pydantic-ai 0.0.32__tar.gz → 0.0.33__tar.gz

{pydantic_ai-0.0.32 → pydantic_ai-0.0.33}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pydantic-ai
-Version: 0.0.32
+Version: 0.0.33
 Summary: Agent Framework / shim to use Pydantic with LLMs
 Project-URL: Homepage, https://ai.pydantic.dev
 Project-URL: Source, https://github.com/pydantic/pydantic-ai
@@ -28,9 +28,9 @@ Classifier: Topic :: Internet
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Classifier: Topic :: Software Development :: Libraries :: Python Modules
 Requires-Python: >=3.9
-Requires-Dist: pydantic-ai-slim[anthropic,cohere,groq,mistral,openai,vertexai]==0.0.32
+Requires-Dist: pydantic-ai-slim[anthropic,cohere,groq,mistral,openai,vertexai]==0.0.33
 Provides-Extra: examples
-Requires-Dist: pydantic-ai-examples==0.0.32; extra == 'examples'
+Requires-Dist: pydantic-ai-examples==0.0.33; extra == 'examples'
 Provides-Extra: logfire
 Requires-Dist: logfire>=2.3; extra == 'logfire'
 Description-Content-Type: text/markdown

{pydantic_ai-0.0.32 → pydantic_ai-0.0.33}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "pydantic-ai"
-version = "0.0.32"
+version = "0.0.33"
 description = "Agent Framework / shim to use Pydantic with LLMs"
 authors = [
     { name = "Samuel Colvin", email = "samuel@pydantic.dev" },
@@ -37,7 +37,7 @@ classifiers = [
 requires-python = ">=3.9"
 
 dependencies = [
-    "pydantic-ai-slim[openai,vertexai,groq,anthropic,mistral,cohere]==0.0.32",
+    "pydantic-ai-slim[openai,vertexai,groq,anthropic,mistral,cohere]==0.0.33",
 ]
 
 [project.urls]
@@ -47,7 +47,7 @@ Documentation = "https://ai.pydantic.dev"
 Changelog = "https://github.com/pydantic/pydantic-ai/releases"
 
 [project.optional-dependencies]
-examples = ["pydantic-ai-examples==0.0.32"]
+examples = ["pydantic-ai-examples==0.0.33"]
 logfire = ["logfire>=2.3"]
 
 [tool.uv.sources]

{pydantic_ai-0.0.32 → pydantic_ai-0.0.33}/tests/conftest.py

@@ -200,6 +200,7 @@ def pytest_recording_configure(config: Any, vcr: VCR):
 @pytest.fixture(scope='module')
 def vcr_config():
     return {
+        'ignore_localhost': True,
         # Note: additional header filtering is done inside the serializer
         'filter_headers': ['authorization', 'x-api-key'],
         'decode_compressed_response': True,

{pydantic_ai-0.0.32 → pydantic_ai-0.0.33}/tests/json_body_serializer.py

@@ -1,5 +1,6 @@
 # pyright: reportUnknownMemberType=false, reportUnknownVariableType=false
 import json
+import urllib.parse
 from typing import TYPE_CHECKING, Any
 
 import yaml
@@ -59,19 +60,24 @@ def serialize(cassette_dict: Any):
             # update headers on source object
             data['headers'] = headers
 
-            content_type = headers.get('content-type', None)
-            if content_type != ['application/json']:
-                continue
-
-            # Parse the body as JSON
-            body: Any = data.get('body', None)
-            assert body is not None, data
-            if isinstance(body, dict):
-                # Responses will have the body under a field called 'string'
-                body = body.get('string')
-            if body is not None:
-                data['parsed_body'] = json.loads(body)
-                del data['body']
+            content_type = headers.get('content-type', [])
+            if any(header.startswith('application/json') for header in content_type):
+                # Parse the body as JSON
+                body: Any = data.get('body', None)
+                assert body is not None, data
+                if isinstance(body, dict):
+                    # Responses will have the body under a field called 'string'
+                    body = body.get('string')
+                if body is not None:
+                    data['parsed_body'] = json.loads(body)
+                    if 'access_token' in data['parsed_body']:
+                        data['parsed_body']['access_token'] = 'scrubbed'
+                    del data['body']
+            if content_type == ['application/x-www-form-urlencoded']:
+                query_params = urllib.parse.parse_qs(data['body'])
+                if 'client_secret' in query_params:
+                    query_params['client_secret'] = ['scrubbed']
+                    data['body'] = urllib.parse.urlencode(query_params)
 
     # Use our custom dumper
     return yaml.dump(cassette_dict, Dumper=LiteralDumper, allow_unicode=True, width=120)

{pydantic_ai-0.0.32 → pydantic_ai-0.0.33}/tests/models/cassettes/test_gemini/test_image_as_binary_content_input.yaml

@@ -38,7 +38,7 @@ interactions:
                 "role": "model"
               },
               "finishReason": "STOP",
-              "avgLogprobs": -0.02894418438275655
+              "avgLogprobs": -0.031536102294921875
             }
           ],
           "usageMetadata": {
@@ -68,11 +68,11 @@ interactions:
       alt-svc:
       - h3=":443"; ma=2592000,h3-29=":443"; ma=2592000
       content-length:
-      - '710'
+      - '711'
       content-type:
       - application/json; charset=UTF-8
       server-timing:
-      - gfet4t7; dur=2776
+      - gfet4t7; dur=2657
       transfer-encoding:
       - chunked
       vary:

{pydantic_ai-0.0.32 → pydantic_ai-0.0.33}/tests/models/cassettes/test_gemini/test_image_url_input.yaml

