diff --git a/examples/example.pkl b/examples/example.pkl
index eecd3ec98..f1912a973 100644
Binary files a/examples/example.pkl and b/examples/example.pkl differ
diff --git a/tests/data/rsp_cache.json b/tests/data/rsp_cache.json
index b83222120..2d43e20c5 100644
--- a/tests/data/rsp_cache.json
+++ b/tests/data/rsp_cache.json
@@ -215,5 +215,15 @@
"You are a python code to Mermaid Sequence Diagram translator in function detail#SYSTEM_MSG_END#```python\n#!/usr/bin/env python\n# -*- coding: utf-8 -*-\n\"\"\"\n@Author : alexanderwu\n@File : write_review.py\n\"\"\"\nimport asyncio\nfrom typing import List, Literal\n\nfrom metagpt.actions import Action\nfrom metagpt.actions.action_node import ActionNode\n\nREVIEW = ActionNode(\n key=\"Review\",\n expected_type=List[str],\n instruction=\"Act as an experienced reviewer and critically assess the given output. Provide specific and\"\n \" constructive feedback, highlighting areas for improvement and suggesting changes.\",\n example=[\n \"The logic in the function `calculate_total` seems flawed. Shouldn't it consider the discount rate as well?\",\n \"The TODO function is not implemented yet? Should we implement it before commit?\",\n ],\n)\n\nREVIEW_RESULT = ActionNode(\n key=\"ReviewResult\",\n expected_type=Literal[\"LGTM\", \"LBTM\"],\n instruction=\"LGTM/LBTM. If the code is fully implemented, \" \"give a LGTM, otherwise provide a LBTM.\",\n example=\"LBTM\",\n)\n\nNEXT_STEPS = ActionNode(\n key=\"NextSteps\",\n expected_type=str,\n instruction=\"Based on the code review outcome, suggest actionable steps. This can include code changes, \"\n \"refactoring suggestions, or any follow-up tasks.\",\n example=\"\"\"1. Refactor the `process_data` method to improve readability and efficiency.\n2. Cover edge cases in the `validate_user` function.\n3. Implement a the TODO in the `calculate_total` function.\n4. Fix the `handle_events` method to update the game state only if a move is successful.\n ```python\n def handle_events(self):\n for event in pygame.event.get():\n if event.type == pygame.QUIT:\n return False\n if event.type == pygame.KEYDOWN:\n moved = False\n if event.key == pygame.K_UP:\n moved = self.game.move('UP')\n elif event.key == pygame.K_DOWN:\n moved = self.game.move('DOWN')\n elif event.key == pygame.K_LEFT:\n moved = self.game.move('LEFT')\n elif event.key == pygame.K_RIGHT:\n moved = self.game.move('RIGHT')\n if moved:\n # Update the game state only if a move was successful\n self.render()\n return True\n ```\n\"\"\",\n)\n\nWRITE_DRAFT = ActionNode(\n key=\"WriteDraft\",\n expected_type=str,\n instruction=\"Could you write draft code for move function in order to implement it?\",\n example=\"Draft: ...\",\n)\n\n\nWRITE_FUNCTION = ActionNode(\n key=\"WriteFunction\",\n expected_type=str,\n instruction=\"write code for the function not implemented.\",\n example=\"\"\"\n```Code\n...\n```\n\"\"\",\n)\n\n\nREWRITE_CODE = ActionNode(\n key=\"RewriteCode\",\n expected_type=str,\n instruction=\"\"\"rewrite code based on the Review and Actions\"\"\",\n example=\"\"\"\n```python\n## example.py\ndef calculate_total(price, quantity):\n total = price * quantity\n```\n\"\"\",\n)\n\n\nCODE_REVIEW_CONTEXT = \"\"\"\n# System\nRole: You are a professional software engineer, and your main task is to review and revise the code. You need to ensure that the code conforms to the google-style standards, is elegantly designed and modularized, easy to read and maintain.\nLanguage: Please use the same language as the user requirement, but the title and code should be still in English. For example, if the user speaks Chinese, the specific text of your answer should also be in Chinese.