value: Any) -> None:
+        """
+        Set an attribute strictly on the currently active OpenTelemetry span.
+
+        - Does not fall back to any root span.
+        - Does not mutate global baggage/context.
+        - If no active recording span is present, logs a warning and returns.
+        """
+        try:
+            span = trace.get_current_span()
+            if span and getattr(span, "is_recording", lambda: False)():
+                # Convert attr_value to a JSON-safe string if needed
+                try:
+                    if isinstance(attr_value, str):
+                        v = attr_value
+                    else:
+                        import json
+
+                        v = json.dumps(attr_value)
+                except Exception:
+                    v = str(attr_value)
+                span.set_attribute(attr_key, v)
+            else:
+                logger.warning("No active span to set attribute '%s'", attr_key)
+        except Exception:
+            logger.exception("Failed to set attribute '%s' on active span", attr_key)

netra_sdk-0.1.38/netra/version.py

@@ -0,0 +1 @@
+__version__ = "0.1.38"

{netra_sdk-0.1.36 → netra_sdk-0.1.38}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "poetry.core.masonry.api"
 
 [project]
 name = "netra-sdk"
-version = "0.1.36"
+version = "0.1.38"
 description = "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."
 authors = [
     {name = "Sooraj Thomas",email = "sooraj@keyvalue.systems"}

netra_sdk-0.1.36/netra/version.py

@@ -1 +0,0 @@
-__version__ = "0.1.36"