add chat function

cloud/sam/Makefile

@@ -3,11 +3,13 @@ check:
 	 --config-env $(ENV)\
 	 --no-execute-changeset\
 	 --no-fail-on-empty-changeset\
+	 --resolve-image-repos\
 	 --parameter-overrides FrontendOrigin=$(TETRIS_FRONT_ORIGIN) SQSUrl=$(SQS_URL) DynamoDBNewsTableName=tetris_news_table DynamoDBTrainingTableName=tetris_training_table TetrisTrainingBucketName=$(TETRIS_TRAINING_BUCKET_NAME)
 
 deploy:
 	sam deploy\
 	 --config-env $(ENV)\
 	 --no-confirm-changeset\
 	 --no-fail-on-empty-changeset\
+	 --resolve-image-repos\
 	 --parameter-overrides FrontendOrigin=$(TETRIS_FRONT_ORIGIN) SQSUrl=$(SQS_URL) DynamoDBNewsTableName=tetris_news_table DynamoDBTrainingTableName=tetris_training_table TetrisTrainingBucketName=$(TETRIS_TRAINING_BUCKET_NAME)

cloud/sam/scripts/chat_function/Dockerfile

@@ -0,0 +1,10 @@
+FROM public.ecr.aws/lambda/python:3.9
+
+COPY requirements.txt ./
+RUN python3.9 -m pip install --upgrade pip
+RUN python3.9 -m pip install -r requirements.txt -t .
+
+COPY . .
+
+# Command can be overwritten by providing a different command in the template directly.
+CMD ["app.lambda_handler"]

cloud/sam/scripts/chat_function/app.py

@@ -0,0 +1,141 @@
+import json
+from tools.utils import DynamoDBManager, OpenAIManager
+from tools.response import HttpResponse
+from role.role_tetris import TetrisAssistant
+import logging
+
+# ロギングの設定
+logger = logging.getLogger()
+logger.setLevel(logging.INFO)
+
+class ChatHandler:
+
+    def __init__(self, event):
+        self.event = event
+        self.db_manager = DynamoDBManager()
+        self.openai_manager = OpenAIManager()
+        self.http_response = HttpResponse()
+        self.tetris_assistant = TetrisAssistant()
+
+    def get_message_from_event(self):
+        body_content = self.event.get('body', None)
+        if not body_content:
+            raise ValueError("The 'body' field in the event is missing or empty.")
+        try:
+            return json.loads(body_content)['input_text']
+        except KeyError:
+            raise ValueError("Invalid input. 'input_text' key is required.")
+
+    def handle_get_request(self):
+        return self.http_response.success('hello world')
+
+    def handle_delete_request(self):
+        # RESTfulな設計では、DELETEはbodyを持たせるべきではないが、他に方法が分からなかった。
+        try:
+            data = json.loads(self.event["body"])
+            user_id = data['identity_id']
+            char_name = data['character_name']
+            try:
+                self.db_manager.delete_items_with_secondary_index(user_id, char_name)
+                return self.http_response.success('delete success')
+
+            except Exception as e:
+                print(f"An error occurred: {e}")
+                return self.http_response.server_error(f'Error during delete operation: {e}')
+        except KeyError as e:
+            print(f"An error occurred: {e}")
+            return self.http_response.client_error(f'Error during delete operation: {e}')
+
+
+    def gpt_function_call(self, response_data, messages, functions, user_id, char_name, max_order_id, input_text):
+        print(f"function call defined\n")
+        # gptが定義した関数を実行し、結果を取得する
+        func_name, args, function_response, function_args = self.openai_manager.execute_function_call(response_data)
+
+        # 2回目のAPI実行のための関数の引数を作成
+        function_args = self.openai_manager.create_function_args(func_name, args)
+        #print(f"function_args: {function_args}")
+        messages.append({"role": "assistant", "content": None, "function_call": function_args})
+        messages.append({"role": "function", "content": function_response, "name": func_name})
+
+        response_2nd = self.openai_manager.get_chat_response_func(messages, functions)
+        response_content = response_2nd.choices[0]["message"]["content"]
+
+        # DynamoDBにトーク履歴を記録
+        self.db_manager.store_conversation(user_id, char_name, max_order_id + 0, "user", input_text)
+        self.db_manager.store_conversation(user_id, char_name, max_order_id + 1, "assistant", None, name=None, function_call=function_args)
+        self.db_manager.store_conversation(user_id, char_name, max_order_id + 2, "function", function_response, name=func_name, function_call=None)
+        self.db_manager.store_conversation(user_id, char_name, max_order_id + 3, "assistant", response_content)
+
+        return response_content
+
+
+    def gpt_simple_response(self, response_data, user_id, char_name, max_order_id, input_text):
+        print(f"function call undefined\n")
+        response_content = response_data["message"]["content"]
+        self.db_manager.store_conversation(user_id, char_name, max_order_id + 0, "user", input_text)
+        self.db_manager.store_conversation(user_id, char_name, max_order_id + 1, "assistant", response_content)
+        return response_content
+
+
+    def call_openai_api(self, messages, functions, user_id, char_name, input_text, max_order_id):
+        response_1st = self.openai_manager.get_chat_response_func(messages, functions)
+        response_data = response_1st["choices"][0]
+        if response_data["finish_reason"] == "function_call":
+            if response_data["message"]["function_call"]["name"]:
+                return self.gpt_function_call(response_data, messages, functions, user_id, char_name, max_order_id, input_text)
+        else:
+            return self.gpt_simple_response(response_data, user_id, char_name, max_order_id, input_text)
+
+
+    def handle_post_request(self):
+        self.openai_manager.get_secret()
+
+        # メッセージを取得
+        try:
+            data = json.loads(self.event["body"])
+            logger.info('Event: %s', json.dumps(data))
+            user_id = data['identity_id']
+            char_name = data['character_name']
+            input_text = data['input_text']
+        except ValueError as e:
+            return self.http_response.client_error(f'Error during post operation: {e}')
+
+        # 過去の応答を取得
+        try:
+            messages = self.tetris_assistant.get_chat_messages()
+            functions = self.tetris_assistant.get_chat_functions()
+            max_order_id,items = self.db_manager.get_max_conversation_id(user_id, char_name)
+
+            #今までの対話をmessagesに並べる
+            messages.extend([
+                {
+                    "role": item["role"],
+                    "content": item["content"],
+                    **({"name": item["name"]} if "name" in item else {}),
+                    **({"function_call": item["function_call"]} if "function_call" in item else {})
+                } 
+                for item in items
+            ])
+            messages.append({"role": "user", "content": data['input_text']})
+
+            #openAIのAPIを叩く
+            response_content = self.call_openai_api(messages, functions, user_id, char_name, input_text, max_order_id)
+            return self.http_response.success(response_content)
+
+        except Exception as e:
+            return self.http_response.server_error(f'Error during post operation: {e}')
+
+    def handle(self):
+        http_method = self.event.get('httpMethod', '')
+        if http_method == 'GET':
+            return self.handle_get_request()
+        elif http_method == 'DELETE':
+            return self.handle_delete_request()
+        elif http_method == 'POST':
+            return self.handle_post_request()
+
+
+def lambda_handler(event, context):
+    handler = ChatHandler(event)
+    return handler.handle()