@@ -6793,7 +6793,7 @@ interactions:
       access-control-allow-origin:
       - '*'
       age:
-      - '42504'
+      - '88'
       alt-svc:
       - h3=":443"; ma=2592000,h3-29=":443"; ma=2592000
       cache-control:
@@ -6809,7 +6809,7 @@ interactions:
       cross-origin-resource-policy:
       - cross-origin
       expires:
-      - Wed, 25 Feb 2026 21:07:23 GMT
+      - Wed, 04 Mar 2026 10:32:03 GMT
       last-modified:
       - Mon, 30 Aug 2123 17:01:05 GMT
       report-to:
@@ -6850,19 +6850,19 @@ interactions:
               "content": {
                 "parts": [
                   {
-                    "text": "This is not a fruit, it is an organ console."
+                    "text": "This is not a fruit; it's a pipe organ console."
                   }
                 ],
                 "role": "model"
               },
               "finishReason": "STOP",
-              "avgLogprobs": -0.4544379711151123
+              "avgLogprobs": -0.31288215092250277
             }
           ],
           "usageMetadata": {
             "promptTokenCount": 1814,
-            "candidatesTokenCount": 12,
-            "totalTokenCount": 1826,
+            "candidatesTokenCount": 14,
+            "totalTokenCount": 1828,
             "promptTokensDetails": [
               {
                 "modality": "TEXT",
@@ -6876,7 +6876,7 @@ interactions:
             "candidatesTokensDetails": [
               {
                 "modality": "TEXT",
-                "tokenCount": 12
+                "tokenCount": 14
               }
             ]
           },
@@ -6886,11 +6886,11 @@ interactions:
       alt-svc:
       - h3=":443"; ma=2592000,h3-29=":443"; ma=2592000
       content-length:
-      - '726'
+      - '730'
       content-type:
       - application/json; charset=UTF-8
       server-timing:
-      - gfet4t7; dur=6687
+      - gfet4t7; dur=1749
       transfer-encoding:
       - chunked
       vary:

{pydantic_ai-0.0.32 → pydantic_ai-0.0.33}/tests/models/test_gemini.py

@@ -45,6 +45,7 @@ from pydantic_ai.models.gemini import (
     _GeminiTools,
     _GeminiUsageMetaData,
 )
+from pydantic_ai.providers.google_gla import GoogleGLAProvider
 from pydantic_ai.result import Usage
 from pydantic_ai.tools import ToolDefinition
 
@@ -55,9 +56,8 @@ pytestmark = pytest.mark.anyio
 
 def test_api_key_arg(env: TestEnv):
     env.set('GEMINI_API_KEY', 'via-env-var')
-    m = GeminiModel('gemini-1.5-flash', api_key='via-arg')
-    assert isinstance(m.auth, ApiKeyAuth)
-    assert m.auth.api_key == 'via-arg'
+    m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(api_key='via-arg'))
+    assert m.client.headers['x-goog-api-key'] == 'via-arg'
 
 
 def test_api_key_env_var(env: TestEnv):
@@ -80,11 +80,10 @@ def test_api_key_empty(env: TestEnv):
 
 
 async def test_model_simple(allow_model_requests: None):
-    m = GeminiModel('gemini-1.5-flash', api_key='via-arg')
-    assert isinstance(m.http_client, httpx.AsyncClient)
+    m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(api_key='via-arg'))
+    assert isinstance(m.client, httpx.AsyncClient)
     assert m.model_name == 'gemini-1.5-flash'
-    assert isinstance(m.auth, ApiKeyAuth)
-    assert m.auth.api_key == 'via-arg'
+    assert 'x-goog-api-key' in m.client.headers
 
     arc = ModelRequestParameters(function_tools=[], allow_text_result=True, result_tools=[])
     tools = m._get_tools(arc)
@@ -94,7 +93,7 @@ async def test_model_simple(allow_model_requests: None):
 
 
 async def test_model_tools(allow_model_requests: None):
-    m = GeminiModel('gemini-1.5-flash', api_key='via-arg')
+    m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(api_key='via-arg'))
     tools = [
         ToolDefinition(
             'foo',
@@ -153,7 +152,7 @@ async def test_model_tools(allow_model_requests: None):
 
 
 async def test_require_response_tool(allow_model_requests: None):
-    m = GeminiModel('gemini-1.5-flash', api_key='via-arg')
+    m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(api_key='via-arg'))
     result_tool = ToolDefinition(
         'result',
         'This is the tool for the final Result',
@@ -212,7 +211,7 @@ async def test_json_def_replaced(allow_model_requests: None):
         }
     )
 
-    m = GeminiModel('gemini-1.5-flash', api_key='via-arg')
+    m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(api_key='via-arg'))
     result_tool = ToolDefinition(
         'result',
         'This is the tool for the final Result',
@@ -259,7 +258,7 @@ async def test_json_def_replaced_any_of(allow_model_requests: None):
 
     json_schema = Locations.model_json_schema()
 
-    m = GeminiModel('gemini-1.5-flash', api_key='via-arg')
+    m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(api_key='via-arg'))
     result_tool = ToolDefinition(
         'result',
         'This is the tool for the final Result',
@@ -322,7 +321,7 @@ async def test_json_def_recursive(allow_model_requests: None):
         }
     )
 
-    m = GeminiModel('gemini-1.5-flash', api_key='via-arg')
+    m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(api_key='via-arg'))
     result_tool = ToolDefinition(
         'result',
         'This is the tool for the final Result',
@@ -354,7 +353,7 @@ async def test_json_def_date(allow_model_requests: None):
         }
     )
 
