pydantic-ai-slim 0.0.15__py3-none-any.whl → 0.0.16__py3-none-any.whl

pydantic_ai/_griffe.py

@@ -4,8 +4,7 @@ import re
 from inspect import Signature
 from typing import Any, Callable, Literal, cast
 
-from _griffe.enumerations import DocstringSectionKind
-from _griffe.models import Docstring, Object as GriffeObject
+from griffe import Docstring, DocstringSectionKind, Object as GriffeObject
 
 DocstringStyle = Literal['google', 'numpy', 'sphinx']
 

pydantic_ai/agent.py

@@ -6,7 +6,6 @@ import inspect
 from collections.abc import AsyncIterator, Awaitable, Iterator, Sequence
 from contextlib import asynccontextmanager, contextmanager
 from contextvars import ContextVar
-from dataclasses import dataclass, field
 from types import FrameType
 from typing import Any, Callable, Generic, Literal, cast, final, overload
 
@@ -40,6 +39,16 @@ __all__ = 'Agent', 'capture_run_messages', 'EndStrategy'
 
 _logfire = logfire_api.Logfire(otel_scope='pydantic-ai')
 
+# while waiting for https://github.com/pydantic/logfire/issues/745
+try:
+    import logfire._internal.stack_info
+except ImportError:
+    pass
+else:
+    from pathlib import Path
+
+    logfire._internal.stack_info.NON_USER_CODE_PREFIXES += (str(Path(__file__).parent.absolute()),)
+
 NoneType = type(None)
 EndStrategy = Literal['early', 'exhaustive']
 """The strategy for handling multiple tool calls when a final result is found.
@@ -50,7 +59,7 @@ EndStrategy = Literal['early', 'exhaustive']
 
 
 @final
-@dataclass(init=False)
+@dataclasses.dataclass(init=False)
 class Agent(Generic[AgentDeps, ResultData]):
     """Class for defining "agents" - a way to have a specific type of "conversation" with an LLM.
 
@@ -90,17 +99,17 @@ class Agent(Generic[AgentDeps, ResultData]):
     be merged with this value, with the runtime argument taking priority.
     """
 
-    _result_schema: _result.ResultSchema[ResultData] | None = field(repr=False)
-    _result_validators: list[_result.ResultValidator[AgentDeps, ResultData]] = field(repr=False)
-    _allow_text_result: bool = field(repr=False)
-    _system_prompts: tuple[str, ...] = field(repr=False)
-    _function_tools: dict[str, Tool[AgentDeps]] = field(repr=False)
-    _default_retries: int = field(repr=False)
-    _system_prompt_functions: list[_system_prompt.SystemPromptRunner[AgentDeps]] = field(repr=False)
-    _deps_type: type[AgentDeps] = field(repr=False)
-    _max_result_retries: int = field(repr=False)
-    _override_deps: _utils.Option[AgentDeps] = field(default=None, repr=False)
-    _override_model: _utils.Option[models.Model] = field(default=None, repr=False)
+    _result_schema: _result.ResultSchema[ResultData] | None = dataclasses.field(repr=False)
+    _result_validators: list[_result.ResultValidator[AgentDeps, ResultData]] = dataclasses.field(repr=False)
+    _allow_text_result: bool = dataclasses.field(repr=False)
+    _system_prompts: tuple[str, ...] = dataclasses.field(repr=False)
+    _function_tools: dict[str, Tool[AgentDeps]] = dataclasses.field(repr=False)
+    _default_retries: int = dataclasses.field(repr=False)
+    _system_prompt_functions: list[_system_prompt.SystemPromptRunner[AgentDeps]] = dataclasses.field(repr=False)
+    _deps_type: type[AgentDeps] = dataclasses.field(repr=False)
+    _max_result_retries: int = dataclasses.field(repr=False)
+    _override_deps: _utils.Option[AgentDeps] = dataclasses.field(default=None, repr=False)
+    _override_model: _utils.Option[models.Model] = dataclasses.field(default=None, repr=False)
 
