refactor: 重构 Agent 模块为 AgentScope，删除旧版 CrewAI/LiteLLM 实现

backend/src/v1/agent/dependencies.py

@@ -13,7 +13,7 @@ from core.agentscope.runtime.tasks import (
     run_command_task_bulk,
     run_command_task_critical,
 )
-from core.agent.infrastructure.storage.tool_result_storage import (
+from core.agentscope.tools.tool_result_storage import (
     create_tool_result_storage,
 )
 from core.config.settings import config

backend/src/v1/agent/router.py

@@ -15,7 +15,8 @@ from fastapi import HTTPException
 from fastapi.responses import JSONResponse, StreamingResponse
 
 from core.agentscope.events import to_sse_event
-from core.agent.domain.agui_input import (
+from core.agentscope.schemas.agui_input import (
+    extract_latest_tool_result,
     parse_run_input,
     validate_run_request_messages_contract,
 )
@@ -29,6 +30,7 @@ from v1.users.dependencies import get_current_user
 router = APIRouter(prefix="/agent", tags=["agent"])
 _LAST_EVENT_ID_RE = re.compile(r"^\d+-\d+$")
 _RUNS_PER_MINUTE = 30
+_TRANSCRIBES_PER_MINUTE = 20
 _MAX_SSE_CONNECTIONS_PER_USER = 3
 _SSE_SLOT_TTL_SECONDS = 15 * 60
 _MAX_TRANSCRIBE_AUDIO_BYTES = 10 * 1024 * 1024
@@ -61,6 +63,19 @@ async def _allow_run_request(*, user_id: str) -> bool:
         return False
 
 
+async def _allow_transcribe_request(*, user_id: str) -> bool:
+    try:
+        redis = await get_or_init_redis_client()
+        minute_bucket = int(time.time() // 60)
+        key = f"agent:transcribe-rate:{user_id}:{minute_bucket}"
+        count = await redis.incr(key)
+        if count == 1:
+            await redis.expire(key, 70)
+        return int(count) <= _TRANSCRIBES_PER_MINUTE
+    except Exception:  # noqa: BLE001
+        return False
+
+
 async def _acquire_sse_slot(*, user_id: str) -> bool:
     try:
         redis = await get_or_init_redis_client()
@@ -130,9 +145,13 @@ async def enqueue_resume(
     if request.thread_id != thread_id:
         raise HTTPException(status_code=422, detail="thread_id path/body mismatch")
     try:
-        parse_run_input(request.model_dump(mode="json", by_alias=True))
+        normalized = parse_run_input(request.model_dump(mode="json", by_alias=True))
+        extract_latest_tool_result(normalized)
     except ValueError as exc:
         raise HTTPException(status_code=422, detail=str(exc)) from exc
+    allowed = await _allow_run_request(user_id=str(current_user.id))
+    if not allowed:
+        raise HTTPException(status_code=429, detail="Too many run requests")
     task = await service.enqueue_resume(
         thread_id=thread_id,
         run_input=request,
@@ -240,9 +259,12 @@ async def transcribe(
     request: Request,
     current_user: Annotated[CurrentUser, Depends(get_current_user)],
 ) -> Union[AsrTranscribeResponse, JSONResponse]:
-    del current_user
     temp_path: str | None = None
     try:
+        allowed = await _allow_transcribe_request(user_id=str(current_user.id))
+        if not allowed:
+            raise HTTPException(status_code=429, detail="Too many transcribe requests")
+
         if audio.content_type not in _ALLOWED_AUDIO_CONTENT_TYPES:
             raise ValueError("Unsupported audio format")