cloud/sam/scripts/chat_function/events/event_delete.json

@@ -0,0 +1,6 @@
+{
+  "httpMethod": "DELETE",
+  "body": "{\"identity_id\": \"id_hoge\", \"character_name\": \"test_user\"}",
+  "resource": "/ask",
+  "path": "/ask"
+}

cloud/sam/scripts/chat_function/events/event_get.json

@@ -0,0 +1,6 @@
+{
+    "httpMethod": "GET",
+    "resource": "/ask",
+    "path": "/ask"
+}
+

cloud/sam/scripts/chat_function/events/event_post.json

@@ -0,0 +1,6 @@
+{
+  "httpMethod": "POST",
+  "body": "{\"identity_id\": \"id_hoge\", \"input_text\": \"Rustはどんな言語ですか？\", \"character_name\": \"test_user\"}",
+  "resource": "/ask",
+  "path": "/ask"
+}

cloud/sam/scripts/chat_function/events/event_post_func.json

@@ -0,0 +1,7 @@
+{
+    "httpMethod": "POST",
+    "body": "{\"identity_id\": \"id_hoge\", \"input_text\": \"ミノを消した時の点数は？\", \"character_name\": \"test_user\"}",
+    "resource": "/ask",
+    "path": "/ask"
+  }
+

cloud/sam/scripts/chat_function/events/test.bash

@@ -0,0 +1 @@
+sam local invoke "ChatFunction" -e scripts/chat_function/events/event_get.json