\n\n# Context\n## System Design\n{\"Implementation approach\": \"我们将使用HTML、CSS和JavaScript来实现这个单机的响应式2048游戏。为了确保游戏性能流畅和响应式设计，我们会选择使用Vue.js框架，因为它易于上手且适合构建交互式界面。我们还将使用localStorage来记录玩家的最高分。\", \"File list\": [\"index.html\", \"styles.css\", \"main.js\", \"game.js\", \"storage.js\"], \"Data structures and interfaces\": \"classDiagram\\\n class Game {\\\n -board Array\\\n -score Number\\\n -bestScore Number\\\n +constructor()\\\n +startGame()\\\n +move(direction: String)\\\n +getBoard() Array\\\n +getScore() Number\\\n +getBestScore() Number\\\n +setBestScore(score: Number)\\\n }\\\n class Storage {\\\n +getBestScore() Number\\\n +setBestScore(score: Number)\\\n }\\\n class Main {\\\n +init()\\\n +bindEvents()\\\n }\\\n Game --> Storage : uses\\\n Main --> Game : uses\", \"Program call flow\": \"sequenceDiagram\\\n participant M as Main\\\n participant G as Game\\\n participant S as Storage\\\n M->>G: init()\\\n G->>S: getBestScore()\\\n S-->>G: return bestScore\\\n M->>G: bindEvents()\\\n M->>G: startGame()\\\n loop Game Loop\\\n M->>G: move(direction)\\\n G->>S: setBestScore(score)\\\n S-->>G: return\\\n end\", \"Anything UNCLEAR\": \"目前项目要求明确，没有不清楚的地方。\"}\n\n## Tasks\n{\"Required Python packages\": [\"无需Python包\"], \"Required Other language third-party packages\": [\"vue.js\"], \"Logic Analysis\": [[\"index.html\", \"作为游戏的入口文件和主要的HTML结构\"], [\"styles.css\", \"包含所有的CSS样式，确保游戏界面美观\"], [\"main.js\", \"包含Main类，负责初始化游戏和绑定事件\"], [\"game.js\", \"包含Game类，负责游戏逻辑，如开始游戏、移动方块等\"], [\"storage.js\", \"包含Storage类，用于获取和设置玩家的最高分\"]], \"Task list\": [\"index.html\", \"styles.css\", \"storage.js\", \"game.js\", \"main.js\"], \"Full API spec\": \"\", \"Shared Knowledge\": \"\\'game.js\\' 包含游戏逻辑相关的函数，被 \\'main.js\\' 调用。\", \"Anything UNCLEAR\": \"目前项目要求明确，没有不清楚的地方。\"}\n\n## Code Files\n----- index.html\n\n\n\n \n \n 2048游戏\n \n \n\n\n
\n
2048
\n
\n
\n
分数
\n
{{ score }}
\n
\n
\n
最高分
\n
{{ bestScore }}
\n
\n
\n
\n
\n
\n {{ cell !== 0 ? cell : \\'\\' }}\n
\n
\n
\n \n
\n\n \n \n \n \n\n\n\n----- styles.css\n/* styles.css */\nbody, html {\n margin: 0;\n padding: 0;\n font-family: \\'Arial\\', sans-serif;\n}\n\n#app {\n text-align: center;\n font-size: 18px;\n color: #776e65;\n}\n\nh1 {\n color: #776e65;\n font-size: 72px;\n font-weight: bold;\n margin: 20px 0;\n}\n\n.scores-container {\n display: flex;\n justify-content: center;\n margin-bottom: 20px;\n}\n\n.score-container, .best-container {\n background: #bbada0;\n padding: 10px;\n border-radius: 5px;\n margin: 0 10px;\n min-width: 100px;\n text-align: center;\n}\n\n.score-header, .best-header {\n color: #eee4da;\n font-size: 18px;\n margin-bottom: 5px;\n}\n\n.game-container {\n max-width: 500px;\n margin: 0 auto 20px;\n background: #bbada0;\n padding: 15px;\n border-radius: 10px;\n position: relative;\n}\n\n.grid-row {\n display: flex;\n}\n\n.grid-cell {\n background: #cdc1b4;\n width: 100px;\n height: 100px;\n margin: 5px;\n display: flex;\n justify-content: center;\n align-items: center;\n font-size: 35px;\n font-weight: bold;\n color: #776e65;\n border-radius: 3px;\n}\n\n/* Dynamic classes for different number cells */\n.number-cell-2 {\n background: #eee4da;\n}\n\n.number-cell-4 {\n background: #ede0c8;\n}\n\n.number-cell-8 {\n background: #f2b179;\n color: #f9f6f2;\n}\n\n.number-cell-16 {\n background: #f59563;\n color: #f9f6f2;\n}\n\n.number-cell-32 {\n background: #f67c5f;\n color: #f9f6f2;\n}\n\n.number-cell-64 {\n background: #f65e3b;\n color: #f9f6f2;\n}\n\n.number-cell-128 {\n background: #edcf72;\n color: #f9f6f2;\n}\n\n.number-cell-256 {\n background: #edcc61;\n color: #f9f6f2;\n}\n\n.number-cell-512 {\n background: #edc850;\n color: #f9f6f2;\n}\n\n.number-cell-1024 {\n background: #edc53f;\n color: #f9f6f2;\n}\n\n.number-cell-2048 {\n background: #edc22e;\n color: #f9f6f2;\n}\n\n/* Larger numbers need smaller font sizes */\n.number-cell-1024, .number-cell-2048 {\n font-size: 30px;\n}\n\nbutton {\n background-color: #8f7a66;\n color: #f9f6f2;\n border: none;\n border-radius: 3px;\n padding: 10px 20px;\n font-size: 18px;\n cursor: pointer;\n outline: none;\n}\n\nbutton:hover {\n background-color: #9f8b76;\n}\n\n----- storage.js\n## storage.js\nclass Storage {\n // 获取最高分\n getBestScore() {\n // 尝试从localStorage中获取最高分，如果不存在则默认为0\n const bestScore = localStorage.getItem(\\'bestScore\\');\n return bestScore ? Number(bestScore) : 0;\n }\n\n // 设置最高分\n setBestScore(score) {\n // 将最高分设置到localStorage中\n localStorage.setItem(\\'bestScore\\', score.toString());\n }\n}\n\n\n\n## Code to be Reviewed: game.js\n```Code\n## game.js\nclass Game {\n constructor() {\n this.board = this.createEmptyBoard();\n this.score = 0;\n this.bestScore = 0;\n }\n\n createEmptyBoard() {\n const board = [];\n for (let i = 0; i < 4; i++) {\n board[i] = [0, 0, 0, 0];\n }\n return board;\n }\n\n startGame() {\n this.board = this.createEmptyBoard();\n this.score = 0;\n this.addRandomTile();\n this.addRandomTile();\n }\n\n addRandomTile() {\n let emptyCells = [];\n for (let r = 0; r < 4; r++) {\n for (let c = 0; c < 4; c++) {\n if (this.board[r][c] === 0) {\n emptyCells.push({ r, c });\n }\n }\n }\n if (emptyCells.length > 0) {\n let randomCell = emptyCells[Math.floor(Math.random() * emptyCells.length)];\n this.board[randomCell.r][randomCell.c] = Math.random() < 0.9 ? 2 : 4;\n }\n }\n\n move(direction) {\n // This function will handle the logic for moving tiles\n // in the specified direction and merging them\n // It will also update the score and add a new random tile if the move is successful\n // The actual implementation of this function is complex and would require\n // a significant amount of code to handle all the cases for moving and merging tiles\n // For the purposes of this example, we will not implement the full logic\n // Instead, we will just call addRandomTile to simulate a move\n this.addRandomTile();\n }\n\n getBoard() {\n return this.board;\n }\n\n getScore() {\n return this.score;\n }\n\n getBestScore() {\n return this.bestScore;\n }\n\n setBestScore(score) {\n this.bestScore = score;\n }\n}\n\n```\n\"\"\"\n\n\nCODE_REVIEW_SMALLEST_CONTEXT = \"\"\"\n## Code to be Reviewed: game.js\n```Code\n// game.js\nclass Game {\n constructor() {\n this.board = this.createEmptyBoard();\n this.score = 0;\n this.bestScore = 0;\n }\n\n createEmptyBoard() {\n const board = [];\n for (let i = 0; i < 4; i++) {\n board[i] = [0, 0, 0, 0];\n }\n return board;\n }\n\n startGame() {\n this.board = this.createEmptyBoard();\n this.score = 0;\n this.addRandomTile();\n this.addRandomTile();\n }\n\n addRandomTile() {\n let emptyCells = [];\n for (let r = 0; r < 4; r++) {\n for (let c = 0; c < 4; c++) {\n if (this.board[r][c] === 0) {\n emptyCells.push({ r, c });\n }\n }\n }\n if (emptyCells.length > 0) {\n let randomCell = emptyCells[Math.floor(Math.random() * emptyCells.length)];\n this.board[randomCell.r][randomCell.c] = Math.random() < 0.9 ? 2 : 4;\n }\n }\n\n move(direction) {\n // This function will handle the logic for moving tiles\n // in the specified direction and merging them\n // It will also update the score and add a new random tile if the move is successful\n // The actual implementation of this function is complex and would require\n // a significant amount of code to handle all the cases for moving and merging tiles\n // For the purposes of this example, we will not implement the full logic\n // Instead, we will just call addRandomTile to simulate a move\n this.addRandomTile();\n }\n\n getBoard() {\n return this.board;\n }\n\n getScore() {\n return this.score;\n }\n\n getBestScore() {\n return this.bestScore;\n }\n\n setBestScore(score) {\n this.bestScore = score;\n }\n}\n\n```\n\"\"\"\n\n\nCODE_REVIEW_SAMPLE = \"\"\"\n## Code Review: game.js\n1. The code partially implements the requirements. The `Game` class is missing the full implementation of the `move` method, which is crucial for the game\\'s functionality.