fix: harden openai-compatible provider responses

src/llm/providers/astraflow.py

@@ -15,6 +15,7 @@ from llm.core.interface import (
     RateLimitError,
 )
 from llm.core.types import LLMInput, LLMOutput, ModelInfo, ProviderType, ToolCall
+from llm.providers.constants import EMPTY_FILTERED_RESPONSE_ERROR
 
 ASTRAFLOW_BASE_URL = "https://api.umodelverse.ai/v1"
 ASTRAFLOW_CN_BASE_URL = "https://api.modelverse.cn/v1"
@@ -55,7 +56,7 @@ class _AstraflowBaseProvider(LLMProvider):
         env_model = os.environ.get(self.model_env)
         fallback_model = os.environ.get(self.fallback_model_env) if self.fallback_model_env else None
         self.default_model = default_model or env_model or fallback_model or DEFAULT_ASTRAFLOW_MODEL
-        self.client = OpenAI(api_key=self.api_key, base_url=self.base_url)
+        self.client = OpenAI(api_key=self.api_key, base_url=self.base_url, _enforce_credentials=False)
         self._models = [
             ModelInfo(
                 name=self.default_model,
@@ -80,7 +81,7 @@ class _AstraflowBaseProvider(LLMProvider):
 
             response = self.client.chat.completions.create(**params)
             if not response.choices or response.choices[0].message is None:
-                raise ValueError("LLM returned empty or filtered response")
+                raise ValueError(EMPTY_FILTERED_RESPONSE_ERROR)
             choice = response.choices[0]
 
             tool_calls = None

src/llm/providers/constants.py

@@ -0,0 +1,3 @@
+"""Shared provider constants."""
+
+EMPTY_FILTERED_RESPONSE_ERROR = "LLM returned empty or filtered response"

src/llm/providers/openai.py

@@ -15,13 +15,18 @@ from llm.core.interface import (
     RateLimitError,
 )
 from llm.core.types import LLMInput, LLMOutput, Message, ModelInfo, ProviderType, ToolCall
+from llm.providers.constants import EMPTY_FILTERED_RESPONSE_ERROR
 
 
 class OpenAIProvider(LLMProvider):
     provider_type = ProviderType.OPENAI
 
     def __init__(self, api_key: str | None = None, base_url: str | None = None) -> None:
-        self.client = OpenAI(api_key=api_key or os.environ.get("OPENAI_API_KEY"), base_url=base_url)
+        self.client = OpenAI(
+            api_key=api_key or os.environ.get("OPENAI_API_KEY"),
+            base_url=base_url,
+            _enforce_credentials=False,
+        )
         self._models = [
             ModelInfo(
                 name="gpt-4o",
@@ -71,7 +76,7 @@ class OpenAIProvider(LLMProvider):
 
             response = self.client.chat.completions.create(**params)
             if not response.choices or response.choices[0].message is None:
-                raise ValueError("LLM returned empty or filtered response")
+                raise ValueError(EMPTY_FILTERED_RESPONSE_ERROR)
             choice = response.choices[0]
 
             tool_calls = None
@@ -85,15 +90,19 @@ class OpenAIProvider(LLMProvider):
                     for tc in choice.message.tool_calls
                 ]
 
-            return LLMOutput(
+            usage = None
-                content=choice.message.content or "",
+            if response.usage:
-                tool_calls=tool_calls,
-                model=response.model,
                 usage = {
                     "prompt_tokens": response.usage.prompt_tokens,
                     "completion_tokens": response.usage.completion_tokens,
                     "total_tokens": response.usage.total_tokens,
-                },
+                }
+
+            return LLMOutput(
+                content=choice.message.content or "",
+                tool_calls=tool_calls,
+                model=response.model,
+                usage=usage,
                 stop_reason=choice.finish_reason,
             )
         except Exception as e:

tests/test_provider_tools.py

@@ -1,7 +1,10 @@
 from types import SimpleNamespace
 
+import pytest
+
 from llm.core.types import LLMInput, Message, Role, ToolDefinition
 from llm.providers.claude import ClaudeProvider
+from llm.providers.constants import EMPTY_FILTERED_RESPONSE_ERROR
 from llm.providers.openai import OpenAIProvider
 
 
@@ -14,21 +17,20 @@ def _tool() -> ToolDefinition:
 
 
 class _OpenAICompletions:
-    def __init__(self) -> None:
+    def __init__(self, response: SimpleNamespace | None = None) -> None:
         self.params = None
+        self.response = response
 
     def create(self, **params):
         self.params = params
-        return SimpleNamespace(
+        if self.response:
-            choices=[SimpleNamespace(message=SimpleNamespace(content="ok", tool_calls=None), finish_reason="stop")],
+            return self.response
-            model=params["model"],
+        return _openai_response(model=params["model"])
-            usage=SimpleNamespace(prompt_tokens=1, completion_tokens=1, total_tokens=2),
-        )
 
 
 class _OpenAIClient:
-    def __init__(self) -> None:
+    def __init__(self, response: SimpleNamespace | None = None) -> None:
-        self.completions = _OpenAICompletions()
+        self.completions = _OpenAICompletions(response=response)
         self.chat = SimpleNamespace(completions=self.completions)
 
 
@@ -52,6 +54,16 @@ class _AnthropicClient:
         self.api_key = "test"
 
 
+def _openai_response(**overrides) -> SimpleNamespace:
+    defaults = {
+        "choices": [SimpleNamespace(message=SimpleNamespace(content="ok", tool_calls=None), finish_reason="stop")],
+        "model": "gpt-4o-mini",
+        "usage": SimpleNamespace(prompt_tokens=1, completion_tokens=1, total_tokens=2),
+    }
+    defaults.update(overrides)
+    return SimpleNamespace(**defaults)
+
+
 def test_openai_provider_serializes_tools_for_chat_completions():
     provider = OpenAIProvider(api_key="test")
     client = _OpenAIClient()
@@ -72,6 +84,36 @@ def test_openai_provider_serializes_tools_for_chat_completions():
     ]
 
 
+def test_openai_provider_can_be_constructed_without_credentials(monkeypatch):
+    monkeypatch.delenv("OPENAI_API_KEY", raising=False)
+
+    provider = OpenAIProvider()
+
+    assert provider.validate_config() is False
+
+
+def test_openai_provider_rejects_empty_or_filtered_responses():
+    provider = OpenAIProvider(api_key="test")
+
+    for response in [
+        _openai_response(choices=[]),
+        _openai_response(choices=[SimpleNamespace(message=None, finish_reason="content_filter")]),
+    ]:
+        provider.client = _OpenAIClient(response=response)
+        with pytest.raises(ValueError, match=EMPTY_FILTERED_RESPONSE_ERROR):
+            provider.generate(LLMInput(messages=[Message(role=Role.USER, content="hi")]))
+
+
+def test_openai_provider_allows_missing_usage():
+    provider = OpenAIProvider(api_key="test")
+    provider.client = _OpenAIClient(response=_openai_response(usage=None))
+
+    output = provider.generate(LLMInput(messages=[Message(role=Role.USER, content="hi")]))
+
+    assert output.content == "ok"
+    assert output.usage is None
+
+
 def test_claude_provider_serializes_tools_for_messages_api():
     provider = ClaudeProvider(api_key="test")
     client = _AnthropicClient()