refactor(core): 扫清 v0.1 Phase A 技术债

- 修复测试编译回归：补全 session.rs/cycle.rs 测试模块导入；
convert.rs 2 处 irrefutable if let 改为 let
- composer.rs 迁移至 IR：OpenaiChatMessage → Message，
ContentField/OpenaiContentPart → ContentBlock；删除 set_message_name
和 build_request；developer 消息映射为 Message::System
- knowledge.rs 锁修复：std::sync::Mutex → tokio::sync::Mutex；
search() 优化锁粒度（锁内仅 clone IDs，避免锁内异步 IO）
- 标记 ChatResponse / ToolDefinition 为废弃（#[deprecated(since = "0.1.0")]），
内部使用点加 #[allow(deprecated)] 抑制警告
- clippy 清零：合并冗余 if、手动 strip_prefix 改 strip_prefix、
多处 dead_code 抑制、测试代码清理

src/llm/convert.rs

@@ -40,15 +40,14 @@ pub fn from_openai(msg: &OpenaiChatMessage) -> Message {
             let mut blocks = content_to_blocks(content);
             if let Some(calls) = tool_calls {
                 for call in calls {
-                    if let OpenaiToolCall::Function { id, function } = call {
-                        let input: Value =
-                            serde_json::from_str(&function.arguments).unwrap_or(Value::Null);
-                        blocks.push(ContentBlock::ToolUse {
-                            id: id.clone(),
-                            name: function.name.clone(),
-                            input,
-                        });
-                    }
+                    let OpenaiToolCall::Function { id, function } = call;
+                    let input: Value =
+                        serde_json::from_str(&function.arguments).unwrap_or(Value::Null);
+                    blocks.push(ContentBlock::ToolUse {
+                        id: id.clone(),
+                        name: function.name.clone(),
+                        input,
+                    });
                 }
             }
             Message::Assistant { content: blocks }
@@ -254,8 +253,7 @@ pub fn blocks_to_content(blocks: &[ContentBlock]) -> ContentField {
 #[cfg(test)]
 mod tests {
     use super::*;
-    use crate::llm::types::message::ImageSource;
-    use crate::llm::types::shared::{AudioFormat, ImageDetail};
+    use crate::llm::types::shared::ImageDetail;
     use serde_json::json;
 
     #[test]
@@ -364,13 +362,10 @@ mod tests {
                 assert!(matches!(content, ContentField::String(_)));
                 let calls = tool_calls.expect("tool_calls");
                 assert_eq!(calls.len(), 1);
-                if let OpenaiToolCall::Function { id, function } = &calls[0] {
-                    assert_eq!(id, "call_1");
-                    assert_eq!(function.name, "search");
-                    assert!(function.arguments.contains("rust"));
-                } else {
-                    panic!("expected Function variant");
-                }
+                let OpenaiToolCall::Function { id, function } = &calls[0];
+                assert_eq!(id, "call_1");
+                assert_eq!(function.name, "search");
+                assert!(function.arguments.contains("rust"));
             }
             _ => panic!("expected Assistant"),
         }

src/llm/cycle.rs

@@ -21,6 +21,7 @@ use crate::llm::stream::StreamEvent;
 use crate::llm::types::message::{ContentBlock, Message};
 use crate::llm::types::request_v2::MessageRequest;
 use crate::llm::types::response_v2::{MessageResponse, StopReason};
+#[allow(deprecated)]
 use crate::llm::types::{ToolChoice, ToolDefinition};
 
 /// LLM 调用周期配置。
@@ -79,6 +80,7 @@ pub struct LlmCycle {
     compact_state: CompactState,
 }
 
+#[allow(deprecated)]
 impl LlmCycle {
     /// 创建一个新的 LlmCycle（持有 `Box<dyn LlmProvider>` 的独占所有权）。
     ///
@@ -175,13 +177,10 @@ impl LlmCycle {
         messages: Vec<Message>,
         tools: Vec<ToolDefinition>,
     ) -> Result<MessageResponse, LlmError> {
-        let ir_messages: Vec<Message> = messages;
-        let ir_tools: Vec<ToolDefinition> = tools.clone();
-
         let request = MessageRequest {
             model: self.config.model.clone(),
-            messages: ir_messages,
-            tools: ir_tools,
+            messages,
+            tools,
             tool_choice: ToolChoice::Auto,
             max_tokens: self.config.max_tokens,
             temperature: self.config.temperature,
@@ -672,6 +671,7 @@ fn truncate_tool_result(s: &str, max_bytes: usize) -> String {
 #[cfg(test)]
 mod tests {
     use super::*;
+    use crate::llm::provider::{ProviderCapabilities, ProviderFeatures};
     use crate::tools::{BaseTool, ToolRegistry};
     use async_trait::async_trait;
     use futures_core::Stream;
@@ -858,8 +858,10 @@ mod tests {
 