-    m = GeminiModel('gemini-1.5-flash', api_key='via-arg')
+    m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(api_key='via-arg'))
     result_tool = ToolDefinition(
         'result',
         'This is the tool for the final Result',
@@ -451,7 +450,7 @@ def example_usage() -> _GeminiUsageMetaData:
 async def test_text_success(get_gemini_client: GetGeminiClient):
     response = gemini_response(_content_model_response(ModelResponse(parts=[TextPart('Hello world')])))
     gemini_client = get_gemini_client(response)
-    m = GeminiModel('gemini-1.5-flash', http_client=gemini_client)
+    m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(http_client=gemini_client))
     agent = Agent(m)
 
     result = await agent.run('Hello')
@@ -493,7 +492,7 @@ async def test_request_structured_response(get_gemini_client: GetGeminiClient):
         _content_model_response(ModelResponse(parts=[ToolCallPart('final_result', {'response': [1, 2, 123]})]))
     )
     gemini_client = get_gemini_client(response)
-    m = GeminiModel('gemini-1.5-flash', http_client=gemini_client)
+    m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(http_client=gemini_client))
     agent = Agent(m, result_type=list[int])
 
     result = await agent.run('Hello')
@@ -540,7 +539,7 @@ async def test_request_tool_call(get_gemini_client: GetGeminiClient):
         gemini_response(_content_model_response(ModelResponse(parts=[TextPart('final response')]))),
     ]
     gemini_client = get_gemini_client(responses)
-    m = GeminiModel('gemini-1.5-flash', http_client=gemini_client)
+    m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(http_client=gemini_client))
     agent = Agent(m, system_prompt='this is the system prompt')
 
     @agent.tool_plain
@@ -622,7 +621,7 @@ async def test_unexpected_response(client_with_handler: ClientWithHandler, env:
         return httpx.Response(401, content='invalid request')
 
     gemini_client = client_with_handler(handler)
-    m = GeminiModel('gemini-1.5-flash', http_client=gemini_client)
+    m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(http_client=gemini_client))
     agent = Agent(m, system_prompt='this is the system prompt')
 
     with pytest.raises(ModelHTTPError) as exc_info:
@@ -639,7 +638,7 @@ async def test_stream_text(get_gemini_client: GetGeminiClient):
     json_data = _gemini_streamed_response_ta.dump_json(responses, by_alias=True)
     stream = AsyncByteStreamList([json_data[:100], json_data[100:200], json_data[200:]])
     gemini_client = get_gemini_client(stream)
-    m = GeminiModel('gemini-1.5-flash', http_client=gemini_client)
+    m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(http_client=gemini_client))
     agent = Agent(m)
 
     async with agent.run_stream('Hello') as result:
@@ -684,7 +683,7 @@ async def test_stream_invalid_unicode_text(get_gemini_client: GetGeminiClient):
 
     stream = AsyncByteStreamList(parts)
     gemini_client = get_gemini_client(stream)
-    m = GeminiModel('gemini-1.5-flash', http_client=gemini_client)
+    m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(http_client=gemini_client))
     agent = Agent(m)
 
     async with agent.run_stream('Hello') as result:
@@ -698,7 +697,7 @@ async def test_stream_text_no_data(get_gemini_client: GetGeminiClient):
     json_data = _gemini_streamed_response_ta.dump_json(responses, by_alias=True)
     stream = AsyncByteStreamList([json_data[:100], json_data[100:200], json_data[200:]])
     gemini_client = get_gemini_client(stream)
-    m = GeminiModel('gemini-1.5-flash', http_client=gemini_client)
+    m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(http_client=gemini_client))
     agent = Agent(m)
     with pytest.raises(UnexpectedModelBehavior, match='Streamed response ended without con'):
         async with agent.run_stream('Hello'):
@@ -714,7 +713,7 @@ async def test_stream_structured(get_gemini_client: GetGeminiClient):
     json_data = _gemini_streamed_response_ta.dump_json(responses, by_alias=True)
     stream = AsyncByteStreamList([json_data[:100], json_data[100:200], json_data[200:]])
     gemini_client = get_gemini_client(stream)
-    model = GeminiModel('gemini-1.5-flash', http_client=gemini_client)
+    model = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(http_client=gemini_client))
     agent = Agent(model, result_type=tuple[int, int])
 
     async with agent.run_stream('Hello') as result:
@@ -744,7 +743,7 @@ async def test_stream_structured_tool_calls(get_gemini_client: GetGeminiClient):
     second_stream = AsyncByteStreamList([d2[:100], d2[100:]])
 
     gemini_client = get_gemini_client([first_stream, second_stream])
-    model = GeminiModel('gemini-1.5-flash', http_client=gemini_client)
+    model = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(http_client=gemini_client))
     agent = Agent(model, result_type=tuple[int, int])
     tool_calls: list[str] = []
 
@@ -817,7 +816,7 @@ async def test_stream_text_heterogeneous(get_gemini_client: GetGeminiClient):
     json_data = _gemini_streamed_response_ta.dump_json(responses, by_alias=True)
     stream = AsyncByteStreamList([json_data[:100], json_data[100:200], json_data[200:]])
     gemini_client = get_gemini_client(stream)
-    m = GeminiModel('gemini-1.5-flash', http_client=gemini_client)
+    m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(http_client=gemini_client))
     agent = Agent(m)
 
     @agent.tool_plain()
@@ -887,7 +886,7 @@ async def test_model_settings(client_with_handler: ClientWithHandler, env: TestE
         )
 
     gemini_client = client_with_handler(handler)
-    m = GeminiModel('gemini-1.5-flash', http_client=gemini_client, api_key='mock')
+    m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(http_client=gemini_client, api_key='mock'))
     agent = Agent(m)
 
     result = await agent.run(
@@ -939,7 +938,8 @@ async def test_safety_settings_unsafe(
             )
 
         gemini_client = client_with_handler(handler)
-        m = GeminiModel('gemini-1.5-flash', http_client=gemini_client, api_key='mock')
+
+        m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(http_client=gemini_client, api_key='mock'))
         agent = Agent(m)
 
         await agent.run(
@@ -975,7 +975,7 @@ async def test_safety_settings_safe(
         )
 
     gemini_client = client_with_handler(handler)
-    m = GeminiModel('gemini-1.5-flash', http_client=gemini_client, api_key='mock')
+    m = GeminiModel('gemini-1.5-flash', provider=GoogleGLAProvider(http_client=gemini_client, api_key='mock'))
     agent = Agent(m)
 
     result = await agent.run(
@@ -994,7 +994,7 @@ async def test_safety_settings_safe(
 async def test_image_as_binary_content_input(
     allow_model_requests: None, gemini_api_key: str, image_content: BinaryContent
 ) -> None:
-    m = GeminiModel('gemini-2.0-flash', api_key=gemini_api_key)
+    m = GeminiModel('gemini-2.0-flash', provider=GoogleGLAProvider(api_key=gemini_api_key))
     agent = Agent(m)
 
     result = await agent.run(['What is the name of this fruit?', image_content])
@@ -1003,10 +1003,10 @@ async def test_image_as_binary_content_input(
 