     def __init__(
         self,
@@ -184,6 +193,7 @@ class Agent(Generic[AgentDeps, ResultData]):
         deps: AgentDeps = None,
         model_settings: ModelSettings | None = None,
         usage_limits: UsageLimits | None = None,
+        usage: result.Usage | None = None,
         infer_name: bool = True,
     ) -> result.RunResult[ResultData]:
         """Run the agent with a user prompt in async mode.
@@ -206,6 +216,7 @@ class Agent(Generic[AgentDeps, ResultData]):
             deps: Optional dependencies to use for this run.
             model_settings: Optional settings to use for this model's request.
             usage_limits: Optional limits on model request count or token usage.
+            usage: Optional usage to start with, useful for resuming a conversation or agents used in tools.
             infer_name: Whether to try to infer the agent name from the call frame if it's not set.
 
         Returns:
@@ -213,7 +224,7 @@ class Agent(Generic[AgentDeps, ResultData]):
         """
         if infer_name and self.name is None:
             self._infer_name(inspect.currentframe())
-        model_used, mode_selection = await self._get_model(model)
+        model_used = await self._get_model(model)
 
         deps = self._get_deps(deps)
         new_message_index = len(message_history) if message_history else 0
@@ -222,40 +233,36 @@ class Agent(Generic[AgentDeps, ResultData]):
             '{agent_name} run {prompt=}',
             prompt=user_prompt,
             agent=self,
-            mode_selection=mode_selection,
             model_name=model_used.name(),
             agent_name=self.name or 'agent',
         ) as run_span:
-            run_context = RunContext(deps, 0, [], None, model_used)
+            run_context = RunContext(deps, model_used, usage or result.Usage(), user_prompt)
             messages = await self._prepare_messages(user_prompt, message_history, run_context)
             run_context.messages = messages
 
             for tool in self._function_tools.values():
                 tool.current_retry = 0
 
-            usage = result.Usage(requests=0)
             model_settings = merge_model_settings(self.model_settings, model_settings)
             usage_limits = usage_limits or UsageLimits()
 
-            run_step = 0
             while True:
-                usage_limits.check_before_request(usage)
+                usage_limits.check_before_request(run_context.usage)
 
-                run_step += 1
-                with _logfire.span('preparing model and tools {run_step=}', run_step=run_step):
+                run_context.run_step += 1
+                with _logfire.span('preparing model and tools {run_step=}', run_step=run_context.run_step):
                     agent_model = await self._prepare_model(run_context)
 
-                with _logfire.span('model request', run_step=run_step) as model_req_span:
+                with _logfire.span('model request', run_step=run_context.run_step) as model_req_span:
                     model_response, request_usage = await agent_model.request(messages, model_settings)
                     model_req_span.set_attribute('response', model_response)
                     model_req_span.set_attribute('usage', request_usage)
 
                 messages.append(model_response)
-                usage += request_usage
-                usage.requests += 1
-                usage_limits.check_tokens(request_usage)
+                run_context.usage.incr(request_usage, requests=1)
+                usage_limits.check_tokens(run_context.usage)
 
-                with _logfire.span('handle model response', run_step=run_step) as handle_span:
+                with _logfire.span('handle model response', run_step=run_context.run_step) as handle_span:
                     final_result, tool_responses = await self._handle_model_response(model_response, run_context)
 
                     if tool_responses:
@@ -266,10 +273,10 @@ class Agent(Generic[AgentDeps, ResultData]):
                     if final_result is not None:
                         result_data = final_result.data
                         run_span.set_attribute('all_messages', messages)
-                        run_span.set_attribute('usage', usage)
+                        run_span.set_attribute('usage', run_context.usage)
                         handle_span.set_attribute('result', result_data)
                         handle_span.message = 'handle model response -> final result'
-                        return result.RunResult(messages, new_message_index, result_data, usage)
+                        return result.RunResult(messages, new_message_index, result_data, run_context.usage)
                     else:
                         # continue the conversation
                         handle_span.set_attribute('tool_responses', tool_responses)
@@ -285,6 +292,7 @@ class Agent(Generic[AgentDeps, ResultData]):
         deps: AgentDeps = None,
         model_settings: ModelSettings | None = None,
         usage_limits: UsageLimits | None = None,
+        usage: result.Usage | None = None,
         infer_name: bool = True,
     ) -> result.RunResult[ResultData]:
         """Run the agent with a user prompt synchronously.
@@ -311,6 +319,7 @@ class Agent(Generic[AgentDeps, ResultData]):
             deps: Optional dependencies to use for this run.
             model_settings: Optional settings to use for this model's request.
             usage_limits: Optional limits on model request count or token usage.
+            usage: Optional usage to start with, useful for resuming a conversation or agents used in tools.
             infer_name: Whether to try to infer the agent name from the call frame if it's not set.
 