     #[tokio::test]
     async fn test_submit_with_tools_max_turns_exceeded() {
-        let mut config = CycleConfig::default();
-        config.max_tool_turns = Some(2);
+        let config = CycleConfig {
+            max_tool_turns: Some(2),
+            ..Default::default()
+        };
         let responses = vec![
             assistant_tool_call_response(vec![("c1", "add", r#"{"a":1,"b":1}"#)]),
             assistant_tool_call_response(vec![("c2", "add", r#"{"a":1,"b":1}"#)]),

src/llm/hooks.rs

@@ -75,6 +75,7 @@ impl<'a> HookContext<'a> {
     }
 
     /// 设置 plan step 序号（仅 OnPlanStepComplete 使用，Phase 4b 新增）。
+    #[allow(dead_code)]
     pub(crate) fn with_plan_step_index(mut self, plan_step_index: usize) -> Self {
         self.plan_step_index = Some(plan_step_index);
         self

src/llm/provider/anthropic.rs

@@ -36,6 +36,7 @@ const DEFAULT_MAX_TOKENS: u32 = 4096;
 pub struct AnthropicProvider {
     http_client: Client,
     base_url: String,
+    #[allow(dead_code)]
     api_key: String,
     model: String,
 }
@@ -586,6 +587,7 @@ enum AnthropicDelta {
 struct AnthropicMessageDeltaInner {
     stop_reason: Option<String>,
     #[serde(default)]
+    #[allow(dead_code)]
     stop_sequence: Option<String>,
 }
 
@@ -599,6 +601,7 @@ pub struct AnthropicSseStream {
     chunks: Pin<Box<dyn Stream<Item = Result<Bytes, LlmError>> + Send>>,
     buffer: String,
     partial: PartialMessageResponse,
+    #[allow(dead_code)]
     next_block_index: u32,
     saw_terminal: bool,
 }

src/llm/provider/openai.rs

@@ -720,8 +720,8 @@ impl Stream for ChunkToEventStream {
                 }
                 let data = if let Some(p) = trimmed.strip_prefix("data: ") {
                     p
-                } else if trimmed.starts_with("data:") {
-                    &trimmed[5..]
+                } else if let Some(p) = trimmed.strip_prefix("data:") {
+                    p
                 } else {
                     continue;
                 };

src/llm/types/mod.rs

@@ -40,6 +40,7 @@ pub use usage::{CompletionTokensDetails, CostTracker, PromptTokensDetails, Usage
 // Phase 1 起移除 `ChatRequest` 别名 —— 新代码统一使用 `MessageRequest`（v2 IR）。
 // `ChatResponse` 结构体仍存在，作为 OpenAI `chat_inner()` 内部 wire-format 转换目标。
 /// 旧 wire-format 响应结构（保留用于 OpenAI 内部转换层）。
+#[deprecated(since = "0.1.0", note = "请改用 MessageResponse")]
 #[derive(Debug, Clone)]
 pub struct ChatResponse {
     pub message: OpenaiChatMessage,
@@ -47,6 +48,7 @@ pub struct ChatResponse {
     pub stop_reason: Option<FinishReason>,
 }
 
+#[allow(deprecated)]
 impl From<OpenaiChatResponse> for ChatResponse {
     fn from(response: OpenaiChatResponse) -> Self {
         let message = response
@@ -63,6 +65,7 @@ impl From<OpenaiChatResponse> for ChatResponse {
     }
 }
 
+#[allow(deprecated)]
 impl From<ChatResponse> for OpenaiChatChunk {
     fn from(response: ChatResponse) -> Self {
         let delta = Delta::from(response.message.clone());
@@ -95,4 +98,5 @@ impl From<ChatResponse> for OpenaiChatChunk {
 }
 
 /// 工具定义别名（无新类型冲突，保留）。
+#[deprecated(since = "0.1.0", note = "ToolDefinition 仍直接对应 OpenAI wire-format；未来 v0.2 引入 IR 工具类型后会再次更新")]
 pub type ToolDefinition = OpenaiToolDefinition;