pydantic-ai-slim 1.1.0__py3-none-any.whl → 1.2.1__py3-none-any.whl

pydantic_ai/_agent_graph.py

@@ -710,6 +710,18 @@ class CallToolsNode(AgentNode[DepsT, NodeRunEndT]):
     __repr__ = dataclasses_no_defaults_repr
 
 
+@dataclasses.dataclass
+class SetFinalResult(AgentNode[DepsT, NodeRunEndT]):
+    """A node that immediately ends the graph run after a streaming response produced a final result."""
+
+    final_result: result.FinalResult[NodeRunEndT]
+
+    async def run(
+        self, ctx: GraphRunContext[GraphAgentState, GraphAgentDeps[DepsT, NodeRunEndT]]
+    ) -> End[result.FinalResult[NodeRunEndT]]:
+        return End(self.final_result)
+
+
 def build_run_context(ctx: GraphRunContext[GraphAgentState, GraphAgentDeps[DepsT, Any]]) -> RunContext[DepsT]:
     """Build a `RunContext` object from the current agent graph run context."""
     return RunContext[DepsT](
@@ -1123,6 +1135,7 @@ def build_agent_graph(
         UserPromptNode[DepsT],
         ModelRequestNode[DepsT],
         CallToolsNode[DepsT],
+        SetFinalResult[DepsT],
     )
     graph = Graph[GraphAgentState, GraphAgentDeps[DepsT, Any], result.FinalResult[OutputT]](
         nodes=nodes,

pydantic_ai/agent/abstract.py

@@ -524,6 +524,14 @@ class AbstractAgent(Generic[AgentDepsT, OutputDataT], ABC):
                                     await stream.get_output(), final_result.tool_name, final_result.tool_call_id
                                 )
 
+                                # When we get here, the `ModelRequestNode` has completed streaming after the final result was found.
+                                # When running an agent with `agent.run`, we'd then move to `CallToolsNode` to execute the tool calls and
+                                # find the final result.
+                                # We also want to execute tool calls (in case `agent.end_strategy == 'exhaustive'`) here, but
+                                # we don't want to use run the `CallToolsNode` logic to determine the final output, as it would be
+                                # wasteful and could produce a different result (e.g. when text output is followed by tool calls).
+                                # So we call `process_tool_calls` directly and then end the run with the found final result.
+
                                 parts: list[_messages.ModelRequestPart] = []
                                 async for _event in _agent_graph.process_tool_calls(
                                     tool_manager=graph_ctx.deps.tool_manager,
@@ -534,9 +542,13 @@ class AbstractAgent(Generic[AgentDepsT, OutputDataT], ABC):
                                     output_parts=parts,
                                 ):
                                     pass
+
+                                # For backwards compatibility, append a new ModelRequest using the tool returns and retries
                                 if parts:
                                     messages.append(_messages.ModelRequest(parts))
 
+                                await agent_run.next(_agent_graph.SetFinalResult(final_result))
+
                             yield StreamedRunResult(
                                 messages,
                                 graph_ctx.deps.new_message_index,

pydantic_ai/models/__init__.py

@@ -55,6 +55,8 @@ KnownModelName = TypeAliasType(
         'anthropic:claude-3-5-sonnet-20240620',
         'anthropic:claude-3-5-sonnet-20241022',
         'anthropic:claude-3-5-sonnet-latest',
+        'anthropic:claude-haiku-4-5',
+        'anthropic:claude-haiku-4-5-20251001',
         'anthropic:claude-3-7-sonnet-20250219',
         'anthropic:claude-3-7-sonnet-latest',
         'anthropic:claude-3-haiku-20240307',

pydantic_ai/models/openai.py

@@ -600,7 +600,7 @@ class OpenAIChatModel(Model):
 