         Returns:
@@ -326,6 +335,7 @@ class Agent(Generic[AgentDeps, ResultData]):
                 deps=deps,
                 model_settings=model_settings,
                 usage_limits=usage_limits,
+                usage=usage,
                 infer_name=False,
             )
         )
@@ -340,6 +350,7 @@ class Agent(Generic[AgentDeps, ResultData]):
         deps: AgentDeps = None,
         model_settings: ModelSettings | None = None,
         usage_limits: UsageLimits | None = None,
+        usage: result.Usage | None = None,
         infer_name: bool = True,
     ) -> AsyncIterator[result.StreamedRunResult[AgentDeps, ResultData]]:
         """Run the agent with a user prompt in async mode, returning a streamed response.
@@ -363,6 +374,7 @@ class Agent(Generic[AgentDeps, ResultData]):
             deps: Optional dependencies to use for this run.
             model_settings: Optional settings to use for this model's request.
             usage_limits: Optional limits on model request count or token usage.
+            usage: Optional usage to start with, useful for resuming a conversation or agents used in tools.
             infer_name: Whether to try to infer the agent name from the call frame if it's not set.
 
         Returns:
@@ -372,7 +384,7 @@ class Agent(Generic[AgentDeps, ResultData]):
             # f_back because `asynccontextmanager` adds one frame
             if frame := inspect.currentframe():  # pragma: no branch
                 self._infer_name(frame.f_back)
-        model_used, mode_selection = await self._get_model(model)
+        model_used = await self._get_model(model)
 
         deps = self._get_deps(deps)
         new_message_index = len(message_history) if message_history else 0
@@ -381,32 +393,29 @@ class Agent(Generic[AgentDeps, ResultData]):
             '{agent_name} run stream {prompt=}',
             prompt=user_prompt,
             agent=self,
-            mode_selection=mode_selection,
             model_name=model_used.name(),
             agent_name=self.name or 'agent',
         ) as run_span:
-            run_context = RunContext(deps, 0, [], None, model_used)
+            run_context = RunContext(deps, model_used, usage or result.Usage(), user_prompt)
             messages = await self._prepare_messages(user_prompt, message_history, run_context)
             run_context.messages = messages
 
             for tool in self._function_tools.values():
                 tool.current_retry = 0
 
-            usage = result.Usage()
             model_settings = merge_model_settings(self.model_settings, model_settings)
             usage_limits = usage_limits or UsageLimits()
 
-            run_step = 0
             while True:
-                run_step += 1
-                usage_limits.check_before_request(usage)
+                run_context.run_step += 1
+                usage_limits.check_before_request(run_context.usage)
 
-                with _logfire.span('preparing model and tools {run_step=}', run_step=run_step):
+                with _logfire.span('preparing model and tools {run_step=}', run_step=run_context.run_step):
                     agent_model = await self._prepare_model(run_context)
 
