Struct AgentSession 

pub struct AgentSession {
    pub total_tool_calls: usize,
    pub nudge_count: usize,
    pub turn_tool_calls: usize,
    /* private fields */
}

Fields

total_tool_calls: usize

Total number of tool calls made in this session (across all turns). Used by middleware for decisions like “first_turn_only” enforcement.

nudge_count: usize

Number of tool-enforcement nudges issued in the current turn. Reset to 0 at the start of each turn (when a new user message arrives). Used by ToolEnforcementMiddleware to cap nudge attempts per turn.

turn_tool_calls: usize

Number of tool calls already executed in the current turn. Reset to 0 at the start of each turn (when a new user message arrives). Used by TurnToolLimitMiddleware to enforce per-turn tool call limits.

Implementations

impl AgentSession

pub fn new(id: SessionId) -> Self

pub fn id(&self) -> Option<SessionId>

pub fn simple_messages(&self) -> Vec<Message>

Derive a simplified Vec<Message> view from the canonical chat_messages. Assistant messages that contain only tool_calls (no text content) are skipped, since they have no corresponding simplified representation.

pub fn chat_messages(&self) -> &[ChatMessage]

pub fn chat_messages_mut(&mut self) -> &mut Vec<ChatMessage>

可变引用，仅用于需要直接操作消息的高级场景。

pub fn is_action_allowed(&self, action_key: &str) -> bool

pub fn allow_action(&mut self, action_key: impl Into<String>)

pub fn push_message(&mut self, role: MessageRole, content: impl Into<String>)

pub fn push_assistant_with_reasoning( &mut self, content: impl Into<String>, reasoning: impl Into<String>, )

Push an assistant message with reasoning/thinking content preserved. This allows the LLM to see its own prior reasoning in subsequent turns, preventing it from re-deriving the same conclusions every turn.

pub fn push_user_message_with_images( &mut self, content: impl Into<String>, images: Vec<ImageAttachment>, )

pub fn push_assistant_tool_call( &mut self, tool_call_id: &str, tool_name: &str, arguments_json: &str, )

pub fn push_assistant_tool_calls( &mut self, tool_calls: &[(String, String, String)], reasoning: Option<String>, )

pub fn push_tool_result( &mut self, tool_call_id: &str, content: impl Into<String>, )

pub fn remove_ephemeral_messages(&mut self)

移除所有临时消息（ephemeral=true）。

在 turn 结束时调用，确保注入的临时内容不残留到下一轮。

pub fn turn_count(&self) -> usize

Count the number of conversation turns. A turn starts with a User message and includes subsequent Assistant/Tool messages.

pub fn trim_oldest_turns(&mut self, max_turns: usize)

Remove the oldest turns from the front until turn count ≤ max_turns. Preserves the System message at index 0 if present.

pub fn pop_last_message(&mut self)

Remove the last message from chat_messages. Used by the max_message_tokens safety valve to discard oversized messages.

pub fn close_dangling_tool_calls(&mut self, error_summary: &str)

pub fn set_chat_messages( &mut self, messages: Vec<ChatMessage>, ) -> Result<(), String>

Replace chat messages — only for persistence restore. Validates message sequence before replacing.

仅供持久化恢复使用。调用方必须保证 messages 序列合法。

Trait Implementations

impl Clone for AgentSession

fn clone(&self) -> AgentSession

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for AgentSession

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for AgentSession

fn default() -> AgentSession

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for AgentSession

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for AgentSession

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.