ninja移行

config/settings.py

@@ -39,7 +39,6 @@ INSTALLED_APPS = [
     "django.contrib.sessions",
     "django.contrib.messages",
     "django.contrib.staticfiles",
-    "rest_framework",
     "ml_api",
 ]
 
@@ -106,9 +105,9 @@ AUTH_PASSWORD_VALIDATORS = [
 # Internationalization
 # https://docs.djangoproject.com/en/6.0/topics/i18n/
 
-LANGUAGE_CODE = "en-us"
+LANGUAGE_CODE = "ja"
 
-TIME_ZONE = "UTC"
+TIME_ZONE = "Asia/Tokyo"
 
 USE_I18N = True
 

config/urls.py

@@ -16,6 +16,11 @@ Including another URLconf
 """
 
 from django.contrib import admin
-from django.urls import include, path
+from django.urls import path
 
-urlpatterns = [path("admin/", admin.site.urls), path("api/", include("ml_api.urls"))]
+from ml_api.api import api
+
+urlpatterns = [
+    path("admin/", admin.site.urls),
+    path("api/", api.urls),  # ここでマウント)
+]

ml_api/api.py

@@ -0,0 +1,55 @@
+import torch
+from ninja import NinjaAPI
+
+from .model_loader import get_model
+from .schemas import ChatInput, ChatOutput
+
+# APIインスタンスの作成
+api = NinjaAPI()
+
+
+@api.post("/chat", response=ChatOutput)
+def chat(request, data: ChatInput):
+    """
+    Qwenモデルを使用したチャットAPI
+    """
+    user_input = data.text  # Schema経由で安全にアクセス
+
+    # モデルのロード（初回のみロードが走る）
+    model, tokenizer = get_model()
+
+    # 1. 会話フォーマットの作成
+    messages = [
+        {
+            "role": "system",
+            "content": "あなたは親切でフレンドリーなAIアシスタント「qwen」です。自然な日本語で簡潔に返事をしてください。",
+        },
+        {"role": "user", "content": user_input},
+    ]
+
+    # 2. プロンプトへの変換
+    text = tokenizer.apply_chat_template(
+        messages, tokenize=False, add_generation_prompt=True
+    )
+
+    inputs = tokenizer([text], return_tensors="pt").to(model.device)
+
+    # 3. 生成
+    with torch.no_grad():
+        generated_ids = model.generate(
+            **inputs,
+            max_new_tokens=256,
+            do_sample=True,
+            temperature=0.7,
+            top_p=0.9,
+        )
+
+    # 4. デコード
+    generated_ids = [
+        output_ids[len(input_ids) :]
+        for input_ids, output_ids in zip(inputs.input_ids, generated_ids)
+    ]
+    response_text = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+
+    # ChatOutputスキーマに合わせてdictを返す
+    return {"result": response_text}

ml_api/schemas.py

@@ -0,0 +1,9 @@
+from ninja import Schema
+
+
+class ChatInput(Schema):
+    text: str
+
+
+class ChatOutput(Schema):
+    result: str

ml_api/tests.py

@@ -1,3 +1 @@
-from django.test import TestCase
-
 # Create your tests here.

ml_api/urls.py

@@ -1,9 +0,0 @@
-from django.urls import path
-
-from .views import ChatView
-
-urlpatterns = [
-    # エンドポイント名を 'chat' とします
-    # アクセスURLの末尾になります
-    path("chat/", ChatView.as_view(), name="chat"),
-]