-                with _logfire.span('model request {run_step=}', run_step=run_step) as model_req_span:
+                with _logfire.span('model request {run_step=}', run_step=run_context.run_step) as model_req_span:
                     async with agent_model.request_stream(messages, model_settings) as model_response:
-                        usage.requests += 1
+                        run_context.usage.requests += 1
                         model_req_span.set_attribute('response_type', model_response.__class__.__name__)
                         # We want to end the "model request" span here, but we can't exit the context manager
                         # in the traditional way
@@ -442,7 +451,6 @@ class Agent(Generic[AgentDeps, ResultData]):
                                 yield result.StreamedRunResult(
                                     messages,
                                     new_message_index,
-                                    usage,
                                     usage_limits,
                                     result_stream,
                                     self._result_schema,
@@ -466,8 +474,8 @@ class Agent(Generic[AgentDeps, ResultData]):
                                 handle_span.message = f'handle model response -> {tool_responses_str}'
                                 # the model_response should have been fully streamed by now, we can add its usage
                                 model_response_usage = model_response.usage()
-                                usage += model_response_usage
-                                usage_limits.check_tokens(usage)
+                                run_context.usage.incr(model_response_usage)
+                                usage_limits.check_tokens(run_context.usage)
 
     @contextmanager
     def override(
@@ -778,14 +786,14 @@ class Agent(Generic[AgentDeps, ResultData]):
 
         self._function_tools[tool.name] = tool
 
-    async def _get_model(self, model: models.Model | models.KnownModelName | None) -> tuple[models.Model, str]:
+    async def _get_model(self, model: models.Model | models.KnownModelName | None) -> models.Model:
         """Create a model configured for this agent.
 
         Args:
             model: model to use for this run, required if `model` was not set when creating the agent.
 
         Returns:
-            a tuple of `(model used, how the model was selected)`
+            The model used
         """
         model_: models.Model
         if some_model := self._override_model:
@@ -796,18 +804,15 @@ class Agent(Generic[AgentDeps, ResultData]):
                     '(Even when `override(model=...)` is customizing the model that will actually be called)'
                 )
             model_ = some_model.value
-            mode_selection = 'override-model'
         elif model is not None:
             model_ = models.infer_model(model)
-            mode_selection = 'custom'
         elif self.model is not None:
             # noinspection PyTypeChecker
             model_ = self.model = models.infer_model(self.model)
-            mode_selection = 'from-agent'
         else:
             raise exceptions.UserError('`model` must be set either when creating the agent or when calling it.')
 
-        return model_, mode_selection
+        return model_
 
     async def _prepare_model(self, run_context: RunContext[AgentDeps]) -> models.AgentModel:
         """Build tools and create an agent model."""
@@ -830,15 +835,15 @@ class Agent(Generic[AgentDeps, ResultData]):
         self, user_prompt: str, message_history: list[_messages.ModelMessage] | None, run_context: RunContext[AgentDeps]
     ) -> list[_messages.ModelMessage]:
         try:
-            messages = _messages_ctx_var.get()
+            ctx_messages = _messages_ctx_var.get()
         except LookupError:
-            messages = []
+            messages: list[_messages.ModelMessage] = []
         else:
-            if messages:
-                raise exceptions.UserError(
-                    'The capture_run_messages() context manager may only be used to wrap '
-                    'one call to run(), run_sync(), or run_stream().'
-                )
+            if ctx_messages.used:
+                messages = []
+            else:
+                messages = ctx_messages.messages
+                ctx_messages.used = True
 
         if message_history:
             # shallow copy messages
@@ -1132,7 +1137,13 @@ class Agent(Generic[AgentDeps, ResultData]):
         raise AttributeError('The `last_run_messages` attribute has been removed, use `capture_run_messages` instead.')
 
 
-_messages_ctx_var: ContextVar[list[_messages.ModelMessage]] = ContextVar('var')
+@dataclasses.dataclass
+class _RunMessages:
+    messages: list[_messages.ModelMessage]
+    used: bool = False
+
+
+_messages_ctx_var: ContextVar[_RunMessages] = ContextVar('var')
 
 
 @contextmanager
@@ -1156,21 +1167,21 @@ def capture_run_messages() -> Iterator[list[_messages.ModelMessage]]:
     ```
 
     !!! note
-        You may not call `run`, `run_sync`, or `run_stream` more than once within a single `capture_run_messages` context.
-        If you try to do so, a [`UserError`][pydantic_ai.exceptions.UserError] will be raised.
+        If you call `run`, `run_sync`, or `run_stream` more than once within a single `capture_run_messages` context,
+        `messages` will represent the messages exchanged during the first call only.
     """
     try:
-        yield _messages_ctx_var.get()
+        yield _messages_ctx_var.get().messages
     except LookupError:
         messages: list[_messages.ModelMessage] = []
-        token = _messages_ctx_var.set(messages)
+        token = _messages_ctx_var.set(_RunMessages(messages))
         try:
             yield messages
         finally:
             _messages_ctx_var.reset(token)
 