\n2. The code logic is not completely correct. The `move` method is not implemented, which means the game cannot process player moves.\n3. The existing code follows the \"Data structures and interfaces\" in terms of class structure but lacks full method implementations.\n4. Not all functions are implemented. The `move` method is incomplete and does not handle the logic for moving and merging tiles.\n5. All necessary pre-dependencies seem to be imported since the code does not indicate the need for additional imports.\n6. The methods from other files (such as `Storage`) are not being used in the provided code snippet, but the class structure suggests that they will be used correctly.\n\n## Actions\n1. Implement the `move` method to handle tile movements and merging. This is a complex task that requires careful consideration of the game\\'s rules and logic. Here is a simplified version of how one might begin to implement the `move` method:\n ```javascript\n move(direction) {\n // Simplified logic for moving tiles up\n if (direction === \\'up\\') {\n for (let col = 0; col < 4; col++) {\n let tiles = this.board.map(row => row[col]).filter(val => val !== 0);\n let merged = [];\n for (let i = 0; i < tiles.length; i++) {\n if (tiles[i] === tiles[i + 1]) {\n tiles[i] *= 2;\n this.score += tiles[i];\n tiles[i + 1] = 0;\n merged.push(i);\n }\n }\n tiles = tiles.filter(val => val !== 0);\n while (tiles.length < 4) {\n tiles.push(0);\n }\n for (let row = 0; row < 4; row++) {\n this.board[row][col] = tiles[row];\n }\n }\n }\n // Additional logic needed for \\'down\\', \\'left\\', \\'right\\'\n // ...\n this.addRandomTile();\n }\n ```\n2. Integrate the `Storage` class methods to handle the best score. This means updating the `startGame` and `setBestScore` methods to use `Storage` for retrieving and setting the best score:\n ```javascript\n startGame() {\n this.board = this.createEmptyBoard();\n this.score = 0;\n this.bestScore = new Storage().getBestScore(); // Retrieve the best score from storage\n this.addRandomTile();\n this.addRandomTile();\n }\n\n setBestScore(score) {\n if (score > this.bestScore) {\n this.bestScore = score;\n new Storage().setBestScore(score); // Set the new best score in storage\n }\n }\n ```\n\n## Code Review Result\nLBTM\n\n```\n\"\"\"\n\n\nWRITE_CODE_NODE = ActionNode.from_children(\"WRITE_REVIEW_NODE\", [REVIEW, REVIEW_RESULT, NEXT_STEPS])\nWRITE_MOVE_NODE = ActionNode.from_children(\"WRITE_MOVE_NODE\", [WRITE_DRAFT, WRITE_FUNCTION])\n\n\nCR_FOR_MOVE_FUNCTION_BY_3 = \"\"\"\nThe move function implementation provided appears to be well-structured and follows a clear logic for moving and merging tiles in the specified direction. However, there are a few potential improvements that could be made to enhance the code:\n\n1. Encapsulation: The logic for moving and merging tiles could be encapsulated into smaller, reusable functions to improve readability and maintainability.\n\n2. Magic Numbers: There are some magic numbers (e.g., 4, 3) used in the loops that could be replaced with named constants for improved readability and easier maintenance.\n\n3. Comments: Adding comments to explain the logic and purpose of each section of the code can improve understanding for future developers who may need to work on or maintain the code.\n\n4. Error Handling: It's important to consider error handling for unexpected input or edge cases to ensure the function behaves as expected in all scenarios.\n\nOverall, the code could benefit from refactoring to improve readability, maintainability, and extensibility. If you would like, I can provide a refactored version of the move function that addresses these considerations.\n\"\"\"\n\n\nclass WriteCodeAN(Action):\n \"\"\"Write a code review for the context.\"\"\"\n\n async def run(self, context):\n self.llm.system_prompt = \"You are an outstanding engineer and can implement any code\"\n return await WRITE_MOVE_NODE.fill(context=context, llm=self.llm, schema=\"json\")\n\n\nasync def main():\n await WriteCodeAN().run(CODE_REVIEW_SMALLEST_CONTEXT)\n\n\nif __name__ == \"__main__\":\n asyncio.run(main())\n\n```\n\n---\nTranslate the code above into Mermaid Sequence Diagram.": "```mermaid\nsequenceDiagram\n participant User\n participant WriteCodeAN\n participant WRITE_MOVE_NODE\n\n User->>WriteCodeAN: Run\n WriteCodeAN->>WRITE_MOVE_NODE: Fill\n WRITE_MOVE_NODE->>WriteCodeAN: Filled\n WriteCodeAN->>User: Return Filled Context\n```",
"You are a function parser.#MSG_SEP#You can convert spoken words into function parameters.#SYSTEM_MSG_END#text_to_image function parameters description:\nparameter `text`: The text used for image conversion.\nparameter `size_type`: size type\n\n---\nExamples:\nIf want you to do `Draw a girl`, return `text_to_image(text=\"Draw a girl\", size_type=\"512x512\")` brief and clear.\nIf want you to do `Draw an apple`, return `text_to_image(text=\"Draw an apple\", size_type=\"512x512\")` brief and clear.\n\n---\n\nRefer to the `text_to_image` function description, and fill in the function parameters according to the example \"I want you to do xx\" in the Examples section.\nNow I want you to do `Draw an apple`, return function parameters in Examples format above, brief and clear.": "`text_to_image(text=\"Draw an apple\", size_type=\"512x512\")`",
"\nNOTICE\nRole: You are a professional engineer; the main goal is to write google-style, elegant, modular, easy to read and maintain code\nLanguage: Please use the same language as the user requirement, but the title and code should be still in English. For example, if the user speaks Chinese, the specific text of your answer should also be in Chinese.\nATTENTION: Use '##' to SPLIT SECTIONS, not '#'. Output format carefully referenced \"Format example\".\n\n# Context\n## Design\n{\"Implementation approach\": \"We will use the Pygame library to create the game interface and handle user input. The game logic will be implemented using Python classes and data structures.\", \"File list\": [\"main.py\", \"game.py\"], \"Data structures and interfaces\": \"classDiagram\\n class Game {\\n -grid: List[List[int]]\\n -score: int\\n -game_over: bool\\n +__init__()\\n +reset_game()\\n +move(direction: str)\\n +is_game_over() bool\\n +get_empty_cells() List[Tuple[int, int]]\\n +add_new_tile()\\n +get_score() int\\n }\\n class UI {\\n -game: Game\\n +__init__(game: Game)\\n +draw_grid()\\n +draw_score()\\n +draw_game_over()\\n +handle_input()\\n }\\n Game --> UI\", \"Program call flow\": \"sequenceDiagram\\n participant M as Main\\n participant G as Game\\n participant U as UI\\n M->>G: reset_game()\\n M->>U: draw_grid()\\n M->>U: draw_score()\\n M->>U: handle_input()\\n U->>G: move(direction)\\n G->>G: add_new_tile()\\n G->>U: draw_grid()\\n G->>U: draw_score()\\n G->>U: draw_game_over()\\n G->>G: is_game_over()\\n G->>G: get_empty_cells()\\n G->>G: get_score()\", \"Anything UNCLEAR\": \"...