 @pytest.mark.vcr()
 async def test_image_url_input(allow_model_requests: None, gemini_api_key: str) -> None:
-    m = GeminiModel('gemini-2.0-flash-exp', api_key=gemini_api_key)
+    m = GeminiModel('gemini-2.0-flash-exp', provider=GoogleGLAProvider(api_key=gemini_api_key))
     agent = Agent(m)
 
     image_url = ImageUrl(url='https://goo.gle/instrument-img')
 
     result = await agent.run(['What is the name of this fruit?', image_url])
-    assert result.data == snapshot('This is not a fruit, it is an organ console.')
+    assert result.data == snapshot("This is not a fruit; it's a pipe organ console.")

{pydantic_ai-0.0.32 → pydantic_ai-0.0.33}/tests/models/test_model.py

@@ -19,7 +19,7 @@ TEST_CASES = [
         'gemini-1.5-flash',
         'google-vertex',
         'vertexai',
-        'VertexAIModel',
+        'GeminiModel',
     ),
     (
         'GEMINI_API_KEY',
@@ -27,7 +27,7 @@ TEST_CASES = [
         'gemini-1.5-flash',
         'google-vertex',
         'vertexai',
-        'VertexAIModel',
+        'GeminiModel',
     ),
     (
         'ANTHROPIC_API_KEY',

{pydantic_ai-0.0.32 → pydantic_ai-0.0.33}/tests/models/test_model_names.py

@@ -1,7 +1,8 @@
 from collections.abc import Iterator
-from typing import Any, get_args
+from typing import Any
 
 import pytest
+from typing_extensions import get_args
 
 from pydantic_ai.models import KnownModelName
 
@@ -40,10 +41,18 @@ def test_known_model_names():
     openai_names = [f'openai:{n}' for n in get_model_names(OpenAIModelName)] + [
         n for n in get_model_names(OpenAIModelName) if n.startswith('o1') or n.startswith('gpt') or n.startswith('o3')
     ]
+    deepseek_names = ['deepseek:deepseek-chat', 'deepseek:deepseek-reasoner']
     extra_names = ['test']
 
     generated_names = sorted(
-        anthropic_names + cohere_names + google_names + groq_names + mistral_names + openai_names + extra_names
+        anthropic_names
+        + cohere_names
+        + google_names
+        + groq_names
+        + mistral_names
+        + openai_names
+        + deepseek_names
+        + extra_names
     )
 
     known_model_names = sorted(get_args(KnownModelName))

{pydantic_ai-0.0.32 → pydantic_ai-0.0.33}/tests/models/test_openai.py

@@ -46,6 +46,7 @@ with try_import() as imports_successful:
     from openai.types.completion_usage import CompletionUsage, PromptTokensDetails
 
     from pydantic_ai.models.openai import OpenAIModel, OpenAISystemPromptRole
+    from pydantic_ai.providers.openai import OpenAIProvider
 
     # note: we use Union here so that casting works with Python 3.9
     MockChatCompletion = Union[chat.ChatCompletion, Exception]
@@ -58,26 +59,26 @@ pytestmark = [
 
 
 def test_init():
-    m = OpenAIModel('gpt-4o', api_key='foobar')
+    m = OpenAIModel('gpt-4o', provider=OpenAIProvider(api_key='foobar'))
     assert str(m.client.base_url) == 'https://api.openai.com/v1/'
     assert m.client.api_key == 'foobar'
     assert m.model_name == 'gpt-4o'
 
 
 def test_init_with_base_url():
-    m = OpenAIModel('gpt-4o', base_url='https://example.com/v1', api_key='foobar')
+    m = OpenAIModel('gpt-4o', provider=OpenAIProvider(base_url='https://example.com/v1', api_key='foobar'))
     assert str(m.client.base_url) == 'https://example.com/v1/'
     assert m.client.api_key == 'foobar'
     assert m.model_name == 'gpt-4o'
 
 
 def test_init_with_no_api_key_will_still_setup_client():
-    m = OpenAIModel('llama3.2', base_url='http://localhost:19434/v1')
+    m = OpenAIModel('llama3.2', provider=OpenAIProvider(base_url='http://localhost:19434/v1'))
     assert str(m.client.base_url) == 'http://localhost:19434/v1/'
 
 
 def test_init_with_non_openai_model():
-    m = OpenAIModel('llama3.2-vision:latest', base_url='https://example.com/v1/')
+    m = OpenAIModel('llama3.2-vision:latest', provider=OpenAIProvider(base_url='https://example.com/v1/'))
     assert m.model_name == 'llama3.2-vision:latest'
 