 
-@dataclass
+@dataclasses.dataclass
 class _MarkFinalResult(Generic[ResultData]):
     """Marker class to indicate that the result is the final result.
 

pydantic_ai/models/gemini.py

@@ -444,7 +444,8 @@ def _content_model_response(m: ModelResponse) -> _GeminiContent:
         if isinstance(item, ToolCallPart):
             parts.append(_function_call_part_from_call(item))
         elif isinstance(item, TextPart):
-            parts.append(_GeminiTextPart(text=item.content))
+            if item.content:
+                parts.append(_GeminiTextPart(text=item.content))
         else:
             assert_never(item)
     return _GeminiContent(role='model', parts=parts)
@@ -701,7 +702,7 @@ class _GeminiJsonSchema:
 
     def _simplify(self, schema: dict[str, Any], refs_stack: tuple[str, ...]) -> None:
         schema.pop('title', None)
-        schema.pop('default', None)
+        default = schema.pop('default', _utils.UNSET)
         if ref := schema.pop('$ref', None):
             # noinspection PyTypeChecker
             key = re.sub(r'^#/\$defs/', '', ref)
@@ -714,8 +715,14 @@ class _GeminiJsonSchema:
             return
 
         if any_of := schema.get('anyOf'):
-            for schema in any_of:
-                self._simplify(schema, refs_stack)
+            for item_schema in any_of:
+                self._simplify(item_schema, refs_stack)
+            if len(any_of) == 2 and {'type': 'null'} in any_of and default is None:
+                for item_schema in any_of:
+                    if item_schema != {'type': 'null'}:
+                        schema.clear()
+                        schema.update(item_schema)
+                        return
 
         type_ = schema.get('type')
 

pydantic_ai/models/ollama.py

@@ -71,6 +71,7 @@ class OllamaModel(Model):
         model_name: OllamaModelName,
         *,
         base_url: str | None = 'http://localhost:11434/v1/',
+        api_key: str = 'ollama',
         openai_client: AsyncOpenAI | None = None,
         http_client: AsyncHTTPClient | None = None,
     ):
@@ -83,6 +84,8 @@ class OllamaModel(Model):
             model_name: The name of the Ollama model to use. List of models available [here](https://ollama.com/library)
                 You must first download the model (`ollama pull <MODEL-NAME>`) in order to use the model
             base_url: The base url for the ollama requests. The default value is the ollama default
+            api_key: The API key to use for authentication. Defaults to 'ollama' for local instances,
+                but can be customized for proxy setups that require authentication
             openai_client: An existing
                 [`AsyncOpenAI`](https://github.com/openai/openai-python?tab=readme-ov-file#async-usage)
                 client to use, if provided, `base_url` and `http_client` must be `None`.
@@ -96,7 +99,7 @@ class OllamaModel(Model):
         else:
             # API key is not required for ollama but a value is required to create the client
             http_client_ = http_client or cached_async_http_client()
-            oai_client = AsyncOpenAI(base_url=base_url, api_key='ollama', http_client=http_client_)
+            oai_client = AsyncOpenAI(base_url=base_url, api_key=api_key, http_client=http_client_)
             self.openai_model = OpenAIModel(model_name=model_name, openai_client=oai_client)
 
