lmnr 0.6.12__tar.gz → 0.6.14__tar.gz

{lmnr-0.6.12 → lmnr-0.6.14}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: lmnr
-Version: 0.6.12
+Version: 0.6.14
 Summary: Python SDK for Laminar
 License: Apache-2.0
 Author: lmnr.ai

{lmnr-0.6.12 → lmnr-0.6.14}/pyproject.toml

@@ -6,7 +6,7 @@
 
 [project]
 name = "lmnr"
-version = "0.6.12"
+version = "0.6.14"
 description = "Python SDK for Laminar"
 authors = [
   { name = "lmnr.ai", email = "founders@lmnr.ai" }

{lmnr-0.6.12 → lmnr-0.6.14}/src/lmnr/opentelemetry_lib/litellm/__init__.py

@@ -347,7 +347,7 @@ try:
                 self._process_response_choices(span, response_dict.get("choices"))
 
 except ImportError as e:
-    logger.warning(f"LiteLLM callback unavailable: {e}")
+    logger.debug(f"LiteLLM callback unavailable: {e}")
 
     # Create a no-op logger when LiteLLM is not available
     class LaminarLiteLLMCallback:

{lmnr-0.6.12 → lmnr-0.6.14}/src/lmnr/opentelemetry_lib/opentelemetry/instrumentation/google_genai/__init__.py

@@ -12,7 +12,6 @@ from .config import (
     Config,
 )
 from .utils import (
-    ProcessedContentPart,
     dont_throw,
     get_content,
     role_from_content_union,
@@ -130,27 +129,30 @@ def _set_request_attributes(span, args, kwargs):
     )
 
     tools: list[types.FunctionDeclaration] = []
-    if kwargs.get("tools"):
-        for tool in kwargs.get("tools"):
+    arg_tools = config_dict.get("tools", kwargs.get("tools"))
+    if arg_tools:
+        for tool in arg_tools:
             if isinstance(tool, types.Tool):
                 tools += tool.function_declarations or []
             elif isinstance(tool, Callable):
                 tools.append(types.FunctionDeclaration.from_callable(tool))
+
     for tool_num, tool in enumerate(tools):
+        tool_dict = to_dict(tool)
         set_span_attribute(
             span,
             f"{SpanAttributes.LLM_REQUEST_FUNCTIONS}.{tool_num}.name",
-            to_dict(tool).get("name"),
+            tool_dict.get("name"),
         )
         set_span_attribute(
             span,
             f"{SpanAttributes.LLM_REQUEST_FUNCTIONS}.{tool_num}.description",
-            to_dict(tool).get("description"),
+            tool_dict.get("description"),
         )
         set_span_attribute(
             span,
             f"{SpanAttributes.LLM_REQUEST_FUNCTIONS}.{tool_num}.parameters",
-            to_dict(tool).get("parameters"),
+            json.dumps(tool_dict.get("parameters")),
         )
 