cloud/sam/scripts/chat_function/local.py

@@ -0,0 +1,41 @@
+import os
+import openai
+
+from dotenv import load_dotenv
+from llama_index import VectorStoreIndex, SimpleDirectoryReader
+from llama_index import StorageContext, load_index_from_storage
+
+# ローカルで辞書を作成し、storageに保存する
+# .envにopenaiのAPIkeyを記述する
+# ./dataに.txtや.md等のテキストファイルを格納し、ローカル環境で実行する
+# ./storageに辞書データを出力し、そのデータを用いて辞書検索を行う
+def main():
+    # APIkeyの設定
+    load_dotenv()
+    try:
+        OPENAI_API_KEY = os.environ["OPENAI_API_KEY"]
+    except KeyError:
+        print("OPENAI_API_KEY environment variable not found. Please make sure it is set.")
+    openai.api_key = OPENAI_API_KEY
+
+    # モデルの読み込み
+    if(1):
+        # 辞書データを作成する
+        documents = SimpleDirectoryReader(input_dir="./data").load_data()
+        print("documents: ", documents)
+        index = VectorStoreIndex.from_documents(documents)
+        # 保存
+        index.storage_context.persist()
+    else:
+        # rebuild storage context
+        storage_context = StorageContext.from_defaults(persist_dir='storage')
+        # load index
+        index = load_index_from_storage(storage_context)
+
+    # クエリの実行
+    query_engine = index.as_query_engine()
+    response = query_engine.query("Dockerイメージの取得途中で止まる原因は?")
+    print("response: ", response)
+
+if __name__ == "__main__":
+    main()

cloud/sam/scripts/chat_function/requirements.txt

@@ -0,0 +1,6 @@
+requests
+urllib3==1.26
+langchain==0.0.234
+llama-index==0.7.9
+openai==0.27.8
+python-dotenv

cloud/sam/scripts/chat_function/role/role_tetris.py

@@ -0,0 +1,68 @@
+import json
+from langchain import OpenAI
+from llama_index import VectorStoreIndex, SimpleDirectoryReader
+from llama_index import StorageContext, load_index_from_storage
+
+class TetrisAssistant:
+    @staticmethod
+    def get_chat_messages():
+        return [
+            {
+                "role": "system",
+                "content":
+"""
+質問には100文字以内で回答する。
+必要に応じて"function"を実行し、その応答を用いて、"user"の言語に合わせた言語で100文字以内で回答する。
+"""
+            },
+            {
+                "role": "user",
+                "content":"テトリスのルールを教えて。"
+            },
+            {
+                "role": "assistant",
+                "content": "テトリスは、異なる形のブロックを落としてきて、行を埋めるゲームです。行を完全に埋めると、その行は消えます。ブロックが画面上部に積み上げられるとゲームオーバーです。"
+            }
+        ]
+
+    @staticmethod
+    def get_chat_functions():
+        return [
+            {
+                "name": "search_tetris_index",
+                "description": 
+"""
+テトリスをpythonで操作することを通してプログラミングを学ぶ時に用いる資料を検索する。
+この資料は以下の内容を含む。
+・テトリスのルール（点数,ルール,ボード情報,フィールド情報）
+・プログラムによるテトリスの操作方法（実行コマンド,ミノの説明,コマンド、各種ファイル,アートの作り方）
+・環境構築（AI,docker,Git,Windows(WSL,PowerShell),Linux,Mac）
+""",
+                "parameters": {
+                    "type": "object",
+                    "properties": {
+                        "SearchContent": {
+                            "type": "string",
+                            "description": "テトリス対戦に関する情報を検索 e.g.ミノを消した時の点数は？"
+                        }
+                    },
+                    "required": ["SearchContent"]
+                }
+            }
+        ]
+
+class TetrisIndexSearch:
+    def __init__(self):
+        # indexの読み込み
+        storage_context = StorageContext.from_defaults(persist_dir='./storage')
+        self.index = load_index_from_storage(storage_context)
+
+    def search_tetris_index(self, SearchContent):
+        # クエリの実行
+        query_engine = self.index.as_query_engine()
+        response = query_engine.query(SearchContent)
+
+        result = {
+            "response": response.response
+        }
+        return json.dumps(result)

cloud/sam/scripts/chat_function/storage/graph_store.json

@@ -0,0 +1 @@
+{"graph_dict": {}}

cloud/sam/scripts/chat_function/tools/config.py

@@ -0,0 +1,5 @@
+import os
+
+DYNAMODB_TABLE_NAME = 'tetris_chat_log'
+OPENAI_MODEL_NAME = "gpt-3.5-turbo-0613"#gpt-3.5-turbo-0613"
+DYNAMODB_INDEX_NAME = 'user_id-char_name-index'

cloud/sam/scripts/chat_function/tools/response.py

@@ -0,0 +1,28 @@
+import json
+
+class HttpResponse:
+
+    @staticmethod
+    def generate_response(status_code, message):
+        return {
+            "statusCode": status_code,
+            "body": json.dumps({
+                "message": message,
+            }),
+        }
+
+    @classmethod
+    def success(cls, message):
+        return cls.generate_response(200, message)
+
+    @classmethod
+    def redirect(cls, message):
+        return cls.generate_response(300, message)
+
+    @classmethod
+    def client_error(cls, message):
+        return cls.generate_response(400, message)
+
+    @classmethod
+    def server_error(cls, message):
+        return cls.generate_response(500, message)