         return ModelResponse(
             parts=items,
-            usage=_map_usage(response),
+            usage=_map_usage(response, self._provider.name, self._provider.base_url, self._model_name),
             model_name=response.model,
             timestamp=timestamp,
             provider_details=vendor_details or None,
@@ -631,6 +631,7 @@ class OpenAIChatModel(Model):
             _response=peekable_response,
             _timestamp=number_to_datetime(first_chunk.created),
             _provider_name=self._provider.name,
+            _provider_url=self._provider.base_url,
         )
 
     def _get_tools(self, model_request_parameters: ModelRequestParameters) -> list[chat.ChatCompletionToolParam]:
@@ -1061,7 +1062,7 @@ class OpenAIResponsesModel(Model):
 
         return ModelResponse(
             parts=items,
-            usage=_map_usage(response),
+            usage=_map_usage(response, self._provider.name, self._provider.base_url, self._model_name),
             model_name=response.model,
             provider_response_id=response.id,
             timestamp=timestamp,
@@ -1088,6 +1089,7 @@ class OpenAIResponsesModel(Model):
             _response=peekable_response,
             _timestamp=number_to_datetime(first_chunk.response.created_at),
             _provider_name=self._provider.name,
+            _provider_url=self._provider.base_url,
         )
 
     @overload
@@ -1589,10 +1591,11 @@ class OpenAIStreamedResponse(StreamedResponse):
     _response: AsyncIterable[ChatCompletionChunk]
     _timestamp: datetime
     _provider_name: str
+    _provider_url: str
 
     async def _get_event_iterator(self) -> AsyncIterator[ModelResponseStreamEvent]:
         async for chunk in self._response:
-            self._usage += _map_usage(chunk)
+            self._usage += _map_usage(chunk, self._provider_name, self._provider_url, self._model_name)
 
             if chunk.id:  # pragma: no branch
                 self.provider_response_id = chunk.id
@@ -1683,12 +1686,13 @@ class OpenAIResponsesStreamedResponse(StreamedResponse):
     _response: AsyncIterable[responses.ResponseStreamEvent]
     _timestamp: datetime
     _provider_name: str
+    _provider_url: str
 
     async def _get_event_iterator(self) -> AsyncIterator[ModelResponseStreamEvent]:  # noqa: C901
         async for chunk in self._response:
             # NOTE: You can inspect the builtin tools used checking the `ResponseCompletedEvent`.
             if isinstance(chunk, responses.ResponseCompletedEvent):
-                self._usage += _map_usage(chunk.response)
+                self._usage += self._map_usage(chunk.response)
 
                 raw_finish_reason = (
                     details.reason if (details := chunk.response.incomplete_details) else chunk.response.status
@@ -1708,7 +1712,7 @@ class OpenAIResponsesStreamedResponse(StreamedResponse):
                     self.provider_response_id = chunk.response.id
 
             elif isinstance(chunk, responses.ResponseFailedEvent):  # pragma: no cover
-                self._usage += _map_usage(chunk.response)
+                self._usage += self._map_usage(chunk.response)
 
             elif isinstance(chunk, responses.ResponseFunctionCallArgumentsDeltaEvent):
                 maybe_event = self._parts_manager.handle_tool_call_delta(
@@ -1722,10 +1726,10 @@ class OpenAIResponsesStreamedResponse(StreamedResponse):
                 pass  # there's nothing we need to do here
 
             elif isinstance(chunk, responses.ResponseIncompleteEvent):  # pragma: no cover
-                self._usage += _map_usage(chunk.response)
+                self._usage += self._map_usage(chunk.response)
 
             elif isinstance(chunk, responses.ResponseInProgressEvent):
-                self._usage += _map_usage(chunk.response)
+                self._usage += self._map_usage(chunk.response)
 
             elif isinstance(chunk, responses.ResponseOutputItemAddedEvent):
                 if isinstance(chunk.item, responses.ResponseFunctionToolCall):
@@ -1906,6 +1910,9 @@ class OpenAIResponsesStreamedResponse(StreamedResponse):
                     UserWarning,
                 )
 