     async def agent_model(

pydantic_ai/models/test.py

@@ -16,6 +16,7 @@ from ..messages import (
     ModelMessage,
     ModelRequest,
     ModelResponse,
+    ModelResponsePart,
     RetryPromptPart,
     TextPart,
     ToolCallPart,
@@ -177,13 +178,23 @@ class TestAgentModel(AgentModel):
             # check if there are any retry prompts, if so retry them
             new_retry_names = {p.tool_name for p in last_message.parts if isinstance(p, RetryPromptPart)}
             if new_retry_names:
-                return ModelResponse(
-                    parts=[
-                        ToolCallPart.from_raw_args(name, self.gen_tool_args(args))
-                        for name, args in self.tool_calls
-                        if name in new_retry_names
-                    ]
-                )
+                # Handle retries for both function tools and result tools
+                # Check function tools first
+                retry_parts: list[ModelResponsePart] = [
+                    ToolCallPart.from_raw_args(name, self.gen_tool_args(args))
+                    for name, args in self.tool_calls
+                    if name in new_retry_names
+                ]
+                # Check result tools
+                if self.result_tools:
+                    retry_parts.extend(
+                        [
+                            ToolCallPart.from_raw_args(tool.name, self.gen_tool_args(tool))
+                            for tool in self.result_tools
+                            if tool.name in new_retry_names
+                        ]
+                    )
+                return ModelResponse(parts=retry_parts)
 
         if response_text := self.result.left:
             if response_text.value is None:

pydantic_ai/result.py

@@ -2,6 +2,7 @@ from __future__ import annotations as _annotations
 
 from abc import ABC, abstractmethod
 from collections.abc import AsyncIterator, Awaitable, Callable
+from copy import copy
 from dataclasses import dataclass, field
 from datetime import datetime
 from typing import Generic, Union, cast
@@ -63,25 +64,33 @@ class Usage:
     details: dict[str, int] | None = None
     """Any extra details returned by the model."""
 
-    def __add__(self, other: Usage) -> Usage:
-        """Add two Usages together.
+    def incr(self, incr_usage: Usage, *, requests: int = 0) -> None:
+        """Increment the usage in place.
 
-        This is provided so it's trivial to sum usage information from multiple requests and runs.
+        Args:
+            incr_usage: The usage to increment by.
+            requests: The number of requests to increment by in addition to `incr_usage.requests`.
         """
-        counts: dict[str, int] = {}
+        self.requests += requests
         for f in 'requests', 'request_tokens', 'response_tokens', 'total_tokens':
             self_value = getattr(self, f)
-            other_value = getattr(other, f)
+            other_value = getattr(incr_usage, f)
             if self_value is not None or other_value is not None:
-                counts[f] = (self_value or 0) + (other_value or 0)
+                setattr(self, f, (self_value or 0) + (other_value or 0))
 
-        details = self.details.copy() if self.details is not None else None
-        if other.details is not None:
-            details = details or {}
-            for key, value in other.details.items():
-                details[key] = details.get(key, 0) + value
+        if incr_usage.details:
+            self.details = self.details or {}
+            for key, value in incr_usage.details.items():
+                self.details[key] = self.details.get(key, 0) + value
 
-        return Usage(**counts, details=details or None)
+    def __add__(self, other: Usage) -> Usage:
+        """Add two Usages together.
+
+        This is provided so it's trivial to sum usage information from multiple requests and runs.
+        """
+        new_usage = copy(self)
+        new_usage.incr(other)
+        return new_usage
 
 
 @dataclass
@@ -136,8 +145,6 @@ class RunResult(_BaseRunResult[ResultData]):
 class StreamedRunResult(_BaseRunResult[ResultData], Generic[AgentDeps, ResultData]):
     """Result of a streamed run that returns structured data via a tool call."""
 
-    usage_so_far: Usage
-    """Usage of the run up until the last request."""
     _usage_limits: UsageLimits | None
     _stream_response: models.EitherStreamedResponse
     _result_schema: _result.ResultSchema[ResultData] | None
@@ -306,7 +313,7 @@ class StreamedRunResult(_BaseRunResult[ResultData], Generic[AgentDeps, ResultDat
         !!! note
             This won't return the full usage until the stream is finished.
         """
-        return self.usage_so_far + self._stream_response.usage()
+        return self._run_ctx.usage + self._stream_response.usage()
 
     def timestamp(self) -> datetime:
         """Get the timestamp of the response."""

pydantic_ai/settings.py

@@ -136,6 +136,6 @@ class UsageLimits:
                 f'Exceeded the response_tokens_limit of {self.response_tokens_limit} ({response_tokens=})'
             )
 
