docs: clarify ModelSettings.truncation behavior

src/agents/agent.py

@@ -38,13 +38,14 @@
 @dataclass
 class ToolsToFinalOutputResult:
     is_final_output: bool
-    """Whether this is the final output. If False, the LLM will run again and receive the tool call
-    output.
+    """If True, this is the final output;
+    if False, LLM will run again and receive the tool call output.
     """
 
     final_output: Any | None = None
-    """The final output. Can be None if `is_final_output` is False, otherwise must match the
-    `output_type` of the agent.
+    """The final output value.
+    - Must be non-None and match the agent's `output_type` if `is_final_output` is True.
+    - Must be None if `is_final_output` is False.
     """
 
 

src/agents/model_settings.py

@@ -93,8 +93,16 @@ class ModelSettings:
 
     truncation: Literal["auto", "disabled"] | None = None
     """The truncation strategy to use when calling the model.
+
+    - "auto": If the context (this response + previous ones) exceeds the model's
+    context window, older input items in the *middle* of the conversation will
+    be dropped so the request still fits. Streaming continues normally.
+    - "disabled" (default): Do not truncate. If input exceeds the model's
+    context window, the request fails with HTTP 400 (no streaming).
+    - None: Defer to the provider's default behavior (typically "disabled").
+
     See [Responses API documentation](https://platform.openai.com/docs/api-reference/responses/create#responses_create-truncation)
-    for more details.
+    for details.
     """
 
     max_tokens: int | None = None