+    def _map_usage(self, response: responses.Response):
+        return _map_usage(response, self._provider_name, self._provider_url, self._model_name)
+
     @property
     def model_name(self) -> OpenAIModelName:
         """Get the model name of the response."""
@@ -1922,55 +1929,45 @@ class OpenAIResponsesStreamedResponse(StreamedResponse):
         return self._timestamp
 
 
-def _map_usage(response: chat.ChatCompletion | ChatCompletionChunk | responses.Response) -> usage.RequestUsage:
+def _map_usage(
+    response: chat.ChatCompletion | ChatCompletionChunk | responses.Response,
+    provider: str,
+    provider_url: str,
+    model: str,
+) -> usage.RequestUsage:
     response_usage = response.usage
     if response_usage is None:
         return usage.RequestUsage()
-    elif isinstance(response_usage, responses.ResponseUsage):
-        details: dict[str, int] = {
-            key: value
-            for key, value in response_usage.model_dump(
-                exclude={'input_tokens', 'output_tokens', 'total_tokens'}
-            ).items()
-            if isinstance(value, int)
-        }
-        # Handle vLLM compatibility - some providers don't include token details
-        if getattr(response_usage, 'input_tokens_details', None) is not None:
-            cache_read_tokens = response_usage.input_tokens_details.cached_tokens
-        else:
-            cache_read_tokens = 0
+
+    usage_data = response_usage.model_dump(exclude_none=True)
+    details = {
+        k: v
+        for k, v in usage_data.items()
+        if k not in {'prompt_tokens', 'completion_tokens', 'input_tokens', 'output_tokens', 'total_tokens'}
+        if isinstance(v, int)
+    }
+    response_data = dict(model=model, usage=usage_data)
+    if isinstance(response_usage, responses.ResponseUsage):
+        api_flavor = 'responses'
 
         if getattr(response_usage, 'output_tokens_details', None) is not None:
             details['reasoning_tokens'] = response_usage.output_tokens_details.reasoning_tokens
         else:
             details['reasoning_tokens'] = 0
-
-        return usage.RequestUsage(
-            input_tokens=response_usage.input_tokens,
-            output_tokens=response_usage.output_tokens,
-            cache_read_tokens=cache_read_tokens,
-            details=details,
-        )
     else:
-        details = {
-            key: value
-            for key, value in response_usage.model_dump(
-                exclude_none=True, exclude={'prompt_tokens', 'completion_tokens', 'total_tokens'}
-            ).items()
-            if isinstance(value, int)
-        }
-        u = usage.RequestUsage(
-            input_tokens=response_usage.prompt_tokens,
-            output_tokens=response_usage.completion_tokens,
-            details=details,
-        )
+        api_flavor = 'chat'
+
         if response_usage.completion_tokens_details is not None:
             details.update(response_usage.completion_tokens_details.model_dump(exclude_none=True))
-            u.output_audio_tokens = response_usage.completion_tokens_details.audio_tokens or 0
-        if response_usage.prompt_tokens_details is not None:
-            u.input_audio_tokens = response_usage.prompt_tokens_details.audio_tokens or 0
-            u.cache_read_tokens = response_usage.prompt_tokens_details.cached_tokens or 0
-        return u
+
+    return usage.RequestUsage.extract(
+        response_data,
+        provider=provider,
+        provider_url=provider_url,
+        provider_fallback='openai',
+        api_flavor=api_flavor,
+        details=details,
+    )
 
 
 def _split_combined_tool_call_id(combined_id: str) -> tuple[str, str | None]:

pydantic_ai/providers/gateway.py

@@ -83,7 +83,7 @@ def gateway_provider(
             ' to use the Pydantic AI Gateway provider.'
         )
 
-    base_url = base_url or os.getenv('PYDANTIC_AI_GATEWAY_BASE_URL', 'http://localhost:8787/proxy')
+    base_url = base_url or os.getenv('PYDANTIC_AI_GATEWAY_BASE_URL', 'https://gateway.pydantic.dev/proxy')
     http_client = http_client or cached_async_http_client(provider=f'gateway-{upstream_provider}')
     http_client.event_hooks = {'request': [_request_hook]}
 

