vllm-judge 0.1.2__py3-none-any.whl → 0.1.3__py3-none-any.whl

vllm_judge/__init__.py

@@ -5,7 +5,7 @@ A lightweight library for evaluating text responses using self-hosted language m
 via vLLM's OpenAI-compatible API.
 """
 
-__version__ = "0.1.2"
+__version__ = "0.1.3"
 
 from vllm_judge.judge import Judge
 from vllm_judge.models import (

vllm_judge/api/client.py

@@ -65,7 +65,7 @@ class JudgeClient:
     
     async def evaluate(
         self,
-        response: Union[str, Dict[str, str]],
+        content: Union[str, Dict[str, str]],
         criteria: str = None,
         rubric: Union[str, Dict[Union[int, float], str]] = None,
         scale: Optional[Tuple[int, int]] = None,
@@ -87,7 +87,7 @@ class JudgeClient:
             EvaluationResult
         """
         request = EvaluateRequest(
-            response=response,
+            response=content,
             criteria=criteria,
             rubric=rubric,
             scale=list(scale) if scale else None,

vllm_judge/cli.py

@@ -75,7 +75,7 @@ def evaluate(
             # Use API client
             async with JudgeClient(api_url) as client:
                 result = await client.evaluate(
-                    response=response,
+                    content=response,
                     criteria=criteria,
                     metric=metric,
                     scale=scale,
@@ -91,7 +91,7 @@ def evaluate(
             judge = Judge.from_url(base_url, model=model)
             async with judge:
                 result = await judge.evaluate(
-                    response=response,
+                    content=response,
                     criteria=criteria,
                     metric=metric,
                     scale=scale,

vllm_judge/judge.py

@@ -63,7 +63,7 @@ class Judge:
     
     async def evaluate(
         self,
-        response: Union[str, Dict[str, str]],
+        content: Union[str, Dict[str, str]],
         criteria: str = None,
         rubric: Union[str, Dict[Union[int, float], str]] = None,
         scale: Optional[Tuple[int, int]] = None,
@@ -79,7 +79,7 @@ class Judge:
         Universal evaluation method that adapts to use case.
         
         Args:
-            response: String for single evaluation, dict {"a": ..., "b": ...} for comparison
+            content: String for single evaluation, dict {"a": ..., "b": ...} for comparison
             criteria: What to evaluate for (can contain template variables)
             rubric: Instructions for evaluation, can be string or dict containing mapping of score to description (can contain template variables)
             scale: Optional numeric scale (min, max)
@@ -101,12 +101,12 @@ class Judge:
         """
         # Handle model-specific metrics
         if isinstance(metric, ModelSpecificMetric):
-            assert isinstance(response, str), "Model-specific metrics only support string content for now"
+            assert isinstance(content, str), "Model-specific metrics only support string content for now"
 
             # logger.info(f"Evaluating model-specific metric {metric.name}.")
             logger.info(f"We assume you're using {metric.model_pattern} type model. If not, please do not use this metric and use a normal metric instead.")
             # Skip ALL our formatting
-            messages = [{"role": "user", "content": response}]
+            messages = [{"role": "user", "content": content}]
             
             # vLLM applies model's chat template automatically
             llm_response = await self._call_model(messages)
@@ -157,7 +157,7 @@ class Judge:
         
         # Build messages
         messages = PromptBuilder.build_messages(
-            response=response,
+            response=content,
             criteria=criteria,
             rubric=rubric,
             scale=scale,

{vllm_judge-0.1.2.dist-info → vllm_judge-0.1.3.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: vllm_judge
-Version: 0.1.2
+Version: 0.1.3
 Summary: LLM-as-a-Judge evaluations for vLLM hosted models
 Author: TrustyAI team
 Author-email: Sai Chandra Pandraju <saichandrapandraju@gmail.com>
@@ -49,9 +49,8 @@ A lightweight library for LLM-as-a-Judge evaluations using vLLM hosted models. E
 - 🚀 **Simple Interface**: Single `evaluate()` method that adapts to any use case
 - 🎯 **Pre-built Metrics**: 20+ ready-to-use evaluation metrics
 - 🛡️ **Model-Specific Support:** Seamlessly works with specialized models like Llama Guard without breaking their trained formats.
-- 🔄 **Async Native**: Built for high-throughput evaluations
+- ⚡ **High Performance**: Async-first design enables high-throughput evaluations
 - 🔧 **Template Support**: Dynamic evaluations with template variables
-- ⚡ **High Performance**: Optimized for vLLM with automatic batching
 - 🌐 **API Mode**: Run as a REST API service
 
 ## Installation
@@ -80,7 +79,7 @@ judge = Judge.from_url("http://vllm-server:8000")
 
 # Simple evaluation
 result = await judge.evaluate(
-    response="The Earth orbits around the Sun.",
+    content="The Earth orbits around the Sun.",
     criteria="scientific accuracy"
 )
 print(f"Decision: {result.decision}")
@@ -90,13 +89,13 @@ print(f"Reasoning: {result.reasoning}")
 from vllm_judge import CODE_QUALITY
 
 result = await judge.evaluate(
-    response="def add(a, b): return a + b",
+    content="def add(a, b): return a + b",
     metric=CODE_QUALITY
 )
 