 
@@ -157,7 +158,7 @@ def completion_message(message: ChatCompletionMessage, *, usage: CompletionUsage
 async def test_request_simple_success(allow_model_requests: None):
     c = completion_message(ChatCompletionMessage(content='world', role='assistant'))
     mock_client = MockOpenAI.create_mock(c)
-    m = OpenAIModel('gpt-4o', openai_client=mock_client)
+    m = OpenAIModel('gpt-4o', provider=OpenAIProvider(openai_client=mock_client))
     agent = Agent(m)
 
     result = await agent.run('hello')
@@ -206,7 +207,7 @@ async def test_request_simple_usage(allow_model_requests: None):
         usage=CompletionUsage(completion_tokens=1, prompt_tokens=2, total_tokens=3),
     )
     mock_client = MockOpenAI.create_mock(c)
-    m = OpenAIModel('gpt-4o', openai_client=mock_client)
+    m = OpenAIModel('gpt-4o', provider=OpenAIProvider(openai_client=mock_client))
     agent = Agent(m)
 
     result = await agent.run('Hello')
@@ -229,7 +230,7 @@ async def test_request_structured_response(allow_model_requests: None):
         )
     )
     mock_client = MockOpenAI.create_mock(c)
-    m = OpenAIModel('gpt-4o', openai_client=mock_client)
+    m = OpenAIModel('gpt-4o', provider=OpenAIProvider(openai_client=mock_client))
     agent = Agent(m, result_type=list[int])
 
     result = await agent.run('Hello')
@@ -305,7 +306,7 @@ async def test_request_tool_call(allow_model_requests: None):
         completion_message(ChatCompletionMessage(content='final response', role='assistant')),
     ]
     mock_client = MockOpenAI.create_mock(responses)
-    m = OpenAIModel('gpt-4o', openai_client=mock_client)
+    m = OpenAIModel('gpt-4o', provider=OpenAIProvider(openai_client=mock_client))
     agent = Agent(m, system_prompt='this is the system prompt')
 
     @agent.tool_plain
@@ -408,7 +409,7 @@ def text_chunk(text: str, finish_reason: FinishReason | None = None) -> chat.Cha
 async def test_stream_text(allow_model_requests: None):
     stream = [text_chunk('hello '), text_chunk('world'), chunk([])]
     mock_client = MockOpenAI.create_mock_stream(stream)
-    m = OpenAIModel('gpt-4o', openai_client=mock_client)
+    m = OpenAIModel('gpt-4o', provider=OpenAIProvider(openai_client=mock_client))
     agent = Agent(m)
 
     async with agent.run_stream('') as result:
@@ -425,7 +426,7 @@ async def test_stream_text_finish_reason(allow_model_requests: None):
         text_chunk('.', finish_reason='stop'),
     ]
     mock_client = MockOpenAI.create_mock_stream(stream)
-    m = OpenAIModel('gpt-4o', openai_client=mock_client)
+    m = OpenAIModel('gpt-4o', provider=OpenAIProvider(openai_client=mock_client))
     agent = Agent(m)
 
     async with agent.run_stream('') as result:
@@ -472,7 +473,7 @@ async def test_stream_structured(allow_model_requests: None):
         chunk([]),
     ]
     mock_client = MockOpenAI.create_mock_stream(stream)
-    m = OpenAIModel('gpt-4o', openai_client=mock_client)
+    m = OpenAIModel('gpt-4o', provider=OpenAIProvider(openai_client=mock_client))
     agent = Agent(m, result_type=MyTypedDict)
 
     async with agent.run_stream('') as result:
@@ -500,7 +501,7 @@ async def test_stream_structured_finish_reason(allow_model_requests: None):
         struc_chunk(None, None, finish_reason='stop'),
     ]
     mock_client = MockOpenAI.create_mock_stream(stream)
-    m = OpenAIModel('gpt-4o', openai_client=mock_client)
+    m = OpenAIModel('gpt-4o', provider=OpenAIProvider(openai_client=mock_client))
     agent = Agent(m, result_type=MyTypedDict)
 
     async with agent.run_stream('') as result:
@@ -520,7 +521,7 @@ async def test_stream_structured_finish_reason(allow_model_requests: None):
 async def test_no_content(allow_model_requests: None):
     stream = [chunk([ChoiceDelta()]), chunk([ChoiceDelta()])]
     mock_client = MockOpenAI.create_mock_stream(stream)
-    m = OpenAIModel('gpt-4o', openai_client=mock_client)
+    m = OpenAIModel('gpt-4o', provider=OpenAIProvider(openai_client=mock_client))
     agent = Agent(m, result_type=MyTypedDict)
 
     with pytest.raises(UnexpectedModelBehavior, match='Received empty model response'):
@@ -535,7 +536,7 @@ async def test_no_delta(allow_model_requests: None):
         text_chunk('world'),
     ]
     mock_client = MockOpenAI.create_mock_stream(stream)
-    m = OpenAIModel('gpt-4o', openai_client=mock_client)
+    m = OpenAIModel('gpt-4o', provider=OpenAIProvider(openai_client=mock_client))
     agent = Agent(m)
 
     async with agent.run_stream('') as result:
@@ -553,7 +554,7 @@ async def test_system_prompt_role(
 
     c = completion_message(ChatCompletionMessage(content='world', role='assistant'))
     mock_client = MockOpenAI.create_mock(c)
-    m = OpenAIModel('gpt-4o', system_prompt_role=system_prompt_role, openai_client=mock_client)
+    m = OpenAIModel('gpt-4o', system_prompt_role=system_prompt_role, provider=OpenAIProvider(openai_client=mock_client))
     assert m.system_prompt_role == system_prompt_role
 
     agent = Agent(m, system_prompt='some instructions')
@@ -579,7 +580,9 @@ async def test_openai_o1_mini_system_role(
     system_prompt_role: Literal['system', 'developer'],
     openai_api_key: str,
 ) -> None:
-    model = OpenAIModel('o1-mini', api_key=openai_api_key, system_prompt_role=system_prompt_role)
+    model = OpenAIModel(
+        'o1-mini', provider=OpenAIProvider(api_key=openai_api_key), system_prompt_role=system_prompt_role
+    )
     agent = Agent(model=model, system_prompt='You are a helpful assistant.')
 