{pydantic_ai_slim-1.1.0.dist-info → pydantic_ai_slim-1.2.1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pydantic-ai-slim
-Version: 1.1.0
+Version: 1.2.1
 Summary: Agent Framework / shim to use Pydantic with LLMs, slim package
 Project-URL: Homepage, https://github.com/pydantic/pydantic-ai/tree/main/pydantic_ai_slim
 Project-URL: Source, https://github.com/pydantic/pydantic-ai/tree/main/pydantic_ai_slim
@@ -29,11 +29,11 @@ Classifier: Topic :: Internet
 Classifier: Topic :: Software Development :: Libraries :: Python Modules
 Requires-Python: >=3.10
 Requires-Dist: exceptiongroup; python_version < '3.11'
-Requires-Dist: genai-prices>=0.0.30
+Requires-Dist: genai-prices>=0.0.31
 Requires-Dist: griffe>=1.3.2
 Requires-Dist: httpx>=0.27
 Requires-Dist: opentelemetry-api>=1.28.0
-Requires-Dist: pydantic-graph==1.1.0
+Requires-Dist: pydantic-graph==1.2.1
 Requires-Dist: pydantic>=2.10
 Requires-Dist: typing-inspection>=0.4.0
 Provides-Extra: a2a
@@ -42,7 +42,7 @@ Provides-Extra: ag-ui
 Requires-Dist: ag-ui-protocol>=0.1.8; extra == 'ag-ui'
 Requires-Dist: starlette>=0.45.3; extra == 'ag-ui'
 Provides-Extra: anthropic
-Requires-Dist: anthropic>=0.69.0; extra == 'anthropic'
+Requires-Dist: anthropic>=0.70.0; extra == 'anthropic'
 Provides-Extra: bedrock
 Requires-Dist: boto3>=1.39.0; extra == 'bedrock'
 Provides-Extra: cli
@@ -57,7 +57,7 @@ Requires-Dist: dbos>=1.14.0; extra == 'dbos'
 Provides-Extra: duckduckgo
 Requires-Dist: ddgs>=9.0.0; extra == 'duckduckgo'
 Provides-Extra: evals
-Requires-Dist: pydantic-evals==1.1.0; extra == 'evals'
+Requires-Dist: pydantic-evals==1.2.1; extra == 'evals'
 Provides-Extra: google
 Requires-Dist: google-genai>=1.31.0; extra == 'google'
 Provides-Extra: groq

