fix: inject s09 memories as turn context

.gitignore

@@ -192,6 +192,7 @@ cython_debug/
 .transcripts/
 
 # Runtime artifacts (generated by agent tests)
+.memory/
 .task_outputs/
 .tasks/
 .teams/

s09_memory/README.md

@@ -23,7 +23,7 @@ LLM 没有持久状态，所有信息都在上下文窗口里。上下文满了
 
 s08 的压缩管线保留，聚焦记忆。存储选文件系统：`.memory/` 目录下，每个记忆一个 `.md` 文件，带 YAML frontmatter（`name` / `description` / `type`）。文件多了需要索引：`MEMORY.md` 一行一个链接，注入 SYSTEM。
 
-关键设计：索引常驻 SYSTEM prompt（可被 prompt cache 缓存），文件内容按需注入（按 filename/description 匹配当前对话，不破坏 cache）。写入分两条路径：用户显式说"记住"，或者每轮结束后后台提取。文件积累多了，定期整理去重。
+关键设计：索引常驻 SYSTEM prompt（可被 prompt cache 缓存），文件内容按需注入到当前 user turn（按 filename/description 匹配当前对话，不破坏 cache）。写入由每轮结束后的提取器完成：用户显式说"记住"或表达稳定偏好时，提取器会保存为记忆。文件积累多了，定期整理去重。
 
 四类记忆，各有用途：
 
@@ -78,7 +78,7 @@ def write_memory_file(name, mem_type, description, body):
 
 **路径一：索引常驻 SYSTEM。** `build_system()` 每轮重建 SYSTEM 时读取 `MEMORY.md`，把记忆清单注入。SYSTEM prompt 中的索引可以被 prompt cache 缓存，不需要每轮重新发送。
 
-**路径二：相关记忆按需注入。** 每轮调用前，`load_memories()` 把最近对话和记忆目录（name + description）一起发给 LLM 做一次轻量 side-query，选出相关的文件名，再读文件内容注入上下文。最多 5 条，控制开销。
+**路径二：相关记忆按需注入。** 每轮调用前，`load_memories()` 把最近对话和记忆目录（name + description）一起发给 LLM 做一次轻量 side-query，选出相关的文件名，再读文件内容临时注入到当前 user turn。最多 5 条，控制开销。
 
 ```python
 def select_relevant_memories(messages, max_items=5):
@@ -93,7 +93,8 @@ def select_relevant_memories(messages, max_items=5):
         "content": f"Select relevant memory indices. Return JSON array.\n\n"
                    f"Recent conversation:\n{recent}\n\nMemory catalog:\n{catalog}"}],
         max_tokens=200)
-    indices = json.loads(re.search(r'\[.*?\]', response.content[0].text).group())
+    text = extract_text(response.content).strip()
+    indices = json.loads(re.search(r'\[.*?\]', text).group())
     return [files[i]["filename"] for i in indices if 0 <= i < len(files)]
 ```
 
@@ -108,7 +109,7 @@ def select_relevant_memories(messages, max_items=5):
 ```python
 # In agent_loop:
 if response.stop_reason != "tool_use":
-    extract_memories(messages)   # 从最近对话提取新记忆
+    extract_memories(pre_compress)   # 从压缩前快照提取新记忆
     consolidate_memories()       # 检查是否需要整理
     return
 ```

s09_memory/code.py

@@ -177,7 +177,7 @@ def select_relevant_memories(messages: list, max_items: int = 5) -> list[str]:
             messages=[{"role": "user", "content": prompt}],
             max_tokens=200,
         )
-        text = response.content[0].text.strip()
+        text = extract_text(response.content).strip()
         # Extract JSON array from response
         match = re.search(r'\[.*?\]', text, re.DOTALL)
         if match:
@@ -259,7 +259,7 @@ def extract_memories(messages: list):
         response = client.messages.create(
             model=MODEL, messages=[{"role": "user", "content": prompt}], max_tokens=800
         )
-        text = response.content[0].text.strip()
+        text = extract_text(response.content).strip()
         # Extract JSON array from response
         match = re.search(r'\[.*\]', text, re.DOTALL)
         if not match:
@@ -309,7 +309,7 @@ def consolidate_memories():
         response = client.messages.create(
             model=MODEL, messages=[{"role": "user", "content": prompt}], max_tokens=3000
         )
-        text = response.content[0].text.strip()
+        text = extract_text(response.content).strip()
         match = re.search(r'\[.*\]', text, re.DOTALL)
         if not match:
             return
@@ -504,7 +504,7 @@ def summarize_history(msgs):
         "Summarize this coding-agent conversation so work can continue.\n"
         "Preserve: 1. current goal, 2. key findings, 3. files changed, 4. remaining work, 5. user constraints.\n\n" + conv}],
         max_tokens=2000)
-    return r.content[0].text.strip()
+    return extract_text(r.content).strip()
 
 def compact_history(msgs):
     write_transcript(msgs)
@@ -550,12 +550,12 @@ MAX_REACTIVE_RETRIES = 1
 
 def agent_loop(messages: list):
     reactive_retries = 0
+    # s09: inject relevant memory content into the current user turn
+    memories_content = load_memories(messages)
+    memory_turn = len(messages) - 1 if messages and isinstance(messages[-1].get("content"), str) else None
     while True:
-        # s09: rebuild system with current memory index + relevant memories
+        # s09: rebuild system with current memory index
         system = build_system()
-        memories_content = load_memories(messages)
-        if memories_content:
-            system += "\n\n" + memories_content
 
         # s09: save pre-compression snapshot for accurate memory extraction
         pre_compress = [m if isinstance(m, dict) else {"role": m.get("role",""),
@@ -571,8 +571,15 @@ def agent_loop(messages: list):
             messages[:] = compact_history(messages)
 
         try:
+            request_messages = messages
+            if memories_content and memory_turn is not None and memory_turn < len(messages):
+                request_messages = messages.copy()
+                request_messages[memory_turn] = {
+                    **messages[memory_turn],
+                    "content": memories_content + "\n\n" + messages[memory_turn]["content"],
+                }
             response = client.messages.create(
-                model=MODEL, system=system, messages=messages, tools=TOOLS, max_tokens=8000
+                model=MODEL, system=system, messages=request_messages, tools=TOOLS, max_tokens=8000
             )
             reactive_retries = 0
         except Exception as e: