feat: nightly hardening - robust HTTP response parsing

promptlens/models/config.py

@@ -2,7 +2,7 @@
 
 from typing import Any, Dict, List, Optional
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, ConfigDict, Field
 
 
 class ProviderConfig(BaseModel):
@@ -114,9 +114,7 @@ class RunConfig(BaseModel):
     execution: ExecutionConfig = Field(default_factory=ExecutionConfig)
     output: OutputConfig = Field(default_factory=OutputConfig)
 
-    class Config:
-        """Pydantic config."""
-        json_schema_extra = {
+    model_config = ConfigDict(json_schema_extra={
             "example": {
                 "golden_set": "./examples/golden_sets/customer_support.yaml",
                 "models": [
@@ -142,4 +140,4 @@ class Config:
                     "formats": ["html", "json"],
                 },
             }
-        }
+        })

promptlens/models/test_case.py

@@ -2,7 +2,7 @@
 
 from typing import Any, Dict, List, Optional
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, ConfigDict, Field
 
 from promptlens.models.tools import ToolDefinition, ExpectedToolCall
 
@@ -50,17 +50,15 @@ class TestCase(BaseModel):
         description="Whether to actually execute tools (default: False, evaluation only)"
     )
 
-    class Config:
-        """Pydantic config."""
-        json_schema_extra = {
+    model_config = ConfigDict(json_schema_extra={
             "example": {
                 "id": "cs-001",
                 "query": "How do I reset my password?",
                 "expected_behavior": "Provide clear step-by-step instructions",
                 "category": "account_management",
                 "tags": ["password", "account"],
             }
-        }
+        })
 
 
 class GoldenSet(BaseModel):
@@ -80,9 +78,7 @@ class GoldenSet(BaseModel):
     test_cases: List[TestCase]
     metadata: Dict[str, Any] = Field(default_factory=dict)
 
-    class Config:
-        """Pydantic config."""
-        json_schema_extra = {
+    model_config = ConfigDict(json_schema_extra={
             "example": {
                 "name": "Customer Support Tests",
                 "description": "Test cases for customer support chatbot",
@@ -97,4 +93,4 @@ class Config:
                     }
                 ],
             }
-        }
+        })

promptlens/models/tools.py

@@ -8,7 +8,7 @@
 """
 
 from typing import Any, Dict, List, Optional
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, ConfigDict, Field
 
 
 class ToolParameter(BaseModel):
@@ -24,8 +24,7 @@ class ToolParameter(BaseModel):
     properties: Optional[Dict[str, "ToolParameter"]] = Field(None, description="For object types, nested properties")
     items: Optional["ToolParameter"] = Field(None, description="For array types, the item schema")
 
-    class Config:
-        extra = "allow"  # Allow additional JSON Schema fields
+    model_config = ConfigDict(extra="allow")  # Allow additional JSON Schema fields
 
 
 class ToolDefinition(BaseModel):

promptlens/providers/http.py

@@ -1,5 +1,6 @@
 """Generic HTTP provider for local models (Ollama, LM Studio, etc.)."""
 
+import json
 import logging
 from datetime import datetime
 from typing import Any, Dict, List, Optional
@@ -119,7 +120,19 @@ async def _make_request() -> ModelResponse:
                         timeout=aiohttp.ClientTimeout(total=self.config.timeout),
                     ) as response:
                         response.raise_for_status()
-                        data = await response.json()
+                        raw_text = await response.text()
+
+                try:
+                    data = json.loads(raw_text)
+                except json.JSONDecodeError as exc:
+                    logger.error(
+                        "HTTP provider returned non-JSON response from %s: %s",
+                        self.endpoint,
+                        raw_text[:500],
+                    )
+                    raise ValueError(
+                        "HTTP provider expected JSON response but received non-JSON content"
+                    ) from exc
 
                 # Extract content (try common response formats)
                 content = self._extract_content(data)

tests/test_http_provider_response_parsing.py

@@ -1,3 +1,5 @@
+import pytest
+
 from promptlens.models.config import ProviderConfig
 from promptlens.providers.http import HTTPProvider
 
@@ -57,3 +59,39 @@ def test_extract_content_returns_empty_for_unknown_shape() -> None:
     provider = _provider()
 
     assert provider._extract_content({"foo": "bar"}) == ""
+
+
+@pytest.mark.asyncio
+async def test_http_provider_returns_clear_error_for_non_json_response() -> None:
+    provider = _provider()
+
+    class MockResponse:
+        async def __aenter__(self):
+            return self
+
+        async def __aexit__(self, exc_type, exc, tb):
+            return False
+
+        def raise_for_status(self) -> None:
+            return None
+
+        async def text(self) -> str:
+            return "not-json"
+
+    class MockSession:
+        async def __aenter__(self):
+            return self
+
+        async def __aexit__(self, exc_type, exc, tb):
+            return False
+
+        def post(self, *args, **kwargs):
+            return MockResponse()
+
+    from unittest.mock import patch
+
+    with patch("promptlens.providers.http.aiohttp.ClientSession", return_value=MockSession()):
+        result = await provider.generate("hello")
+
+    assert result.error is not None
+    assert "expected JSON response" in result.error