-        total_tokens = request_tokens + response_tokens
+        total_tokens = usage.total_tokens or 0
         if self.total_tokens_limit is not None and total_tokens > self.total_tokens_limit:
             raise UsageLimitExceeded(f'Exceeded the total_tokens_limit of {self.total_tokens_limit} ({total_tokens=})')

pydantic_ai/tools.py

@@ -4,7 +4,7 @@ import dataclasses
 import inspect
 from collections.abc import Awaitable
 from dataclasses import dataclass, field
-from typing import Any, Callable, Generic, TypeVar, Union, cast
+from typing import TYPE_CHECKING, Any, Callable, Generic, TypeVar, Union, cast
 
 from pydantic import ValidationError
 from pydantic_core import SchemaValidator
@@ -13,6 +13,9 @@ from typing_extensions import Concatenate, ParamSpec, TypeAlias
 from . import _pydantic, _utils, messages as _messages, models
 from .exceptions import ModelRetry, UnexpectedModelBehavior
 
+if TYPE_CHECKING:
+    from .result import Usage
+
 __all__ = (
     'AgentDeps',
     'RunContext',
@@ -37,14 +40,20 @@ class RunContext(Generic[AgentDeps]):
 
     deps: AgentDeps
     """Dependencies for the agent."""
-    retry: int
-    """Number of retries so far."""
-    messages: list[_messages.ModelMessage]
-    """Messages exchanged in the conversation so far."""
-    tool_name: str | None
-    """Name of the tool being called."""
     model: models.Model
     """The model used in this run."""
+    usage: Usage
+    """LLM usage associated with the run."""
+    prompt: str
+    """The original user prompt passed to the run."""
+    messages: list[_messages.ModelMessage] = field(default_factory=list)
+    """Messages exchanged in the conversation so far."""
+    tool_name: str | None = None
+    """Name of the tool being called."""
+    retry: int = 0
+    """Number of retries so far."""
+    run_step: int = 0
+    """The current step in the run."""
 
     def replace_with(
         self, retry: int | None = None, tool_name: str | None | _utils.Unset = _utils.UNSET

{pydantic_ai_slim-0.0.15.dist-info → pydantic_ai_slim-0.0.16.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pydantic-ai-slim
-Version: 0.0.15
+Version: 0.0.16
 Summary: Agent Framework / shim to use Pydantic with LLMs, slim package
 Author-email: Samuel Colvin <samuel@pydantic.dev>
 License-Expression: MIT