\"}\n\n## Tasks\n{\"Required Python packages\": [\"pygame==2.0.1\"], \"Required Other language third-party packages\": [\"No third-party dependencies required\"], \"Logic Analysis\": [[\"game.py\", \"Contains Game class and related functions for game logic\"], [\"main.py\", \"Contains main function, initializes the game and UI\"]], \"Task list\": [\"game.py\", \"main.py\"], \"Full API spec\": \"\", \"Shared Knowledge\": \"The game logic will be implemented using Python classes and data structures. The Pygame library will be used to create the game interface and handle user input.\", \"Anything UNCLEAR\": \"...\"}\n\n## Legacy Code\n```Code\n----- main.py\nif __name__ == \"__main__\":\nmain()\n```\n\n## Debug logs\n```text\n\n\n\n```\n\n## Bug Feedback logs\n```text\n\n```\n\n# Format example\n## Code: game.py\n```python\n## game.py\n...\n```\n\n# Instruction: Based on the context, follow \"Format example\", write code.\n\n## Code: game.py. Write code with triple quoto, based on the following attentions and context.\n1. Only One file: do your best to implement THIS ONLY ONE FILE.\n2. COMPLETE CODE: Your code will be part of the entire project, so please implement complete, reliable, reusable code snippets.\n3. Set default value: If there is any setting, ALWAYS SET A DEFAULT VALUE, ALWAYS USE STRONG TYPE AND EXPLICIT VARIABLE. AVOID circular import.\n4. Follow design: YOU MUST FOLLOW \"Data structures and interfaces\". DONT CHANGE ANY DESIGN. Do not use public member functions that do not exist in your design.\n5. CAREFULLY CHECK THAT YOU DONT MISS ANY NECESSARY CLASS/FUNCTION IN THIS FILE.\n6. Before using a external variable/module, make sure you import it first.\n7. Write out EVERY CODE DETAIL, DON'T LEAVE TODO.\n\n": "```python\n## game.py\n\nfrom typing import List, Tuple\nimport random\n\nclass Game:\n def __init__(self):\n self.grid: List[List[int]] = [[0 for _ in range(4)] for _ in range(4)]\n self.score: int = 0\n self.game_over: bool = False\n\n def reset_game(self):\n self.grid = [[0 for _ in range(4)] for _ in range(4)]\n self.score = 0\n self.game_over = False\n self.add_new_tile()\n self.add_new_tile()\n\n def move(self, direction: str):\n # Implement the logic to move the tiles in the specified direction\n pass\n\n def is_game_over(self) -> bool:\n # Implement the logic to check if the game is over\n pass\n\n def get_empty_cells(self) -> List[Tuple[int, int]]:\n # Implement the logic to get the coordinates of empty cells in the grid\n pass\n\n def add_new_tile(self):\n empty_cells = self.get_empty_cells()\n if empty_cells:\n row, col = random.choice(empty_cells)\n self.grid[row][col] = 2 if random.random() < 0.9 else 4\n\n def get_score(self) -> int:\n return self.score\n```",
- "\n## context\n\n### Project Name\n\n\n### Original Requirements\n开发一个基于大语言模型与私有知识库的搜索引擎，希望可以基于大语言模型进行搜索总结\n\n### Search Information\n-\n\n\n-----\n\n## format example\n[CONTENT]\n{\n \"Language\": \"en_us\",\n \"Programming Language\": \"Python\",\n \"Original Requirements\": \"Create a 2048 game\",\n \"Project Name\": \"game_2048\",\n \"Product Goals\": [\n \"Create an engaging user experience\",\n \"Improve accessibility, be responsive\",\n \"More beautiful UI\"\n ],\n \"User Stories\": [\n \"As a player, I want to be able to choose difficulty levels\",\n \"As a player, I want to see my score after each game\",\n \"As a player, I want to get restart button when I lose\",\n \"As a player, I want to see beautiful UI that make me feel good\",\n \"As a player, I want to play game via mobile phone\"\n ],\n \"Competitive Analysis\": [\n \"2048 Game A: Simple interface, lacks responsive features\",\n \"play2048.co: Beautiful and responsive UI with my best score shown\",\n \"2048game.com: Responsive UI with my best score shown, but many ads\"\n ],\n \"Competitive Quadrant Chart\": \"quadrantChart\\n title \\\"Reach and engagement of campaigns\\\"\\n x-axis \\\"Low Reach\\\" --> \\\"High Reach\\\"\\n y-axis \\\"Low Engagement\\\" --> \\\"High Engagement\\\"\\n quadrant-1 \\\"We should expand\\\"\\n quadrant-2 \\\"Need to promote\\\"\\n quadrant-3 \\\"Re-evaluate\\\"\\n quadrant-4 \\\"May be improved\\\"\\n \\\"Campaign A\\\": [0.3, 0.6]\\n \\\"Campaign B\\\": [0.45, 0.23]\\n \\\"Campaign C\\\": [0.57, 0.69]\\n \\\"Campaign D\\\": [0.78, 0.34]\\n \\\"Campaign E\\\": [0.40, 0.34]\\n \\\"Campaign F\\\": [0.35, 0.78]\\n \\\"Our Target Product\\\": [0.5, 0.6]\",\n \"Requirement Analysis\": \"\",\n \"Requirement Pool\": [\n [\n \"P0\",\n \"The main code ...\"\n ],\n [\n \"P0\",\n \"The game algorithm ...\"\n ]\n ],\n \"UI Design draft\": \"Basic function description with a simple style and layout.\",\n \"Anything UNCLEAR\": \"\"\n}\n[/CONTENT]\n\n## nodes: \": # \"\n- Language: # Provide the language used in the project, typically matching the user's requirement language.\n- Programming Language: # Python/JavaScript or other mainstream programming language.\n- Original Requirements: # Place the original user's requirements here.\n- Project Name: # According to the content of \"Original Requirements,\" name the project using snake case style , like 'game_2048' or 'simple_crm.\n- Product Goals: typing.List[str] # Provide up to three clear, orthogonal product goals.\n- User Stories: typing.List[str] # Provide up to 3 to 5 scenario-based user stories.\n- Competitive Analysis: typing.List[str] # Provide 5 to 7 competitive products.\n- Competitive Quadrant Chart: # Use mermaid quadrantChart syntax. Distribute scores evenly between 0 and 1\n- Requirement Analysis: # Provide a detailed analysis of the requirements.\n- Requirement Pool: typing.List[typing.List[str]] # List down the top-5 requirements with their priority (P0, P1, P2).\n- UI Design draft: # Provide a simple description of UI elements, functions, style, and layout.\n- Anything UNCLEAR: # Mention any aspects of the project that are unclear and try to clarify them.\n\n\n## constraint\nLanguage: Please use the same language as Human INPUT.\nFormat: output wrapped inside [CONTENT][/CONTENT] like format example, nothing else.\n\n## action\nFollow instructions of nodes, generate output and make sure it follows the format example.\n": "[CONTENT]\n{\n \"Language\": \"zh_cn\",\n \"Programming Language\": \"Python\",\n \"Original Requirements\": \"开发一个基于大语言模型与私有知识库的搜索引擎，希望可以基于大语言模型进行搜索总结\",\n \"Project Name\": \"language_model_search\",\n \"Product Goals\": [\n \"提供基于大语言模型的高效搜索功能\",\n \"整合私有知识库，实现个性化搜索结果\",\n \"实现搜索结果的自动总结功能\"\n ],\n \"User Stories\": [\n \"作为用户，我希望能够快速找到我需要的信息\",\n \"作为用户，我希望搜索结果能够根据我的偏好进行个性化排序\",\n \"作为用户，我希望搜索结果能够自动总结，方便我快速了解内容\"\n ],\n \"Competitive Analysis\": [\n \"搜索引擎A: 提供基于大语言模型的搜索功能，但个性化程度较低\",\n \"知识库B: 整合私有知识库，但搜索速度较慢\",\n \"语言模型搜索C: 提供搜索结果自动总结功能，但搜索准确度有待提高\"\n ],\n \"Competitive Quadrant Chart\": \"quadrantChart\\n title \\\"搜索引擎竞争分析\\\"\\n x-axis \\\"低搜索准确度\\\" --> \\\"高搜索准确度\\\"\\n y-axis \\\"低个性化程度\\\" --> \\\"高个性化程度\\\"\\n quadrant-1 \\\"需改进\\\"\\n quadrant-2 \\\"需提升\\\"\\n quadrant-3 \\\"重新评估\\\"\\n quadrant-4 \\\"扩展发展\\\"\\n \\\"搜索引擎A\\\": [0.6, 0.3]\\n \\\"知识库B\\\": [0.4, 0.2]\\n \\\"语言模型搜索C\\\": [0.7, 0.5]\\n \\\"我们的目标产品\\\": [0.8, 0.7]\",\n \"Requirement Analysis\": \"\",\n \"Requirement Pool\": [\n [\n \"P0\",\n \"基于大语言模型的高效搜索功能\"\n ],\n [\n \"P1\",\n \"整合私有知识库，实现个性化搜索结果\"\n ],\n [\n \"P2\",\n \"实现搜索结果的自动总结功能\"\n ]\n ],\n \"UI Design draft\": \"搜索页面简洁明了，搜索结果清晰展示，提供个性化排序和自动总结功能。\",\n \"Anything UNCLEAR\": \"\"\n}\n[/CONTENT]"