{pydantic_ai_slim-1.1.0.dist-info → pydantic_ai_slim-1.2.1.dist-info}/RECORD

@@ -1,7 +1,7 @@
 pydantic_ai/__init__.py,sha256=IgLTfgpGwbYsT_d_2wSucOfFyIMl1GH6v-yfkNs_zrM,5149
 pydantic_ai/__main__.py,sha256=Q_zJU15DUA01YtlJ2mnaLCoId2YmgmreVEERGuQT-Y0,132
 pydantic_ai/_a2a.py,sha256=3_pl7JW2yHdu31qLgCrdcTZTqXaJNjAwUV6zavah_w8,12159
-pydantic_ai/_agent_graph.py,sha256=D_Oo_LbPqGwkVAaEAcFFERwrg9pVkxeZ45vtTMhES8M,54230
+pydantic_ai/_agent_graph.py,sha256=beDv9Ixg3Q3LGgNK6BWyvsjjqA27gq3zHvceT1rXuBE,54688
 pydantic_ai/_cli.py,sha256=iZTCFrpJy3aUZ49nJQ5nw2INFw6gPVQd8EhB0rahVcI,14005
 pydantic_ai/_function_schema.py,sha256=UnDGh7Wh5z70pEaRujXF_hKsSibQdN2ywI6lZGz3LUo,11663
 pydantic_ai/_griffe.py,sha256=BphvTL00FHxsSY56GM-bNyCOdwrpL0T3LbDQITWUK_Q,5280
@@ -32,7 +32,7 @@ pydantic_ai/settings.py,sha256=0mr6KudxKKjTG8e3nsv_8vDLxNhu_1-WvefCOzCGSYM,3565
 pydantic_ai/tools.py,sha256=dCecmJtRkF1ioqFYbfT00XGGqzGB4PPO9n6IrHCQtnc,20343
 pydantic_ai/usage.py,sha256=_xXoPIfpENghWcjBvMj0URXQV6YwHWxxZYma4WZ4vUg,15710
 pydantic_ai/agent/__init__.py,sha256=VigDqMYLKQHsNYWYy6qPkqN0yfdffqxBYEA5YyxkIBM,67111
-pydantic_ai/agent/abstract.py,sha256=69kTaR-ZMEmLJ4tD3oGQS5VuomXtNL8t5mxmPz8Ao50,54587
+pydantic_ai/agent/abstract.py,sha256=Akq1NvfzXbIEJwwvo_t-FQ6MobW_cPWSeUXffdUN7Og,55651
 pydantic_ai/agent/wrapper.py,sha256=ygwfMq24mGe3pGIK-TtPAy3cV7M8VZJW3ulEHvwNTck,10293
 pydantic_ai/common_tools/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 pydantic_ai/common_tools/duckduckgo.py,sha256=1ae_o3zqMGrC6KFqAmuqPwJqQgNBTisuvU2jX9KU8PI,2273
@@ -62,7 +62,7 @@ pydantic_ai/durable_exec/temporal/_toolset.py,sha256=IlPQrumm2MpZrb518ru15s0jIl8
 pydantic_ai/ext/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 pydantic_ai/ext/aci.py,sha256=YWYLXzTQJ6hS7qfgNycA8cRl69gogGgThqEU6II7eMA,2527
 pydantic_ai/ext/langchain.py,sha256=kmbbV3Cx2BiNYEJCZMHVYQquUQD-zG2L_bwDangy0Ww,2317
-pydantic_ai/models/__init__.py,sha256=D-lOP764gxfjdaNjUujpHlvpJQuIjGfYO8ljcohQ_rE,35736
+pydantic_ai/models/__init__.py,sha256=_47Vu42NfhDkm5TeRU3s0ibVBOGZr44uyhMLzA8Gg0I,35821
 pydantic_ai/models/anthropic.py,sha256=-vW7aoPrELKnJzbooCEhMu8__jY6iqvWdFJbIKeQPa8,38087
 pydantic_ai/models/bedrock.py,sha256=fha8zVZgDFYgDqO5nvBkZ2CEv4GV92yq_YnK4qmD73E,33639
 pydantic_ai/models/cohere.py,sha256=_ccK7XBts1OwD-RP8puU3z425SZ4PeJGts1WFhPjikg,14051
@@ -75,7 +75,7 @@ pydantic_ai/models/huggingface.py,sha256=711C0ysjLYKriGfSxPiaF6lqjGcNmIaJaCvAXou
 pydantic_ai/models/instrumented.py,sha256=J8eVTutr3UP1r_wd5sM5c0BIdzkRqT-EGgd2NiF0ssQ,22319
 pydantic_ai/models/mcp_sampling.py,sha256=qY4y4nXbRpNp2QbkfjzWLvF_8KLZGXypz4cc0lYRHXU,3553
