langwatch 0.8.0__py3-none-any.whl → 0.9.0__py3-none-any.whl

langwatch/__version__.py

@@ -1,3 +1,3 @@
 """Version information for LangWatch."""
 
-__version__ = "0.8.0" # x-release-please-version
+__version__ = "0.9.0" # x-release-please-version

langwatch/batch_evaluation.py

@@ -24,6 +24,7 @@ from tqdm import tqdm
 import pandas as pd
 
 from langwatch.types import Money
+from langwatch.utils.exceptions import better_raise_for_status
 
 
 class EvaluationResult(BaseModel):
@@ -150,7 +151,7 @@ class BatchEvaluation:
             raise ValueError(
                 "API key is not valid, please try to login again with langwatch.login()"
             )
-        response.raise_for_status()
+        better_raise_for_status(response)
         experiment_path = response.json()["path"]
         self.experiment_slug = response.json()["slug"]
 
@@ -368,7 +369,7 @@ class BatchEvaluation:
             json=body,
             timeout=60,
         )
-        response.raise_for_status()
+        better_raise_for_status(response)
 
     def wait_for_completion(self):
         async def wait_for_completion(self):
@@ -414,7 +415,7 @@ async def run_evaluation(
 
         async with httpx.AsyncClient(timeout=900) as client:
             response = await client.post(**request_params)
-            response.raise_for_status()
+            better_raise_for_status(response)
 
         result = response.json()
 
@@ -462,7 +463,7 @@ def get_dataset(
 
     with httpx.Client(timeout=300) as client:
         response = client.get(**request_params)
-        response.raise_for_status()
+        better_raise_for_status(response)
 
     result = response.json()
 

langwatch/dspy/__init__.py

@@ -5,6 +5,7 @@ import time
 import warnings
 import dspy
 from typing import Callable, List, Optional, Any, Type, Union
+from langwatch.utils.exceptions import better_raise_for_status
 from langwatch.utils.transformation import truncate_object_recursively
 from langwatch.telemetry.tracing import LangWatchTrace
 from typing_extensions import TypedDict
@@ -193,7 +194,7 @@ class LangWatchDSPy:
             raise ValueError(
                 "API key is not valid, please try to login again with langwatch.login()"
             )
-        response.raise_for_status()
+        better_raise_for_status(response)
 
         if optimizer and evaluator:
             raise ValueError("You can only provide an optimizer or an evaluator, not both.")
@@ -386,7 +387,7 @@ class LangWatchDSPy:
             data=json.dumps(data),  # type: ignore
             timeout=60,
         )
-        response.raise_for_status()
+        better_raise_for_status(response)
         self.steps_buffer = []
 
     def tracer(self, trace: LangWatchTrace):
@@ -736,10 +737,6 @@ class DSPyTracer:
             dspy.Module.__original_call__ = dspy.Module.__call__  # type: ignore
             dspy.Module.__call__ = self.patched_module_call()
 
-        if not hasattr(dspy.Predict, "__original_forward__"):
-            dspy.Predict.__original_forward__ = dspy.Predict.forward  # type: ignore
-            dspy.Predict.forward = self.patched_predict_forward()
-
         language_model_classes = dspy.LM.__subclasses__()
         for lm in language_model_classes:
             if not hasattr(lm, "__original_basic_request__") and hasattr(
@@ -775,7 +772,7 @@ class DSPyTracer:
     def patched_module_call(self):
         self_ = self
 
-        @langwatch.span(ignore_missing_trace_warning=True, type="module")
+        @langwatch.span(ignore_missing_trace_warning=True, type="module", capture_output=False)
         def __call__(self: dspy.Module, *args, **kwargs):
             span = self_.safe_get_current_span()
             signature = (
@@ -800,34 +797,10 @@ class DSPyTracer:
 
         return __call__
 
-    def patched_predict_forward(self):
-        self_ = self
-
-        @langwatch.span(ignore_missing_trace_warning=True, type="module")
-        def forward(self: dspy.Predict, **kwargs):
-            span = self_.safe_get_current_span()
-            signature = kwargs.get("signature", self.signature)
-
-            if span and signature and hasattr(signature, "__name__"):
-                span.update(name=f"{self.__class__.__name__}({signature.__name__})")
-            elif span:
-                span.update(name=f"{self.__class__.__name__}.forward")
-
-            prediction = self.__class__.__original_forward__(self, **kwargs)  # type: ignore
-
-            if span and isinstance(prediction, dspy.Prediction):
-                span.update(output=prediction._store)  # type: ignore
-            elif span:
-                span.update(output=prediction)  # type: ignore
-
-            return prediction
-
-        return forward
-
     def patched_language_model_call(self):
         self_ = self
 
-        @langwatch.span(ignore_missing_trace_warning=True, type="llm")
+        @langwatch.span(ignore_missing_trace_warning=True, type="llm", capture_output=False)
         def call(self: dspy.LM, prompt=None, messages=None, **kwargs):
             all_kwargs = self.kwargs | kwargs
             model = self.model
@@ -894,7 +867,7 @@ class DSPyTracer:
     def patched_legacy_language_model_request(self):
         self_ = self
 
-        @langwatch.span(ignore_missing_trace_warning=True, type="llm")
+        @langwatch.span(ignore_missing_trace_warning=True, type="llm", capture_output=False)
         def basic_request(self: dspy.LM, prompt, **kwargs):
             all_kwargs = self.kwargs | kwargs
             model = all_kwargs.get("model", None)
@@ -946,7 +919,7 @@ class DSPyTracer:
             ) is not getattr(dspy.Retrieve, "forward", None):
                 return self.__class__.__original_forward__(self, *args, **kwargs)  # type: ignore
 
-            @langwatch.span(ignore_missing_trace_warning=True, type="rag")
+            @langwatch.span(ignore_missing_trace_warning=True, type="rag", capture_output=False)
             def forward(self, *args, **kwargs):
                 result = self.__class__.__original_forward__(self, *args, **kwargs)  # type: ignore
 

langwatch/evaluation/__init__.py

@@ -1,9 +1,36 @@
 from typing import Optional
 from langwatch.evaluation.evaluation import Evaluation
-from .evaluation import Evaluation
+from langwatch.evaluation.platform_run import (
+    evaluate,
+    run,  # Deprecated, kept for backwards compatibility
+    EvaluationRunResult,
+    EvaluationRunSummary,
+    EvaluationNotFoundError,
+    EvaluationTimeoutError,
+    EvaluationRunFailedError,
+    EvaluationsApiError,
+    TargetStats,
+    EvaluatorStats,
+)
 
 
 def init(name: str, *, run_id: Optional[str] = None) -> Evaluation:
     evaluation = Evaluation(name, run_id=run_id)
     evaluation.init()
     return evaluation
+
+
+__all__ = [
+    "init",
+    "evaluate",
+    "run",  # Deprecated
+    "Evaluation",
+    "EvaluationRunResult",
+    "EvaluationRunSummary",
+    "EvaluationNotFoundError",
+    "EvaluationTimeoutError",
+    "EvaluationRunFailedError",
+    "EvaluationsApiError",
+    "TargetStats",
+    "EvaluatorStats",
+]