test against real models

.github/workflows/ci.yml

@@ -56,6 +56,22 @@ jobs:
 
       - run: tree site
 
+  test-live:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: astral-sh/setup-uv@v3
+        with:
+          enable-cache: true
+
+      - run: uv run --python 3.12 --frozen pytest tests/test_live.py
+        if: github.repository_owner == 'pydantic'
+        env:
+          PYDANTIC_AI_LIVE_TEST_DANGEROUS: 'CHARGE-ME!'
+          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+          GEMINI_API_KEY: ${{ secrets.GEMINI_API_KEY }}
+
   test:
     name: test on ${{ matrix.python-version }}
     runs-on: ubuntu-latest
@@ -64,7 +80,7 @@ jobs:
       matrix:
         python-version: ["3.9", "3.10", "3.11", "3.12", "3.13"]
     env:
-      PYTHON: ${{ matrix.python-version }}
+      UV_PYTHON: ${{ matrix.python-version }}
     steps:
       - uses: actions/checkout@v4
 
@@ -73,15 +89,15 @@ jobs:
           enable-cache: true
 
       - run: mkdir coverage
-      - run: uv run --frozen --python ${{ matrix.python-version }} coverage run -m pytest
+      - run: uv run --frozen coverage run -m pytest
         env:
           COVERAGE_FILE: coverage/.coverage.${{ runner.os }}-py${{ matrix.python-version }}
 
-      - run: uv run --frozen --all-extras --python ${{ matrix.python-version }} coverage run -m pytest
+      - run: uv run --frozen --all-extras coverage run -m pytest
         env:
           COVERAGE_FILE: coverage/.coverage.${{ runner.os }}-py${{ matrix.python-version }}-all-extras
 
-      - run: uv run --frozen --all-extras --python ${{ matrix.python-version }} python tests/import_examples.py
+      - run: uv run --frozen --all-extras python tests/import_examples.py
 
       - name: store coverage files
         uses: actions/upload-artifact@v4
@@ -117,7 +133,7 @@ jobs:
   # https://github.com/marketplace/actions/alls-green#why used for branch protection checks
   check:
     if: always()
-    needs: [lint, docs, test, coverage]
+    needs: [lint, docs, test-live, test, coverage]
     runs-on: ubuntu-latest
 
     steps:

pydantic_ai/models/gemini.py

@@ -187,6 +187,7 @@ async def _make_request(self, messages: list[Message], streamed: bool) -> AsyncI
 
         async with self.http_client.stream('POST', url, content=request_json, headers=headers) as r:
             if r.status_code != 200:
+                await r.aread()
                 raise exceptions.UnexpectedModelBehaviour(f'Unexpected response from gemini {r.status_code}', r.text)
             yield r
 

pyproject.toml

@@ -146,6 +146,7 @@ filterwarnings = [
 [tool.coverage.run]
 # required to avoid warnings about files created by create_module fixture
 include = ["pydantic_ai/**/*.py", "tests/**/*.py"]
+omit = ["tests/test_live.py"]
 branch = true
 
 # https://coverage.readthedocs.io/en/latest/config.html#report

tests/test_live.py

@@ -0,0 +1,66 @@
+"""Tests of pydantic-ai actually connecting to OpenAI and Gemini models.
+
+WARNING: running these tests will consume your OpenAI and Gemini credits.
+"""
+
+import os
+
+import httpx
+import pytest
+
+from pydantic_ai import Agent
+from pydantic_ai.models.gemini import GeminiModel
+from pydantic_ai.models.openai import OpenAIModel
+
+pytestmark = [
+    pytest.mark.skipif(os.getenv('PYDANTIC_AI_LIVE_TEST_DANGEROUS') != 'CHARGE-ME!', reason='live tests disabled'),
+    pytest.mark.anyio,
+]
+
+
+@pytest.fixture
+async def http_client():
+    async with httpx.AsyncClient(timeout=30) as client:
+        yield client
+
+
+async def test_openai(http_client: httpx.AsyncClient):
+    agent = Agent(OpenAIModel('gpt-3.5-turbo', http_client=http_client))
+    result = await agent.run('What is the capital of France?')
+    print('OpenAI response:', result.data)
+    assert 'paris' in result.data.lower()
+    print('OpenAI cost:', result.cost())
+    cost = result.cost()
+    assert cost.total_tokens is not None and cost.total_tokens > 0
+
+
+async def test_openai_stream(http_client: httpx.AsyncClient):
+    agent = Agent(OpenAIModel('gpt-3.5-turbo', http_client=http_client))
+    async with agent.run_stream('What is the capital of France?') as result:
+        data = await result.get_data()
+    print('OpenAI stream response:', data)
+    assert 'paris' in data.lower()
+    print('OpenAI stream cost:', result.cost())
+    cost = result.cost()
+    assert cost.total_tokens is not None and cost.total_tokens > 0
+
+
+async def test_gemini(http_client: httpx.AsyncClient):
+    agent = Agent(GeminiModel('gemini-1.5-flash', http_client=http_client))
+    result = await agent.run('What is the capital of France?')
+    print('Gemini response:', result.data)
+    assert 'paris' in result.data.lower()
+    print('Gemini cost:', result.cost())
+    cost = result.cost()
+    assert cost.total_tokens is not None and cost.total_tokens > 0
+
+
+async def test_gemini_stream(http_client: httpx.AsyncClient):
+    agent = Agent(GeminiModel('gemini-1.5-flash', http_client=http_client))
+    async with agent.run_stream('What is the capital of France?') as result:
+        data = await result.get_data()
+    print('Gemini stream response:', data)
+    assert 'paris' in data.lower()
+    print('Gemini stream cost:', result.cost())
+    cost = result.cost()
+    assert cost.total_tokens is not None and cost.total_tokens > 0