ml_api/views.py

@@ -1,86 +0,0 @@
-import torch
-from rest_framework.response import Response
-from rest_framework.views import APIView
-
-from ml_api.model_loader import get_model
-
-
-class ChatView(APIView):
-    def post(self, request):
-        user_input = request.data.get("text", "")
-
-        # ここで呼び出す（初回のみロードが走る）
-        model, tokenizer = get_model()
-
-        # 1. 会話フォーマットの作成
-        messages = [
-            {
-                "role": "system",
-                "content": "あなたは親切でフレンドリーなAIアシスタント「qwen」です。。自然な日本語で簡潔に返事をしてください。",
-            },
-            {"role": "user", "content": user_input},
-        ]
-        print(user_input)
-
-        # 2. プロンプトへの変換
-        text = tokenizer.apply_chat_template(
-            messages, tokenize=False, add_generation_prompt=True
-        )
-
-        inputs = tokenizer([text], return_tensors="pt").to(model.device)
-
-        # 3. 生成
-        with torch.no_grad():
-            generated_ids = model.generate(
-                **inputs,
-                max_new_tokens=256,
-                do_sample=True,
-                temperature=0.7,
-                top_p=0.9,
-            )
-
-        # 4. デコード
-        generated_ids = [
-            output_ids[len(input_ids) :]
-            for input_ids, output_ids in zip(inputs.input_ids, generated_ids)
-        ]
-        response_text = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[
-            0
-        ]
-
-        return Response({"result": response_text})
-
-
-# class ChatView(APIView):
-#     def post(self, request):
-#         input_text = request.data.get("text", "")
-
-#         # 簡易的なプロンプトエンジニアリング
-#         # モデルに「会話」であることを認識させるフォーマット
-#         prompt = f"ユーザー: {input_text}\nシステム: "
-
-#         app_config = apps.get_app_config("ml_api")
-#         tokenizer = app_config.tokenizer
-#         model = app_config.model
-
-#         # トークン化
-#         inputs = tokenizer(prompt, return_tensors="pt", add_special_tokens=False)
-
-#         # 生成
-#         with torch.no_grad():
-#             output_ids = model.generate(
-#                 inputs["input_ids"],
-#                 max_new_tokens=50,  # 返信の長さ
-#                 do_sample=True,
-#                 temperature=0.7,  # 創造性（高いほどランダム）
-#                 pad_token_id=tokenizer.pad_token_id,
-#                 eos_token_id=tokenizer.eos_token_id,
-#             )
-
-#         # デコード
-#         output = tokenizer.decode(output_ids.tolist()[0])
-
-#         # プロンプト部分を除去して返信部分だけ抽出
-#         response_text = output.split("システム: ")[-1].strip()
-
-#         return Response({"result": response_text})

pyproject.toml

@@ -7,7 +7,6 @@ requires-python = ">=3.12"
 dependencies = [
     "accelerate>=1.12.0",
     "django>=6.0",
-    "djangorestframework>=3.16.1",
     "protobuf>=6.33.2",
     "sentencepiece>=0.2.1",
     "torch>=2.9.1",

requirements.txt

@@ -9,10 +9,6 @@ certifi==2025.11.12
 charset-normalizer==3.4.4
     # via requests
 django==6.0
-    # via
-    #   django-ai-chat (pyproject.toml)
-    #   djangorestframework
-djangorestframework==3.16.1
     # via django-ai-chat (pyproject.toml)
 filelock==3.20.0
     # via

uv.lock

@@ -125,7 +125,6 @@ source = { virtual = "." }
 dependencies = [
     { name = "accelerate" },
     { name = "django" },
-    { name = "djangorestframework" },
     { name = "protobuf" },
     { name = "sentencepiece" },
     { name = "torch" },
@@ -136,25 +135,12 @@ dependencies = [
 requires-dist = [
     { name = "accelerate", specifier = ">=1.12.0" },
     { name = "django", specifier = ">=6.0" },
-    { name = "djangorestframework", specifier = ">=3.16.1" },
     { name = "protobuf", specifier = ">=6.33.2" },
     { name = "sentencepiece", specifier = ">=0.2.1" },
     { name = "torch", specifier = ">=2.9.1" },
     { name = "transformers", specifier = ">=4.57.3" },
 ]
 
-[[package]]
-name = "djangorestframework"
-version = "3.16.1"
-source = { registry = "https://pypi.org/simple" }
-dependencies = [
-    { name = "django" },
-]
-sdist = { url = "https://files.pythonhosted.org/packages/8a/95/5376fe618646fde6899b3cdc85fd959716bb67542e273a76a80d9f326f27/djangorestframework-3.16.1.tar.gz", hash = "sha256:166809528b1aced0a17dc66c24492af18049f2c9420dbd0be29422029cfc3ff7", size = 1089735, upload-time = "2025-08-06T17:50:53.251Z" }
-wheels = [
-    { url = "https://files.pythonhosted.org/packages/b0/ce/bf8b9d3f415be4ac5588545b5fcdbbb841977db1c1d923f7568eeabe1689/djangorestframework-3.16.1-py3-none-any.whl", hash = "sha256:33a59f47fb9c85ede792cbf88bde71893bcda0667bc573f784649521f1102cec", size = 1080442, upload-time = "2025-08-06T17:50:50.667Z" },
-]
-
 [[package]]
 name = "filelock"
 version = "3.20.0"