     with pytest.raises(ModelHTTPError, match=r".*Unsupported value: 'messages\[0\]\.role' does not support.*"):
@@ -602,7 +605,7 @@ async def test_parallel_tool_calls(allow_model_requests: None, parallel_tool_cal
         )
     )
     mock_client = MockOpenAI.create_mock(c)
-    m = OpenAIModel('gpt-4o', openai_client=mock_client)
+    m = OpenAIModel('gpt-4o', provider=OpenAIProvider(openai_client=mock_client))
     agent = Agent(m, result_type=list[int], model_settings=ModelSettings(parallel_tool_calls=parallel_tool_calls))
 
     await agent.run('Hello')
@@ -612,7 +615,7 @@ async def test_parallel_tool_calls(allow_model_requests: None, parallel_tool_cal
 async def test_image_url_input(allow_model_requests: None):
     c = completion_message(ChatCompletionMessage(content='world', role='assistant'))
     mock_client = MockOpenAI.create_mock(c)
-    m = OpenAIModel('gpt-4o', openai_client=mock_client)
+    m = OpenAIModel('gpt-4o', provider=OpenAIProvider(openai_client=mock_client))
     agent = Agent(m)
 
     result = await agent.run(
@@ -650,7 +653,7 @@ async def test_image_url_input(allow_model_requests: None):
 async def test_image_as_binary_content_input(
     allow_model_requests: None, image_content: BinaryContent, openai_api_key: str
 ):
-    m = OpenAIModel('gpt-4o', api_key=openai_api_key)
+    m = OpenAIModel('gpt-4o', provider=OpenAIProvider(api_key=openai_api_key))
     agent = Agent(m)
 
     result = await agent.run(['What fruit is in the image?', image_content])
@@ -661,7 +664,7 @@ async def test_image_as_binary_content_input(
 async def test_audio_as_binary_content_input(
     allow_model_requests: None, audio_content: BinaryContent, openai_api_key: str
 ):
-    m = OpenAIModel('gpt-4o-audio-preview', api_key=openai_api_key)
+    m = OpenAIModel('gpt-4o-audio-preview', provider=OpenAIProvider(api_key=openai_api_key))
     agent = Agent(m)
 
     result = await agent.run(['Whose name is mentioned in the audio?', audio_content])
@@ -676,7 +679,7 @@ def test_model_status_error(allow_model_requests: None) -> None:
             body={'error': 'test error'},
         )
     )
-    m = OpenAIModel('gpt-4o', openai_client=mock_client)
+    m = OpenAIModel('gpt-4o', provider=OpenAIProvider(openai_client=mock_client))
     agent = Agent(m)
     with pytest.raises(ModelHTTPError) as exc_info:
         agent.run_sync('hello')

{pydantic_ai-0.0.32 → pydantic_ai-0.0.33}/tests/models/test_vertexai.py

@@ -1,3 +1,4 @@
+# pyright: reportDeprecated=false
 from __future__ import annotations as _annotations
 
 import json
@@ -22,6 +23,8 @@ with try_import() as imports_successful:
 pytestmark = [
     pytest.mark.skipif(not imports_successful(), reason='google-auth not installed'),
     pytest.mark.anyio,
+    # This ignore is added because we should just remove the `VertexAIModel` class.
+    pytest.mark.filterwarnings('ignore::DeprecationWarning'),
 ]
 
 

pydantic_ai-0.0.33/tests/providers/test_deepseek.py

@@ -0,0 +1,48 @@
+import os
+import re
+from unittest.mock import patch
+
+import httpx
+import pytest
+
+from ..conftest import try_import
+
+with try_import() as imports_successful:
+    import openai
+
+    from pydantic_ai.providers.deepseek import DeepSeekProvider
+
+
+pytestmark = pytest.mark.skipif(not imports_successful(), reason='openai not installed')
+
+
+def test_deep_seek_provider():
+    provider = DeepSeekProvider(api_key='api-key')
+    assert provider.name == 'deepseek'
+    assert provider.base_url == 'https://api.deepseek.com'
+    assert isinstance(provider.client, openai.AsyncOpenAI)
+    assert provider.client.api_key == 'api-key'
+
+
+def test_deep_seek_provider_need_api_key() -> None:
+    with patch.dict(os.environ, {}, clear=True):
+        with pytest.raises(
+            ValueError,
+            match=re.escape(
+                'Set the `DEEPSEEK_API_KEY` environment variable or pass it via `DeepSeekProvider(api_key=...)`'
+                'to use the DeepSeek provider.'
+            ),
+        ):
+            DeepSeekProvider()
+
+
+def test_deep_seek_provider_pass_http_client() -> None:
+    http_client = httpx.AsyncClient()
+    provider = DeepSeekProvider(http_client=http_client, api_key='api-key')
+    assert provider.client._client == http_client  # type: ignore[reportPrivateUsage]
+
+
+def test_deep_seek_pass_openai_client() -> None:
+    openai_client = openai.AsyncOpenAI(api_key='api-key')
+    provider = DeepSeekProvider(openai_client=openai_client)
+    assert provider.client == openai_client

pydantic_ai-0.0.33/tests/providers/test_google_gla.py

@@ -0,0 +1,19 @@
+import os
+import re
+from unittest.mock import patch
+
+import pytest
+
+from pydantic_ai.providers.google_gla import GoogleGLAProvider
+
+
+def test_google_gla_provider_need_api_key() -> None:
+    with patch.dict(os.environ, {}, clear=True):
+        with pytest.raises(
+            ValueError,
+            match=re.escape(
+                'Set the `GEMINI_API_KEY` environment variable or pass it via `GoogleGLAProvider(api_key=...)`'
+                'to use the Google GLA provider.'
+            ),
+        ):
+            GoogleGLAProvider()

