add fastapi_mcp

mcp/scripts/ask.py

@@ -1,55 +1,198 @@
-import os
+## scripts/ask.py
+import sys
 import json
-import httpx
-import openai
+import requests
+from config import load_config
+from datetime import datetime, timezone
 
-from context_loader import load_context_from_repo
-from prompt_template import PROMPT_TEMPLATE
+def build_payload_openai(cfg, message: str):
+    return {
+        "model": cfg["model"],
+        "tools": [
+            {
+                "type": "function",
+                "function": {
+                    "name": "ask_message",
+                    "description": "過去の記憶を検索します",
+                    "parameters": {
+                        "type": "object",
+                        "properties": {
+                            "query": {
+                                "type": "string",
+                                "description": "検索したい語句"
+                            }
+                        },
+                        "required": ["query"]
+                    }
+                }
+            }
+        ],
+        "tool_choice": "auto",
+        "messages": [
+            {"role": "system", "content": "あなたは親しみやすいAIで、必要に応じて記憶から情報を検索して応答します。"},
+            {"role": "user", "content": message}
+        ]
+    }
 
-PROVIDER = os.getenv("PROVIDER", "ollama")  # "ollama" or "openai"
-
-# Ollama用
-OLLAMA_HOST = os.getenv("OLLAMA_HOST", "http://localhost:11434")
-OLLAMA_URL = f"{OLLAMA_HOST}/api/generate"
-OLLAMA_MODEL = os.getenv("MODEL", "syui/ai")
-
-# OpenAI用
-OPENAI_BASE = os.getenv("OPENAI_API_BASE", "https://api.openai.com/v1")
-OPENAI_KEY = os.getenv("OPENAI_API_KEY", "")
-OPENAI_MODEL = os.getenv("MODEL", "gpt-4o-mini")
-
-def ask_question(question, repo_path="."):
-    context = load_context_from_repo(repo_path)
-    prompt = PROMPT_TEMPLATE.format(context=context[:10000], question=question)
-
-    if PROVIDER == "ollama":
-        payload = {
-            "model": OLLAMA_MODEL,
-            "prompt": prompt,
-            "stream": False
+def build_payload_mcp(message: str):
+    return {
+        "tool": "ask_message",  # MCPサーバー側で定義されたツール名
+        "input": {
+            "message": message
         }
-        response = httpx.post(OLLAMA_URL, json=payload, timeout=60.0)
-        result = response.json()
-        return result.get("response", "返答がありませんでした。")
+    }
 
-    elif PROVIDER == "openai":
-        import openai
-        openai.api_key = OPENAI_KEY
-        openai.api_base = OPENAI_BASE
+def build_payload_openai(cfg, message: str):
+    return {
+        "model": cfg["model"],
+        "messages": [
+            {"role": "system", "content": "あなたは思いやりのあるAIです。"},
+            {"role": "user", "content": message}
+        ],
+        "temperature": 0.7
+    }
 
-        client = openai.OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
-        response = client.chat.completions.create(
-            model=OPENAI_MODEL,
-            messages=[{"role": "user", "content": prompt}]
-        )
-        return response.choices[0].message.content
+def call_mcp(cfg, message: str):
+    payload = build_payload_mcp(message)
+    headers = {"Content-Type": "application/json"}
+    response = requests.post(cfg["url"], headers=headers, json=payload)
+    response.raise_for_status()
+    return response.json().get("output", {}).get("response", "❓ 応答が取得できませんでした")
 