     if should_send_prompts():
@@ -162,7 +164,9 @@ def _set_request_attributes(span, args, kwargs):
             set_span_attribute(
                 span,
                 f"{gen_ai_attributes.GEN_AI_PROMPT}.{i}.content",
-                (get_content(process_content_union(system_instruction)) or {}).get("text", ""),
+                (get_content(process_content_union(system_instruction)) or {}).get(
+                    "text", ""
+                ),
             )
             set_span_attribute(
                 span, f"{gen_ai_attributes.GEN_AI_PROMPT}.{i}.role", "system"
@@ -174,6 +178,7 @@ def _set_request_attributes(span, args, kwargs):
         for content in contents:
             processed_content = process_content_union(content)
             content_str = get_content(processed_content)
+
             set_span_attribute(
                 span,
                 f"{gen_ai_attributes.GEN_AI_PROMPT}.{i}.content",
@@ -188,26 +193,35 @@ def _set_request_attributes(span, args, kwargs):
                 if isinstance(processed_content, list)
                 else [processed_content]
             )
-            for j, block in enumerate(blocks):
+            tool_call_index = 0
+            for block in blocks:
                 block_dict = to_dict(block)
+
                 if not block_dict.get("function_call"):
                     continue
                 function_call = to_dict(block_dict.get("function_call", {}))
+
                 set_span_attribute(
                     span,
-                    f"{gen_ai_attributes.GEN_AI_PROMPT}.{i}.tool_calls.{j}.name",
+                    f"{gen_ai_attributes.GEN_AI_PROMPT}.{i}.tool_calls.{tool_call_index}.name",
                     function_call.get("name"),
                 )
                 set_span_attribute(
                     span,
-                    f"{gen_ai_attributes.GEN_AI_PROMPT}.{i}.tool_calls.{j}.id",
-                    function_call.get("id"),
+                    f"{gen_ai_attributes.GEN_AI_PROMPT}.{i}.tool_calls.{tool_call_index}.id",
+                    (
+                        function_call.get("id")
+                        if function_call.get("id") is not None
+                        else function_call.get("name")
+                    ),  # google genai doesn't support tool call ids
                 )
                 set_span_attribute(
                     span,
-                    f"{gen_ai_attributes.GEN_AI_PROMPT}.{i}.tool_calls.{j}.arguments",
+                    f"{gen_ai_attributes.GEN_AI_PROMPT}.{i}.tool_calls.{tool_call_index}.arguments",
                     json.dumps(function_call.get("arguments")),
                 )
+                tool_call_index += 1
+
             set_span_attribute(
                 span,
                 f"{gen_ai_attributes.GEN_AI_PROMPT}.{i}.role",
@@ -258,21 +272,8 @@ def _set_response_attributes(span, response: types.GenerateContentResponse):
         candidates_list = candidates if isinstance(candidates, list) else [candidates]
         for i, candidate in enumerate(candidates_list):
             processed_content = process_content_union(candidate.content)
-            if isinstance(processed_content, list):
-                if all(
-                    isinstance(item, dict) and item.get("type") == "text"
-                    for item in processed_content
-                ):
-                    content_str = processed_content[0]["text"]
-                elif all(
-                    isinstance(item, ProcessedContentPart) and item.content
-                    for item in processed_content
-                ):
-                    content_str = processed_content[0].content
-                else:
-                    content_str = get_content(processed_content)
-            else:
-                content_str = get_content(processed_content)
+            content_str = get_content(processed_content)
+
             set_span_attribute(
                 span, f"{gen_ai_attributes.GEN_AI_COMPLETION}.{i}.role", "model"
             )
@@ -290,26 +291,33 @@ def _set_response_attributes(span, response: types.GenerateContentResponse):
                 if isinstance(processed_content, list)
                 else [processed_content]
             )
-            for j, block in enumerate(blocks):
+
+            tool_call_index = 0
+            for block in blocks:
                 block_dict = to_dict(block)
                 if not block_dict.get("function_call"):
                     continue
                 function_call = to_dict(block_dict.get("function_call", {}))
                 set_span_attribute(
                     span,
-                    f"{gen_ai_attributes.GEN_AI_COMPLETION}.{i}.tool_calls.{j}.name",
+                    f"{gen_ai_attributes.GEN_AI_COMPLETION}.{i}.tool_calls.{tool_call_index}.name",
                     function_call.get("name"),
                 )
                 set_span_attribute(
                     span,
-                    f"{gen_ai_attributes.GEN_AI_COMPLETION}.{i}.tool_calls.{j}.id",
-                    function_call.get("id"),
+                    f"{gen_ai_attributes.GEN_AI_COMPLETION}.{i}.tool_calls.{tool_call_index}.id",
+                    (
+                        function_call.get("id")
+                        if function_call.get("id") is not None
+                        else function_call.get("name")
+                    ),  # google genai doesn't support tool call ids
                 )
                 set_span_attribute(
                     span,
-                    f"{gen_ai_attributes.GEN_AI_COMPLETION}.{i}.tool_calls.{j}.arguments",
+                    f"{gen_ai_attributes.GEN_AI_COMPLETION}.{i}.tool_calls.{tool_call_index}.arguments",
                     json.dumps(function_call.get("arguments")),
                 )
+                tool_call_index += 1
 
 
 @dont_throw

{lmnr-0.6.12 → lmnr-0.6.14}/src/lmnr/opentelemetry_lib/opentelemetry/instrumentation/google_genai/utils.py

