Make input/new items available in the run context

src/agents/result.py

@@ -8,6 +8,8 @@
 
 from typing_extensions import TypeVar
 
+from agents.run_context import RunContextWrapper
+
 from ._run_impl import QueueCompleteSentinel
 from .agent import Agent
 from .agent_output import AgentOutputSchemaBase
@@ -50,6 +52,9 @@ class RunResultBase(abc.ABC):
     output_guardrail_results: list[OutputGuardrailResult]
     """Guardrail results for the final output of the agent."""
 
+    context_wrapper: RunContextWrapper[Any]
+    """The context wrapper that was used to run the agent."""
+
     @property
     @abc.abstractmethod
     def last_agent(self) -> Agent[Any]:

src/agents/run.py

@@ -170,6 +170,7 @@ async def run(
 
             context_wrapper: RunContextWrapper[TContext] = RunContextWrapper(
                 context=context,  # type: ignore
+                _input=copy.deepcopy(input),
             )
 
             input_guardrail_results: list[InputGuardrailResult] = []
@@ -255,6 +256,9 @@ async def run(
                     original_input = turn_result.original_input
                     generated_items = turn_result.generated_items
 
+                    context_wrapper._input = copy.deepcopy(original_input)
+                    context_wrapper._new_items = copy.deepcopy(generated_items)
+
                     if isinstance(turn_result.next_step, NextStepFinalOutput):
                         output_guardrail_results = await cls._run_output_guardrails(
                             current_agent.output_guardrails + (run_config.output_guardrails or []),
@@ -270,6 +274,7 @@ async def run(
                             _last_agent=current_agent,
                             input_guardrail_results=input_guardrail_results,
                             output_guardrail_results=output_guardrail_results,
+                            context_wrapper=context_wrapper,
                         )
                     elif isinstance(turn_result.next_step, NextStepHandoff):
                         current_agent = cast(Agent[TContext], turn_result.next_step.new_agent)
@@ -407,7 +412,8 @@ def run_streamed(
 
         output_schema = cls._get_output_schema(starting_agent)
         context_wrapper: RunContextWrapper[TContext] = RunContextWrapper(
-            context=context  # type: ignore
+            context=context,  # type: ignore
+            _input=copy.deepcopy(input),
         )
 
         streamed_result = RunResultStreaming(
@@ -423,6 +429,7 @@ def run_streamed(
             output_guardrail_results=[],
             _current_agent_output_schema=output_schema,
             trace=new_trace,
+            context_wrapper=context_wrapper,
         )
 
         # Kick off the actual agent loop in the background and return the streamed result object.
@@ -576,6 +583,9 @@ async def _run_streamed_impl(
                     streamed_result.input = turn_result.original_input
                     streamed_result.new_items = turn_result.generated_items
 
+                    context_wrapper._input = copy.deepcopy(streamed_result.input)
+                    context_wrapper._new_items = copy.deepcopy(streamed_result.new_items)
+
                     if isinstance(turn_result.next_step, NextStepHandoff):
                         current_agent = turn_result.next_step.new_agent
                         current_span.finish(reset_current=True)

src/agents/run_context.py

@@ -1,8 +1,12 @@
+from __future__ import annotations
+
+import copy
 from dataclasses import dataclass, field
 from typing import Any, Generic
 
 from typing_extensions import TypeVar
 
+from .items import RunItem, TResponseInputItem
 from .usage import Usage
 
 TContext = TypeVar("TContext", default=Any)
@@ -24,3 +28,21 @@ class RunContextWrapper(Generic[TContext]):
     """The usage of the agent run so far. For streamed responses, the usage will be stale until the
     last chunk of the stream is processed.
     """
+
+    _new_items: list[RunItem] = field(default_factory=list, repr=False)
+    """The new items created during the agent run."""
+
+    _input: str | list[TResponseInputItem] = field(default_factory=list, repr=False)
+    """The original input that you passed to `Runner.run()`."""
+
+    @property
+    def input(self) -> str | list[TResponseInputItem]:
+        """The original input that you passed to `Runner.run()`."""
+
+        return copy.deepcopy(self._input)
+
+    @property
+    def new_items(self) -> list[RunItem]:
+        """The new items created during the agent run."""
+
+        return copy.deepcopy(self._new_items)