pydantic_ai-0.0.33/tests/providers/test_google_vertex.py

@@ -0,0 +1,110 @@
+from __future__ import annotations as _annotations
+
+import json
+from dataclasses import dataclass
+from pathlib import Path
+from unittest.mock import patch
+
+import httpx
+import pytest
+from inline_snapshot import snapshot
+
+from ..conftest import try_import
+
+with try_import() as imports_successful:
+    from google.auth.transport.requests import Request
+
+    from pydantic_ai.providers.google_vertex import GoogleVertexProvider
+
+pytestmark = [
+    pytest.mark.skipif(not imports_successful(), reason='google-genai not installed'),
+    pytest.mark.anyio(),
+]
+
+
+@pytest.fixture()
+def http_client():
+    async def handler(request: httpx.Request):
+        if (
+            request.url.path
+            == '/v1/projects/my-project-id/locations/us-central1/publishers/google/models/gemini-1.0-pro:generateContent'
+        ):
+            return httpx.Response(200, json={'content': 'success'})
+        raise NotImplementedError(f'Unexpected request: {request.url!r}')  # pragma: no cover
+
+    return httpx.AsyncClient(transport=httpx.MockTransport(handler=handler))
+
+
+def test_google_vertex_provider(allow_model_requests: None) -> None:
+    provider = GoogleVertexProvider()
+    assert provider.name == 'google-vertex'
+    assert provider.base_url == snapshot(
+        'https://us-central1-aiplatform.googleapis.com/v1/projects/None/locations/us-central1/publishers/google/models/'
+    )
+    assert isinstance(provider.client, httpx.AsyncClient)
+
+
+@dataclass
+class NoOpCredentials:
+    token = 'my-token'
+
+    def refresh(self, request: Request): ...
+
+
+@patch('pydantic_ai.providers.google_vertex.google.auth.default', return_value=(NoOpCredentials(), 'my-project-id'))
+async def test_google_vertex_provider_auth(allow_model_requests: None, http_client: httpx.AsyncClient):
+    provider = GoogleVertexProvider(http_client=http_client)
+    await provider.client.post('/gemini-1.0-pro:generateContent')
+    assert provider.region == 'us-central1'
+    assert getattr(provider.client.auth, 'project_id') == 'my-project-id'
+    assert getattr(provider.client.auth, 'token_created') is not None
+
+
+async def test_google_vertex_provider_service_account_file(
+    monkeypatch: pytest.MonkeyPatch, tmp_path: Path, allow_model_requests: None
+):
+    service_account_path = tmp_path / 'service_account.json'
+    save_service_account(service_account_path, 'my-project-id')
+
+    provider = GoogleVertexProvider(service_account_file=service_account_path)
+    monkeypatch.setattr(provider.client.auth, '_refresh_token', lambda: 'my-token')
+    await provider.client.post('/gemini-1.0-pro:generateContent')
+    assert provider.region == 'us-central1'
+    assert getattr(provider.client.auth, 'project_id') == 'my-project-id'
+    assert getattr(provider.client.auth, 'token_created') is not None
+
+
+def save_service_account(service_account_path: Path, project_id: str) -> None:
+    service_account = {
+        'type': 'service_account',
+        'project_id': project_id,
+        'private_key_id': 'abc',
+        # this is just a random private key I created with `openssl genpke ...`, it doesn't do anything
+        'private_key': (
+            '-----BEGIN PRIVATE KEY-----\n'
+            'MIICdgIBADANBgkqhkiG9w0BAQEFAASCAmAwggJcAgEAAoGBAMFrZYX4gZ20qv88\n'
+            'jD0QCswXgcxgP7Ta06G47QEFprDVcv4WMUBDJVAKofzVcYyhsasWsOSxcpA8LIi9\n'
+            '/VS2Otf8CmIK6nPBCD17Qgt8/IQYXOS4U2EBh0yjo0HQ4vFpkqium4lLWxrAZohA\n'
+            '8r82clV08iLRUW3J+xvN23iPHyVDAgMBAAECgYBScRJe3iNxMvbHv+kOhe30O/jJ\n'
+            'QiUlUzhtcEMk8mGwceqHvrHTcEtRKJcPC3NQvALcp9lSQQhRzjQ1PLXkC6BcfKFd\n'
+            '03q5tVPmJiqsHbSyUyHWzdlHP42xWpl/RmX/DfRKGhPOvufZpSTzkmKWtN+7osHu\n'
+            '7eiMpg2EDswCvOgf0QJBAPXLYwHbZLaM2KEMDgJSse5ZTE/0VMf+5vSTGUmHkr9c\n'
+            'Wx2G1i258kc/JgsXInPbq4BnK9hd0Xj2T5cmEmQtm4UCQQDJc02DFnPnjPnnDUwg\n'
+            'BPhrCyW+rnBGUVjehveu4XgbGx7l3wsbORTaKdCX3HIKUupgfFwFcDlMUzUy6fPO\n'
+            'IuQnAkA8FhVE/fIX4kSO0hiWnsqafr/2B7+2CG1DOraC0B6ioxwvEqhHE17T5e8R\n'
+            '5PzqH7hEMnR4dy7fCC+avpbeYHvVAkA5W58iR+5Qa49r/hlCtKeWsuHYXQqSuu62\n'
+            'zW8QWBo+fYZapRsgcSxCwc0msBm4XstlFYON+NoXpUlsabiFZOHZAkEA8Ffq3xoU\n'
+            'y0eYGy3MEzxx96F+tkl59lfkwHKWchWZJ95vAKWJaHx9WFxSWiJofbRna8Iim6pY\n'
+            'BootYWyTCfjjwA==\n'
+            '-----END PRIVATE KEY-----\n'
+        ),
+        'client_email': 'testing-pydantic-ai@pydantic-ai.iam.gserviceaccount.com',
+        'client_id': '123',
+        'auth_uri': 'https://accounts.google.com/o/oauth2/auth',
+        'token_uri': 'https://oauth2.googleapis.com/token',
+        'auth_provider_x509_cert_url': 'https://www.googleapis.com/oauth2/v1/certs',
+        'client_x509_cert_url': 'https://www.googleapis.com/...',
+        'universe_domain': 'googleapis.com',
+    }
+
+    service_account_path.write_text(json.dumps(service_account, indent=2))