 pydantic_ai/models/mistral.py,sha256=fi57hADjYxZw8wEpAcNI6mqY32VG9hHK9GGRQ-9vlZg,33905
-pydantic_ai/models/openai.py,sha256=nCqFy2sRihygbBbTYSbMy_W9LiXwNoRzIUNHSJk5ctc,99861
+pydantic_ai/models/openai.py,sha256=1DRmsFx2beuxH5RAy2o_PRMAtxAmF2Gba-EmqyK_9BM,99457
 pydantic_ai/models/test.py,sha256=5ER66nwZG7Iwm-KkzPo4vwNd3rulzgkpgysu4YcT1W4,20568
 pydantic_ai/models/wrapper.py,sha256=nwh8Gea59blbr1JDKlUnkYICuI9TUubC4qP7iZRRW28,2440
 pydantic_ai/profiles/__init__.py,sha256=UHknN-CYsQexUaxfsgz_J_uSZ9QwistLSuAErQkvbcM,3385
@@ -100,7 +100,7 @@ pydantic_ai/providers/cerebras.py,sha256=3rIu092TYYuI5S4mlRjWxay5uomPbEDyHWIBMfr
 pydantic_ai/providers/cohere.py,sha256=L3wgvcbxRRPrIKoZka_DQl1Uvi1VxBPMJikrzJ85iHE,2839
 pydantic_ai/providers/deepseek.py,sha256=zop0sb1XzdzSuI2dCNXrinfMdxoqB8H_rp2zw6ItbKc,3023
 pydantic_ai/providers/fireworks.py,sha256=t4PznbxnD9GnzZ3wYqSn6xdxRRJlYzNKf_EZzX0UWl8,3585
-pydantic_ai/providers/gateway.py,sha256=Xgns651ndehQ3F6PkMw8CBgaFBmxQx25LjrQA8PuPjk,6937
+pydantic_ai/providers/gateway.py,sha256=QEpTLOgUCoDiL8ZOm5JZtGMm7t7Jfluk_nTd6a8LfbA,6944
 pydantic_ai/providers/github.py,sha256=yi7c16_Ao1E1QmehVfdsO9NrjDGK1moaHTK-P5cIrsI,4369
 pydantic_ai/providers/google.py,sha256=scCHek7whNEbi742hnRlKItboYOoxtYosgNN7wDjvpM,6019
 pydantic_ai/providers/google_gla.py,sha256=PnmnzgCOPJB1kMVnNVqZu2Cdzk7K9jx2z0MpbJ6EkII,1951
@@ -130,8 +130,8 @@ pydantic_ai/toolsets/prefixed.py,sha256=0KwcDkW8OM36ZUsOLVP5h-Nj2tPq78L3_E2c-1Fb
 pydantic_ai/toolsets/prepared.py,sha256=Zjfz6S8In6PBVxoKFN9sKPN984zO6t0awB7Lnq5KODw,1431
 pydantic_ai/toolsets/renamed.py,sha256=JuLHpi-hYPiSPlaTpN8WiXLiGsywYK0axi2lW2Qs75k,1637
 pydantic_ai/toolsets/wrapper.py,sha256=KRzF1p8dncHbva8CE6Ud-IC5E_aygIHlwH5atXK55k4,1673
-pydantic_ai_slim-1.1.0.dist-info/METADATA,sha256=T4LppdZllkCLfqapnpsV7PxT7tyntYq4wCc-UDV8b9A,4703
-pydantic_ai_slim-1.1.0.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-pydantic_ai_slim-1.1.0.dist-info/entry_points.txt,sha256=kbKxe2VtDCYS06hsI7P3uZGxcVC08-FPt1rxeiMpIps,50
-pydantic_ai_slim-1.1.0.dist-info/licenses/LICENSE,sha256=vA6Jc482lEyBBuGUfD1pYx-cM7jxvLYOxPidZ30t_PQ,1100
-pydantic_ai_slim-1.1.0.dist-info/RECORD,,
+pydantic_ai_slim-1.2.1.dist-info/METADATA,sha256=ToerQFMS6lH9tRpFuCCdu3IPnjTb7RqVYp6DUk1XF8k,4703
+pydantic_ai_slim-1.2.1.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+pydantic_ai_slim-1.2.1.dist-info/entry_points.txt,sha256=kbKxe2VtDCYS06hsI7P3uZGxcVC08-FPt1rxeiMpIps,50
+pydantic_ai_slim-1.2.1.dist-info/licenses/LICENSE,sha256=vA6Jc482lEyBBuGUfD1pYx-cM7jxvLYOxPidZ30t_PQ,1100
+pydantic_ai_slim-1.2.1.dist-info/RECORD,,