 # With template variables
 result = await judge.evaluate(
-    response="Essay content here...",
+    content="Essay content here...",
     criteria="Evaluate this {doc_type} for {audience}",
     template_vars={
         "doc_type": "essay",
@@ -108,7 +107,7 @@ result = await judge.evaluate(
 from vllm_judge import LLAMA_GUARD_3_SAFETY
 
 result = await judge.evaluate(
-    response="How do I make a bomb?",
+    content="How do I make a bomb?",
     metric=LLAMA_GUARD_3_SAFETY  # Automatically uses Llama Guard format
 )
 # Result: decision="unsafe", reasoning="S9"
@@ -129,7 +128,7 @@ from vllm_judge.api import JudgeClient
 
 client = JudgeClient("http://localhost:9090")
 result = await client.evaluate(
-    response="Python is great!",
+    content="Python is great!",
     criteria="technical accuracy"
 )
 ```

{vllm_judge-0.1.2.dist-info → vllm_judge-0.1.3.dist-info}/RECORD

@@ -1,20 +1,20 @@
-vllm_judge/__init__.py,sha256=TcPeBC1yv3oDT5c8NvikyOL9cZyDZRnHD2Aeu0ynGuo,2469
+vllm_judge/__init__.py,sha256=TBS7fQ4n7QEVwNtr4ErJu-T3m4c-8BwW4zDltt8S6Ko,2469
 vllm_judge/batch.py,sha256=68jKgRTMzZXw4bxAiGp73NZzHOd1tKK763nBNjrr6gg,4842
-vllm_judge/cli.py,sha256=KQtUt_L4u5TPrS8xoyiKYt_hQ_FiHtGcrkecGEtktI8,10685
+vllm_judge/cli.py,sha256=mdoxNA5gQ1m3XBnNJYCE8uoi0RxrS9d3YIlrtdxRcME,10683
 vllm_judge/client.py,sha256=QPz64q9-7XEOOJiKQU7FBkGFWocJ-WGUmpETKSLQYDI,8386
 vllm_judge/exceptions.py,sha256=X9YxnukDuI3RwJPkabj3pl6v0JIbflvhUaWrdAW4RTM,1066
-vllm_judge/judge.py,sha256=Wn1ez1HJKb2U0Fu-kcIo7Ls3-ph7hVtb6K5Rlk0NfGw,15225
+vllm_judge/judge.py,sha256=FKMpl6ubugHqKlR-W1-arr4J2rkwnC76QM5oAFv_HyM,15220
 vllm_judge/metrics.py,sha256=lQOBaHqlX79L8yP9_YYd-dTaqvfOPo0nDMY0dtsnKvI,15960
 vllm_judge/models.py,sha256=aEXZmP2sM-9aetstzHE3ngZwvCcvnrqzcj-8oV0NCJA,7889
 vllm_judge/prompts.py,sha256=jAsBdshCCdgGF3UUAM0Wbb6MN1AB2jgHh1NmtXLbyrc,6345
 vllm_judge/templating.py,sha256=LjVFXFcwHl8xnBLLVr_IIqtN-EbLp0HZ5ndNbBpcJTQ,6998
 vllm_judge/utils.py,sha256=lhByBIMS_1EwvxEe31jFgVcTwcFwm5mWoJDXG4TnbvQ,509
 vllm_judge/api/__init__.py,sha256=aPQ1o7_ZzbJJpm2UyX3H35snbOGbgQJoglJjzdnc1LU,762
-vllm_judge/api/client.py,sha256=mcpdH-9ko6aEh_JAybpPPVhHqlO3l5K-lTujTlkTw8c,11302
+vllm_judge/api/client.py,sha256=XRiveUw1edcknxO3zLFkYX_YbOObipx7dMFeSUjMSwk,11300
 vllm_judge/api/models.py,sha256=tPEePecZbKb9ZbjwusdJwhLiBK9Rd5xqiOqjklDKJ9s,4781
 vllm_judge/api/server.py,sha256=mbQ45YC0RYGONdy1oIcRIxUvByLtKXXrrMTpE9l2y1w,17818
-vllm_judge-0.1.2.dist-info/METADATA,sha256=DtXmkJ_sIXp49PuIL3CZJzkPRHsR8zhhwMaJFm6bUYg,4307
-vllm_judge-0.1.2.dist-info/WHEEL,sha256=iAkIy5fosb7FzIOwONchHf19Qu7_1wCWyFNR5gu9nU0,91
-vllm_judge-0.1.2.dist-info/entry_points.txt,sha256=F3plmbMXOQ0pBIh0clqWPVIJWl20_1LZ7QHxC2XF5Lg,51
-vllm_judge-0.1.2.dist-info/top_level.txt,sha256=bqtMvn2y13cHSz_1-HKCBMzYSTfDHsTQBG6U5STHvwM,11
-vllm_judge-0.1.2.dist-info/RECORD,,
+vllm_judge-0.1.3.dist-info/METADATA,sha256=L_Kf2ic1W5wn1D1Y4amZaxO6E2i6bEKjZ4JFVvh3-YA,4251
+vllm_judge-0.1.3.dist-info/WHEEL,sha256=iAkIy5fosb7FzIOwONchHf19Qu7_1wCWyFNR5gu9nU0,91
+vllm_judge-0.1.3.dist-info/entry_points.txt,sha256=F3plmbMXOQ0pBIh0clqWPVIJWl20_1LZ7QHxC2XF5Lg,51
+vllm_judge-0.1.3.dist-info/top_level.txt,sha256=bqtMvn2y13cHSz_1-HKCBMzYSTfDHsTQBG6U5STHvwM,11
+vllm_judge-0.1.3.dist-info/RECORD,,