@@ -1,3 +1,4 @@
+import base64
 import logging
 import traceback
 
@@ -74,7 +75,8 @@ def to_dict(obj: BaseModel | pydantic.BaseModel | dict) -> dict[str, Any]:
             return obj
         else:
             return dict(obj)
-    except Exception:
+    except Exception as e:
+        logging.error(f"Error converting to dict: {obj}, error: {e}")
         return dict(obj)
 
 
@@ -96,7 +98,7 @@ def get_content(
         else:
             return None
     elif isinstance(content, list):
-        return [get_content(item) or "" for item in content if item is not None]
+        return [get_content(item) for item in content]
     elif isinstance(content, str):
         return {
             "type": "text",
@@ -226,11 +228,15 @@ def _process_image_item(
     content_index: int,
 ) -> ProcessedContentPart | dict | None:
     # Convert to openai format, so backends can handle it
+    data = blob.get("data")
+    encoded_data = (
+        base64.b64encode(data).decode("utf-8") if isinstance(data, bytes) else data
+    )
     return (
         ProcessedContentPart(
             image_url=ImageUrl(
                 image_url=ImageUrlInner(
-                    url=f"data:image/{blob.get('mime_type').split('/')[1]};base64,{blob.get('data')}",
+                    url=f"data:image/{blob.get('mime_type').split('/')[1]};base64,{encoded_data}",
                 )
             )
         )

{lmnr-0.6.12 → lmnr-0.6.14}/src/lmnr/sdk/client/asynchronous/resources/evals.py

@@ -15,13 +15,14 @@ class AsyncEvals(BaseAsyncResource):
     """Resource for interacting with Laminar evaluations API."""
 
     async def init(
-        self, name: str | None = None, group_name: str | None = None
+        self, name: str | None = None, group_name: str | None = None, metadata: dict[str, Any] | None = None
     ) -> InitEvaluationResponse:
         """Initialize a new evaluation.
 
         Args:
             name (str | None, optional): Name of the evaluation. Defaults to None.
             group_name (str | None, optional): Group name for the evaluation. Defaults to None.
+            metadata (dict[str, Any] | None, optional): Metadata to associate with. Defaults to None.
 
         Returns:
             InitEvaluationResponse: The response from the initialization request.
@@ -31,6 +32,7 @@ class AsyncEvals(BaseAsyncResource):
             json={
                 "name": name,
                 "groupName": group_name,
+                "metadata": metadata,
             },
             headers=self._headers(),
         )
@@ -45,6 +47,7 @@ class AsyncEvals(BaseAsyncResource):
         self,
         name: str | None = None,
         group_name: str | None = None,
+        metadata: dict[str, Any] | None = None,
     ) -> uuid.UUID:
         """
         Create a new evaluation and return its ID.
@@ -52,11 +55,12 @@ class AsyncEvals(BaseAsyncResource):
         Parameters:
             name (str | None, optional): Optional name of the evaluation.
             group_name (str | None, optional): An identifier to group evaluations.
-        
+            metadata (dict[str, Any] | None, optional): Metadata to associate with. Defaults to None.
+
         Returns:
             uuid.UUID: The evaluation ID.
         """
-        evaluation = await self.init(name=name, group_name=group_name)
+        evaluation = await self.init(name=name, group_name=group_name, metadata=metadata)
         return evaluation.id
 
     async def create_datapoint(

{lmnr-0.6.12 → lmnr-0.6.14}/src/lmnr/sdk/client/synchronous/resources/evals.py

@@ -17,13 +17,14 @@ class Evals(BaseResource):
     """Resource for interacting with Laminar evaluations API."""
 
     def init(
-        self, name: str | None = None, group_name: str | None = None
+        self, name: str | None = None, group_name: str | None = None, metadata: dict[str, Any] | None = None
     ) -> InitEvaluationResponse:
         """Initialize a new evaluation.
 
         Args:
             name (str | None, optional): Name of the evaluation. Defaults to None.
             group_name (str | None, optional): Group name for the evaluation. Defaults to None.
+            metadata (dict[str, Any] | None, optional): Metadata to associate with. Defaults to None.
 
