added fastapi

.gitignore

@@ -173,4 +173,4 @@ cython_debug/
 
 # PyPI configuration file
 .pypirc
-
+.idea/**

main.py

@@ -1,29 +1,105 @@
 import asyncio
+import json
+import urllib.request
+from contextlib import asynccontextmanager
+from typing import Optional
+
+from fastapi import FastAPI
+from pydantic import BaseModel
 
 from mcp_agent.app import MCPApp
 from mcp_agent.agents.agent import Agent
 from mcp_agent.workflows.llm.augmented_llm_openai import OpenAIAugmentedLLM
 
 
-app = MCPApp(name="weather_agent")
+OLLAMA_MODEL = "qwen3:1.7b"
+OLLAMA_KEEP_ALIVE = "10h"
 
-
-async def main():
-    async with app.run():
-        agent = Agent(
-            name="weather",
-            instruction=(
+AGENT_INSTRUCTION = (
     "Ты ассистент по погоде. "
     "Если нужны актуальные данные — используй доступные инструменты."
-            ),
-            server_names=["weather"],
+    "ОТВЕЧАЙ ТОЛЬКО НА РУССКОМ ЯЗЫКЕ!"
+    "Не выводи блоки <think> или внутренние рассуждения, отвечай только итогом."
 )
 
-        async with agent:
-            llm = await agent.attach_llm(OpenAIAugmentedLLM)
-            answer = await llm.generate_str("Какая сейчас точная погода в Берлине?")
-            print(answer)
+MCP_SERVER_NAMES = ["weather"]
 
 
-if __name__ == "__main__":
-    asyncio.run(main())
+def warmup_ollama(model: str, keep_alive: str = "10h") -> None:
+    payload = {
+        "model": model,
+        "prompt": "",
+        "stream": False,
+        "keep_alive": keep_alive,
+    }
+    req = urllib.request.Request(
+        "http://localhost:11434/api/generate",
+        data=json.dumps(payload).encode("utf-8"),
+        headers={"Content-Type": "application/json"},
+        method="POST",
+    )
+    try:
+        urllib.request.urlopen(req, timeout=5).read()
+    except Exception:
+        pass
+
+
+class AskRequest(BaseModel):
+    question: str
+
+
+class AskResponse(BaseModel):
+    answer: str
+
+
+app_mcp = MCPApp(name="weather_agent")
+_agent: Optional[Agent] = None
+_llm: Optional[OpenAIAugmentedLLM] = None
+_lock: asyncio.Lock = asyncio.Lock()
+
+
+@asynccontextmanager
+async def lifespan(app: FastAPI):
+    """
+    Startup:
+      - прогреваем Ollama
+      - поднимаем MCPApp
+      - создаём Agent и подключаем LLM один раз
+    Shutdown:
+      - корректно закрываем agent и MCPApp
+    """
+    global _agent, _llm
+
+    warmup_ollama(OLLAMA_MODEL, keep_alive=OLLAMA_KEEP_ALIVE)
+
+    async with app_mcp.run():
+        _agent = Agent(
+            name="weather",
+            instruction=AGENT_INSTRUCTION,
+            server_names=MCP_SERVER_NAMES,
+        )
+
+        async with _agent:
+            _llm = await _agent.attach_llm(OpenAIAugmentedLLM)
+            yield
+
+api = FastAPI(title="Weather MCP Agent API", lifespan=lifespan)
+
+
+@api.post("/ask", response_model=AskResponse)
+async def ask(req: AskRequest) -> AskResponse:
+    """
+    Принимает вопрос, возвращает краткий ответ на русском.
+    """
+    if _llm is None:
+        return AskResponse(answer="Сервис не готов. Попробуйте позже.")
+
+    async with _lock:
+        answer = await _llm.generate_str(req.question)
+
+    return AskResponse(answer=answer)
+
+
+@api.get("/health")
+async def health():
+    return {"status": "ok"}

mcp_agent.config.yaml

@@ -8,4 +8,4 @@ mcp:
 
 openai:
   base_url: "http://localhost:11434/v1"
-  default_model: "qwen3:1.7b"
+  default_model: "llama3.2:1b"