+def call_openai(cfg, message: str):
+    # ツール定義
+    tools = [
+        {
+            "type": "function",
+            "function": {
+                "name": "memory",
+                "description": "記憶を検索する",
+                "parameters": {
+                    "type": "object",
+                    "properties": {
+                        "query": {
+                            "type": "string",
+                            "description": "検索する語句"
+                        }
+                    },
+                    "required": ["query"]
+                }
+            }
+        }
+    ]
+
+    # 最初のメッセージ送信
+    payload = {
+        "model": cfg["model"],
+        "messages": [
+            {"role": "system", "content": "あなたはAIで、必要に応じてツールmemoryを使って記憶を検索します。"},
+            {"role": "user", "content": message}
+        ],
+        "tools": tools,
+        "tool_choice": "auto"
+    }
+
+    headers = {
+        "Authorization": f"Bearer {cfg['api_key']}",
+        "Content-Type": "application/json",
+    }
+
+    res1 = requests.post(cfg["url"], headers=headers, json=payload)
+    res1.raise_for_status()
+    result = res1.json()
+
+    # 🧠 tool_call されたか確認
+    if "tool_calls" in result["choices"][0]["message"]:
+        tool_call = result["choices"][0]["message"]["tool_calls"][0]
+        if tool_call["function"]["name"] == "memory":
+            args = json.loads(tool_call["function"]["arguments"])
+            query = args.get("query", "")
+            print(f"🛠️ ツール実行: memory(query='{query}')")
+
+            # MCPエンドポイントにPOST
+            memory_res = requests.post("http://127.0.0.1:5000/memory/search", json={"query": query})
+            memory_json = memory_res.json()
+            tool_output = memory_json.get("result", "なし")
+
+            # tool_outputをAIに返す
+            followup = {
+                "model": cfg["model"],
+                "messages": [
+                    {"role": "system", "content": "あなたはAIで、必要に応じてツールmemoryを使って記憶を検索します。"},
+                    {"role": "user", "content": message},
+                    {"role": "assistant", "tool_calls": result["choices"][0]["message"]["tool_calls"]},
+                    {"role": "tool", "tool_call_id": tool_call["id"], "name": "memory", "content": tool_output}
+                ]
+            }
+
+            res2 = requests.post(cfg["url"], headers=headers, json=followup)
+            res2.raise_for_status()
+            final_response = res2.json()
+            return final_response["choices"][0]["message"]["content"]
+            #print(tool_output)
+            #print(cfg["model"])
+            #print(final_response)
+
+    # ツール未使用 or 通常応答
+    return result["choices"][0]["message"]["content"]
+
+def call_ollama(cfg, message: str):
+    payload = {
+            "model": cfg["model"],
+            "prompt": message,  # `prompt` → `message` にすべき（変数未定義エラー回避）
+            "stream": False
+            }
+    headers = {"Content-Type": "application/json"}
+    response = requests.post(cfg["url"], headers=headers, json=payload)
+    response.raise_for_status()
+    return response.json().get("response", "❌ 応答が取得できませんでした")
+def main():
+    if len(sys.argv) < 2:
+        print("Usage: ask.py 'your message'")
+        return
+
+    message = sys.argv[1]
+    cfg = load_config()
+
+    print(f"🔍 使用プロバイダー: {cfg['provider']}")
+
+    try:
+        if cfg["provider"] == "openai":
+            response = call_openai(cfg, message)
+        elif cfg["provider"] == "mcp":
+            response = call_mcp(cfg, message)
+        elif cfg["provider"] == "ollama":
+            response = call_ollama(cfg, message)
+        else:
+            raise ValueError(f"未対応のプロバイダー: {cfg['provider']}")
+
+        print("💬 応答:")
+        print(response)
+
+        # ログ保存（オプション）
+        save_log(message, response)
+
+    except Exception as e:
+        print(f"❌ 実行エラー: {e}")
+
+def save_log(user_msg, ai_msg):
+    from config import MEMORY_DIR
+    date_str = datetime.now().strftime("%Y-%m-%d")
+    path = MEMORY_DIR / f"{date_str}.json"
+    path.parent.mkdir(parents=True, exist_ok=True)
+
+    if path.exists():
+        with open(path, "r") as f:
+            logs = json.load(f)
     else:
-        return f"❌ 未知のプロバイダです: {PROVIDER}"
+        logs = []
 
+    now = datetime.now(timezone.utc).isoformat()
+    logs.append({"timestamp": now, "sender": "user", "message": user_msg})
+    logs.append({"timestamp": now, "sender": "ai", "message": ai_msg})
+
+    with open(path, "w") as f:
+        json.dump(logs, f, indent=2, ensure_ascii=False)
 
 if __name__ == "__main__":
-    import sys
-    question = " ".join(sys.argv[1:])
-    answer = ask_question(question)
-    print("\n🧠 回答:\n", answer)
+    main()