{pydantic_ai_slim-0.0.15.dist-info → pydantic_ai_slim-0.0.16.dist-info}/RECORD

@@ -1,26 +1,26 @@
 pydantic_ai/__init__.py,sha256=FbYetEgT6OO25u2KF5ZnFxKpz5DtnSpfckRXP4mjl8E,489
-pydantic_ai/_griffe.py,sha256=pRjCJ6B1hhx6k46XJgl9zF6aRYxRmqEZKFok8unp4Iw,3449
+pydantic_ai/_griffe.py,sha256=Wqk3AuyeWuPwE5s1GbMeCsERelx1B4QcU9uYZSoko8s,3409
 pydantic_ai/_pydantic.py,sha256=qXi5IsyiYOHeg_-qozCdxkfeqw2z0gBTjqgywBCiJWo,8125
 pydantic_ai/_result.py,sha256=cUSugZQV0n5Z4fFHiMqua-2xs_0S6m-rr-yd6QS3nFE,10317
 pydantic_ai/_system_prompt.py,sha256=MZJWksIoS5GM3Au5lznlcQnC-h7eqwtE7oI5WFgRcOg,1090
 pydantic_ai/_utils.py,sha256=skWNgm89US_x1EpxdRy5wCkghBrm1XgxFCiEh6wAkAo,8753
-pydantic_ai/agent.py,sha256=qa3Ox5pXEDzxcTJgwN0gebV37qQKizVc0PW-1q5MMn4,51662
+pydantic_ai/agent.py,sha256=NJTcPSlqb4Fd-x9pDPuoXGCwFGF1GHcHevutoB0Busw,52333
 pydantic_ai/exceptions.py,sha256=eGDKX6bGhgVxXBzu81Sk3iiAkXr0GUtgT7bD5Rxlqpg,2028
 pydantic_ai/messages.py,sha256=ImbWY8Ft3mxInUQ08EmIWywf4nJBvTiJhmsECRYDkSQ,8968
 pydantic_ai/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-pydantic_ai/result.py,sha256=n2cEFwm8WhFzHuT6KRhZ2itQVPShGUd7ECbOPmRIIoM,15335
-pydantic_ai/settings.py,sha256=R71rBg2u2SgjxKWcpUdSzm7icV5_apF3b0BlBqa2lpA,4927
-pydantic_ai/tools.py,sha256=wNYzfdp1XjIVw_8bqh5GP3x3k_12gHTC74HWNvHAAwI,11447
+pydantic_ai/result.py,sha256=LbZVHZnJnQwgegSz5PtwS9r_ifrJnLRpsa9xjYnHg1g,15549
+pydantic_ai/settings.py,sha256=W8krcFsujjhE03rwckrz39F4Dz_9RwdBSeEF3izK0-Y,4918
+pydantic_ai/tools.py,sha256=mnh3Lvs0Ri0FkqpV1MUooExNN4epTCcBKw6DyZvNSQ8,11745
 pydantic_ai/models/__init__.py,sha256=XHt02IDQAircb-lEkIbIcuabSAIh5_UKnz2V1xN0Glw,10926
 pydantic_ai/models/anthropic.py,sha256=EUZgmvT0jhMDbooBp_jfW0z2cM5jTMuAhVws1XKgaNs,13451
 pydantic_ai/models/function.py,sha256=i7qkS_31aHrTbYVh6OzQ7Cwucz44F5PjT2EJK3GMphw,10573
-pydantic_ai/models/gemini.py,sha256=8vdcW4izL9NUGFj6lcD9yIPaakCtsmHauTvKwlTzD14,28207
+pydantic_ai/models/gemini.py,sha256=Sr19D2hN8iEAcoLlzv5883pto90TgEr_xiGlV8hMOwA,28572
 pydantic_ai/models/groq.py,sha256=ZoPkuWJrf78JPnTRfZhi7v0ETgxJKNN5dH8BLWagGGk,15770
 pydantic_ai/models/mistral.py,sha256=xGVI6-b8-9vnFickPPI2cRaHEWLc0jKKUM_vMjipf-U,25894
-pydantic_ai/models/ollama.py,sha256=i3mMXkXu9xL6f4c52Eyx3j4aHKfYoloFondlGHPtkS4,3971
+pydantic_ai/models/ollama.py,sha256=ELqxhcNcnvQBnadd3gukS01zprUp6v8N_h1P5K-uf6c,4188
 pydantic_ai/models/openai.py,sha256=qFFInL3NbgfGcsAWigxMP5mscp76hC-jJimHc9woU6Y,16518
-pydantic_ai/models/test.py,sha256=pty5qaudHsSDvdE89HqMj-kmd4UMV9VJI2YGtdfOX1o,15960
+pydantic_ai/models/test.py,sha256=u2pdZd9OLXQ_jI6CaVt96udXuIcv0Hfnfqd3pFGmeJM,16514
 pydantic_ai/models/vertexai.py,sha256=DBCBfpvpIhZaMG7cKvRl5rugCZqJqqEFm74uBc45weo,9259
-pydantic_ai_slim-0.0.15.dist-info/METADATA,sha256=CM_cQ6RRb9PFJVXKVa0JIVsp_bCrdCKWGnEu0KBiD0c,2730
-pydantic_ai_slim-0.0.15.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-pydantic_ai_slim-0.0.15.dist-info/RECORD,,
+pydantic_ai_slim-0.0.16.dist-info/METADATA,sha256=4udd7j2erIuMC0ekYgmgQAqsKfhA5sLsKzTcD_QyOeo,2730
+pydantic_ai_slim-0.0.16.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+pydantic_ai_slim-0.0.16.dist-info/RECORD,,