         Returns:
             InitEvaluationResponse: The response from the initialization request.
@@ -33,6 +34,7 @@ class Evals(BaseResource):
             json={
                 "name": name,
                 "groupName": group_name,
+                "metadata": metadata,
             },
             headers=self._headers(),
         )
@@ -47,6 +49,7 @@ class Evals(BaseResource):
         self,
         name: str | None = None,
         group_name: str | None = None,
+        metadata: dict[str, Any] | None = None,
     ) -> uuid.UUID:
         """
         Create a new evaluation and return its ID.
@@ -54,11 +57,12 @@ class Evals(BaseResource):
         Parameters:
             name (str | None, optional): Optional name of the evaluation.
             group_name (str | None, optional): An identifier to group evaluations.
-        
+            metadata (dict[str, Any] | None, optional): Metadata to associate with. Defaults to None.
+
         Returns:
             uuid.UUID: The evaluation ID.
         """
-        evaluation = self.init(name=name, group_name=group_name)
+        evaluation = self.init(name=name, group_name=group_name, metadata=metadata)
         return evaluation.id
 
     def create_datapoint(

{lmnr-0.6.12 → lmnr-0.6.14}/src/lmnr/sdk/evaluations.py

@@ -104,6 +104,7 @@ class Evaluation:
         evaluators: dict[str, EvaluatorFunction | HumanEvaluator],
         name: str | None = None,
         group_name: str | None = None,
+        metadata: dict[str, Any] | None = None,
         concurrency_limit: int = DEFAULT_BATCH_SIZE,
         project_api_key: str | None = None,
         base_url: str | None = None,
@@ -143,6 +144,7 @@ class Evaluation:
                 evaluations. Only evaluations within the same group_name can be\
                 visually compared. If not provided, "default" is assigned.
                 Defaults to None
+            metadata (dict[str, Any] | None): optional metadata to associate with\
             concurrency_limit (int, optional): The concurrency limit for\
                 evaluation. This many data points will be evaluated in parallel\
                 with a pool of workers.
@@ -192,6 +194,7 @@ class Evaluation:
         self.evaluators = evaluators
         self.group_name = group_name
         self.name = name
+        self.metadata = metadata
         self.concurrency_limit = concurrency_limit
         self.batch_size = concurrency_limit
         self._logger = get_default_logger(self.__class__.__name__)
@@ -242,7 +245,7 @@ class Evaluation:
         self.reporter.start(len(self.data))
         try:
             evaluation = await self.client.evals.init(
-                name=self.name, group_name=self.group_name
+                name=self.name, group_name=self.group_name, metadata=self.metadata
             )
             result_datapoints = await self._evaluate_in_batches(evaluation.id)
 
@@ -409,6 +412,7 @@ def evaluate(
     evaluators: dict[str, EvaluatorFunction | HumanEvaluator],
     name: str | None = None,
     group_name: str | None = None,
+    metadata: dict[str, Any] | None = None,
     concurrency_limit: int = DEFAULT_BATCH_SIZE,
     project_api_key: str | None = None,
     base_url: str | None = None,
@@ -452,6 +456,7 @@ def evaluate(
             Only evaluations within the same group_name can be visually compared.\
             If not provided, set to "default".
             Defaults to None
+        metadata (dict[str, Any] | None, optional): Optional metadata to associate with\
         concurrency_limit (int, optional): The concurrency limit for evaluation.
                         Defaults to DEFAULT_BATCH_SIZE.
         project_api_key (str | None, optional): The project API key.
@@ -478,6 +483,7 @@ def evaluate(
         executor=executor,
         evaluators=evaluators,
         group_name=group_name,
+        metadata=metadata,
         name=name,
         concurrency_limit=concurrency_limit,
         project_api_key=project_api_key,

{lmnr-0.6.12 → lmnr-0.6.14}/src/lmnr/version.py

@@ -3,7 +3,7 @@ import httpx
 from packaging import version
 
 
-__version__ = "0.6.12"
+__version__ = "0.6.14"
 PYTHON_VERSION = f"{sys.version_info.major}.{sys.version_info.minor}"