pydantic_ai-0.0.33/tests/providers/test_provider_names.py

@@ -0,0 +1,44 @@
+from __future__ import annotations as _annotations
+
+import os
+from typing import Any
+from unittest.mock import patch
+
+import pytest
+
+from pydantic_ai.providers import Provider, infer_provider
+
+from ..conftest import try_import
+
+with try_import() as imports_successful:
+    from pydantic_ai.providers.deepseek import DeepSeekProvider
+    from pydantic_ai.providers.google_gla import GoogleGLAProvider
+    from pydantic_ai.providers.google_vertex import GoogleVertexProvider
+    from pydantic_ai.providers.openai import OpenAIProvider
+
+    test_infer_provider_params = [
+        ('deepseek', DeepSeekProvider, 'DEEPSEEK_API_KEY'),
+        ('openai', OpenAIProvider, None),
+        ('google-vertex', GoogleVertexProvider, None),
+        ('google-gla', GoogleGLAProvider, 'GEMINI_API_KEY'),
+    ]
+
+if not imports_successful():
+    test_infer_provider_params = []
+
+pytestmark = pytest.mark.skipif(not imports_successful(), reason='need to install all extra packages')
+
+
+@pytest.fixture(scope='module', autouse=True)
+def empty_env():
+    with patch.dict(os.environ, {}, clear=True):
+        yield
+
+
+@pytest.mark.parametrize(('provider', 'provider_cls', 'exception_has'), test_infer_provider_params)
+def test_infer_provider(provider: str, provider_cls: type[Provider[Any]], exception_has: str | None):
+    if exception_has is not None:
+        with pytest.raises(ValueError, match=rf'.*{exception_has}.*'):
+            infer_provider(provider)
+    else:
+        assert isinstance(infer_provider(provider), provider_cls)

{pydantic_ai-0.0.32 → pydantic_ai-0.0.33}/tests/test_examples.py

@@ -36,9 +36,9 @@ from pydantic_ai.models.test import TestModel
 from .conftest import ClientWithHandler, TestEnv
 
 try:
-    from pydantic_ai.models.vertexai import VertexAIModel
+    from pydantic_ai.providers.google_vertex import GoogleVertexProvider
 except ImportError:
-    VertexAIModel = None
+    GoogleVertexProvider = None
 
 
 try:
@@ -47,7 +47,9 @@ except ImportError:
     logfire = None
 
 
-pytestmark = pytest.mark.skipif(VertexAIModel is None or logfire is None, reason='google-auth or logfire not installed')
+pytestmark = pytest.mark.skipif(
+    GoogleVertexProvider is None or logfire is None, reason='google-auth or logfire not installed'
+)
 
 
 def find_filter_examples() -> Iterable[CodeExample]:

{pydantic_ai-0.0.32 → pydantic_ai-0.0.33}/tests/test_live.py

@@ -23,23 +23,29 @@ pytestmark = [
 
 def openai(http_client: httpx.AsyncClient, _tmp_path: Path) -> Model:
     from pydantic_ai.models.openai import OpenAIModel
+    from pydantic_ai.providers.openai import OpenAIProvider
 
-    return OpenAIModel('gpt-4o-mini', http_client=http_client)
+    return OpenAIModel('gpt-4o-mini', provider=OpenAIProvider(http_client=http_client))
 
 
 def gemini(http_client: httpx.AsyncClient, _tmp_path: Path) -> Model:
     from pydantic_ai.models.gemini import GeminiModel
+    from pydantic_ai.providers.google_gla import GoogleGLAProvider
 
-    return GeminiModel('gemini-1.5-pro', http_client=http_client)
+    return GeminiModel('gemini-1.5-pro', provider=GoogleGLAProvider(http_client=http_client))
 
 
 def vertexai(http_client: httpx.AsyncClient, tmp_path: Path) -> Model:
-    from pydantic_ai.models.vertexai import VertexAIModel
+    from pydantic_ai.models.gemini import GeminiModel
+    from pydantic_ai.providers.google_vertex import GoogleVertexProvider
 
     service_account_content = os.environ['GOOGLE_SERVICE_ACCOUNT_CONTENT']
     service_account_path = tmp_path / 'service_account.json'
     service_account_path.write_text(service_account_content)
-    return VertexAIModel('gemini-1.5-flash', service_account_file=service_account_path, http_client=http_client)
+    return GeminiModel(
+        'gemini-1.5-flash',
+        provider=GoogleVertexProvider(service_account_file=service_account_path, http_client=http_client),
+    )
 
 
 def groq(http_client: httpx.AsyncClient, _tmp_path: Path) -> Model:
@@ -56,8 +62,11 @@ def anthropic(http_client: httpx.AsyncClient, _tmp_path: Path) -> Model:
 
 def ollama(http_client: httpx.AsyncClient, _tmp_path: Path) -> Model:
     from pydantic_ai.models.openai import OpenAIModel
+    from pydantic_ai.providers.openai import OpenAIProvider
 
-    return OpenAIModel('qwen2:0.5b', base_url='http://localhost:11434/v1/', http_client=http_client)
+    return OpenAIModel(
+        'qwen2:0.5b', provider=OpenAIProvider(base_url='http://localhost:11434/v1/', http_client=http_client)
+    )
 
 
 def mistral(http_client: httpx.AsyncClient, _tmp_path: Path) -> Model: