langwatch 0.9.0__tar.gz → 0.10.0__tar.gz

{langwatch-0.9.0 → langwatch-0.10.0}/CHANGELOG.md

@@ -1,5 +1,24 @@
 # Changelog
 
+## [0.10.0](https://github.com/langwatch/langwatch/compare/python-sdk@v0.9.0...python-sdk@v0.10.0) (2026-01-21)
+
+
+### Features
+
+* **sdk:** add online evaluations API and ensureSetup for TypeScript ([2209258](https://github.com/langwatch/langwatch/commit/22092580433b9a3014123e62f39cf8c44543d8cc))
+
+
+### Bug Fixes
+
+* **python-sdk:** resolve name collision between Evaluation TypedDict and class alias ([873909a](https://github.com/langwatch/langwatch/commit/873909a09414bb309fa955b68e4a19bddc93e21f))
+
+
+### Code Refactoring
+
+* **sdk:** fix evaluation API naming consistency ([3aef2a3](https://github.com/langwatch/langwatch/commit/3aef2a353cc4939c8087b53044f441e7501530ae))
+* **sdk:** rename evaluation API to experiment for new terminology ([f10326c](https://github.com/langwatch/langwatch/commit/f10326c0f2ee5818fbcf51507a95933bfe83caeb))
+* **sdk:** rename internal evaluation classes to experiment ([ff70cab](https://github.com/langwatch/langwatch/commit/ff70cab904905b85a77a68e7b1eb60e9c364a18d))
+
 ## [0.9.0](https://github.com/langwatch/langwatch/compare/python-sdk@v0.8.1...python-sdk@v0.9.0) (2026-01-18)
 
 

{langwatch-0.9.0 → langwatch-0.10.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: langwatch
-Version: 0.9.0
+Version: 0.10.0
 Summary: LangWatch Python SDK, for monitoring your LLMs
 Author-email: Langwatch Engineers <engineering@langwatch.ai>
 License: MIT

{langwatch-0.9.0 → langwatch-0.10.0}/examples/guardrails.py

@@ -8,7 +8,7 @@ from openai import OpenAI
 client = OpenAI()
 
 import langwatch
-import langwatch.guardrails
+import langwatch.evaluation
 
 
 @cl.on_message
@@ -23,14 +23,15 @@ async def main(message: cl.Message):
         content="",
     )
 
-    jailbreak_guardrail = langwatch.get_current_span().evaluate(
+    # New API: Use langwatch.evaluation.evaluate() instead of span.evaluate()
+    jailbreak_guardrail = langwatch.evaluation.evaluate(
         "jailbreak-detection",
+        data={"input": message.content},
         name="Jailbreak Detection",
         as_guardrail=True,
-        input=message.content,
     )
     if not jailbreak_guardrail.passed:
-        await msg.stream_token(f"I'm sorry, I can't help you with that.")
+        await msg.stream_token("I'm sorry, I can't help you with that.")
         await msg.update()
         return
 

{langwatch-0.9.0 → langwatch-0.10.0}/examples/guardrails_parallel.py

@@ -8,8 +8,8 @@ from openai import AsyncOpenAI
 
 client = AsyncOpenAI()
 
-import langwatch.openai
-import langwatch.guardrails
+import langwatch
+import langwatch.evaluation
 
 
 @cl.on_message
@@ -24,14 +24,19 @@ async def main(message: cl.Message):
         content="",
     )
 
+    # New API: Use langwatch.evaluation.async_evaluate() for parallel guardrails
     jailbreak_guardrail = asyncio.create_task(
-        langwatch.get_current_span().async_evaluate(
-            "jailbreak-detection", as_guardrail=True, input=message.content
+        langwatch.evaluation.async_evaluate(
+            "jailbreak-detection",
+            data={"input": message.content},
+            as_guardrail=True,
         )
     )
     off_topic_guardrail = asyncio.create_task(
-        langwatch.get_current_span().async_evaluate(
-            "off-topic-evaluator", as_guardrail=True, input=message.content
+        langwatch.evaluation.async_evaluate(
+            "off-topic-evaluator",
+            data={"input": message.content},
+            as_guardrail=True,
         )
     )
 
@@ -39,14 +44,14 @@ async def main(message: cl.Message):
         if jailbreak_guardrail.done():
             result = await jailbreak_guardrail
             if not result.passed:
-                await msg.stream_token(f"I'm sorry, I can't help you with that.")
+                await msg.stream_token("I'm sorry, I can't help you with that.")
                 await msg.update()
                 return True
 
         if off_topic_guardrail.done():
             result = await off_topic_guardrail
             if not result.passed:
-                await msg.stream_token(f"I'm sorry, I can't help you with that.")
+                await msg.stream_token("I'm sorry, I can't help you with that.")
                 await msg.update()
                 return True
 

{langwatch-0.9.0 → langwatch-0.10.0}/examples/guardrails_without_tracing.py

@@ -7,7 +7,7 @@ from openai import OpenAI
 
 client = OpenAI()
 
-import langwatch.guardrails
+import langwatch.evaluation
 
 
 @cl.on_message
@@ -16,11 +16,14 @@ async def main(message: cl.Message):
         content="",
     )
 
-    jailbreak_guardrail = langwatch.guardrails.evaluate(
-        "jailbreak-detection", input=message.content
+    # New API: Use langwatch.evaluation.evaluate() with as_guardrail=True
+    jailbreak_guardrail = langwatch.evaluation.evaluate(
+        "jailbreak-detection",
+        data={"input": message.content},
+        as_guardrail=True,
     )
     if not jailbreak_guardrail.passed:
-        await msg.stream_token(f"I'm sorry, I can't help you with that.")
+        await msg.stream_token("I'm sorry, I can't help you with that.")
         await msg.update()
         return
 

{langwatch-0.9.0 → langwatch-0.10.0}/examples/langwatch_docs_rag_evaluation.ipynb

@@ -244,7 +244,7 @@
         "# Or retrieve it from LangWatch:\n",
         "# df = langwatch.dataset.get_dataset(\"CEtFivQeya4kyAzy9eJht\").to_pandas()  # dataset--rSAYL4HxQRXHSayq6c7A\n",
         "\n",
-        "evaluation = langwatch.evaluation.init(\"my-incredible-experiment\")\n",
+        "evaluation = langwatch.experiment.init(\"my-incredible-experiment\")\n",
         "\n",
         "for index, row in evaluation.loop(df.iterrows()):\n",
         "    def task(index, row):\n",

{langwatch-0.9.0 → langwatch-0.10.0}/examples/offline_evaluation.ipynb

@@ -89,7 +89,7 @@
         "    ]\n",
         ")\n",
         "\n",
-        "evaluation = langwatch.evaluation.init(\"my-incredible-experiment\")\n",
+        "evaluation = langwatch.experiment.init(\"my-incredible-experiment\")\n",
         "\n",
         "\n",
         "@langwatch.trace()\n",
@@ -141,7 +141,7 @@
         "import time\n",
         "\n",
         "langwatch.setup()\n",
-        "evaluation = langwatch.evaluation.init(\"my-incredible-experiment\")\n",
+        "evaluation = langwatch.experiment.init(\"my-incredible-experiment\")\n",
         "\n",
         "@langwatch.trace()\n",
         "def agent(question):\n",
@@ -194,7 +194,7 @@
         "import time\n",
         "\n",
         "langwatch.setup()\n",
-        "evaluation = langwatch.evaluation.init(\"model-comparison-experiment\")\n",
+        "evaluation = langwatch.experiment.init(\"model-comparison-experiment\")\n",
         "\n",
         "\n",
         "@langwatch.trace()\n",

{langwatch-0.9.0 → langwatch-0.10.0}/examples/pdf_parsing_evaluation.ipynb

@@ -396,7 +396,7 @@
         }
       ],
       "source": [
-        "evaluation = langwatch.evaluation.init(\"pdf-parsing-evaluation\")\n",
+        "evaluation = langwatch.experiment.init(\"pdf-parsing-evaluation\")\n",
         "\n",
         "for index, row in evaluation.loop(df.iterrows()):\n",
         "    def task(index, row):\n",

{langwatch-0.9.0 → langwatch-0.10.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "langwatch"
-version = "0.9.0" # remember to also update it in src/langwatch/__version__.py
+version = "0.10.0" # remember to also update it in src/langwatch/__version__.py
 description = "LangWatch Python SDK, for monitoring your LLMs"
 authors = [{ name = "Langwatch Engineers", email = "engineering@langwatch.ai" }]
 requires-python = ">=3.10,<3.14"

{langwatch-0.9.0 → langwatch-0.10.0}/src/langwatch/__init__.py

@@ -17,7 +17,8 @@ from typing import TYPE_CHECKING
 # Type hints for IntelliSense (only imported for typing)
 if TYPE_CHECKING:
     import langwatch.evaluations as evaluations
-    import langwatch.evaluation as evaluation
+    import langwatch.experiment as experiment
+    import langwatch.evaluation as evaluation  # Deprecated, use experiment
     import langwatch.dataset as dataset
     import langwatch.dspy as dspy
     import langwatch.langchain as langchain
@@ -41,7 +42,8 @@ def _api_key():
 # Lazy loading configuration
 _LAZY_MODULES = {
     "evaluations": "langwatch.evaluations",
-    "evaluation": "langwatch.evaluation",
+    "experiment": "langwatch.experiment",
+    "evaluation": "langwatch.evaluation",  # Deprecated, use experiment
     "dataset": "langwatch.dataset",
     "dspy": "langwatch.dspy",  # Special handling
     "langchain": "langwatch.langchain",  # Special handling
@@ -150,7 +152,8 @@ __all__ = [
     "ensure_setup",
     "get_current_trace",
     "get_current_span",
-    "evaluation",
+    "experiment",
+    "evaluation",  # Deprecated, use experiment
     "dataset",
     "evaluations",
     "langchain",

langwatch-0.10.0/src/langwatch/__version__.py

@@ -0,0 +1,3 @@
+"""Version information for LangWatch."""
+
+__version__ = "0.10.0" # x-release-please-version