+ "\n## context\n\n### Project Name\n\n\n### Original Requirements\n开发一个基于大语言模型与私有知识库的搜索引擎，希望可以基于大语言模型进行搜索总结\n\n### Search Information\n-\n\n\n-----\n\n## format example\n[CONTENT]\n{\n \"Language\": \"en_us\",\n \"Programming Language\": \"Python\",\n \"Original Requirements\": \"Create a 2048 game\",\n \"Project Name\": \"game_2048\",\n \"Product Goals\": [\n \"Create an engaging user experience\",\n \"Improve accessibility, be responsive\",\n \"More beautiful UI\"\n ],\n \"User Stories\": [\n \"As a player, I want to be able to choose difficulty levels\",\n \"As a player, I want to see my score after each game\",\n \"As a player, I want to get restart button when I lose\",\n \"As a player, I want to see beautiful UI that make me feel good\",\n \"As a player, I want to play game via mobile phone\"\n ],\n \"Competitive Analysis\": [\n \"2048 Game A: Simple interface, lacks responsive features\",\n \"play2048.co: Beautiful and responsive UI with my best score shown\",\n \"2048game.com: Responsive UI with my best score shown, but many ads\"\n ],\n \"Competitive Quadrant Chart\": \"quadrantChart\\n title \\\"Reach and engagement of campaigns\\\"\\n x-axis \\\"Low Reach\\\" --> \\\"High Reach\\\"\\n y-axis \\\"Low Engagement\\\" --> \\\"High Engagement\\\"\\n quadrant-1 \\\"We should expand\\\"\\n quadrant-2 \\\"Need to promote\\\"\\n quadrant-3 \\\"Re-evaluate\\\"\\n quadrant-4 \\\"May be improved\\\"\\n \\\"Campaign A\\\": [0.3, 0.6]\\n \\\"Campaign B\\\": [0.45, 0.23]\\n \\\"Campaign C\\\": [0.57, 0.69]\\n \\\"Campaign D\\\": [0.78, 0.34]\\n \\\"Campaign E\\\": [0.40, 0.34]\\n \\\"Campaign F\\\": [0.35, 0.78]\\n \\\"Our Target Product\\\": [0.5, 0.6]\",\n \"Requirement Analysis\": \"\",\n \"Requirement Pool\": [\n [\n \"P0\",\n \"The main code ...\"\n ],\n [\n \"P0\",\n \"The game algorithm ...\"\n ]\n ],\n \"UI Design draft\": \"Basic function description with a simple style and layout.\",\n \"Anything UNCLEAR\": \"\"\n}\n[/CONTENT]\n\n## nodes: \": # \"\n- Language: # Provide the language used in the project, typically matching the user's requirement language.\n- Programming Language: # Python/JavaScript or other mainstream programming language.\n- Original Requirements: # Place the original user's requirements here.\n- Project Name: # According to the content of \"Original Requirements,\" name the project using snake case style , like 'game_2048' or 'simple_crm.\n- Product Goals: typing.List[str] # Provide up to three clear, orthogonal product goals.\n- User Stories: typing.List[str] # Provide up to 3 to 5 scenario-based user stories.\n- Competitive Analysis: typing.List[str] # Provide 5 to 7 competitive products.\n- Competitive Quadrant Chart: # Use mermaid quadrantChart syntax. Distribute scores evenly between 0 and 1\n- Requirement Analysis: # Provide a detailed analysis of the requirements.\n- Requirement Pool: typing.List[typing.List[str]] # List down the top-5 requirements with their priority (P0, P1, P2).\n- UI Design draft: # Provide a simple description of UI elements, functions, style, and layout.\n- Anything UNCLEAR: # Mention any aspects of the project that are unclear and try to clarify them.\n\n\n## constraint\nLanguage: Please use the same language as Human INPUT.\nFormat: output wrapped inside [CONTENT][/CONTENT] like format example, nothing else.\n\n## action\nFollow instructions of nodes, generate output and make sure it follows the format example.\n": "[CONTENT]\n{\n \"Language\": \"zh_cn\",\n \"Programming Language\": \"Python\",\n \"Original Requirements\": \"开发一个基于大语言模型与私有知识库的搜索引擎，希望可以基于大语言模型进行搜索总结\",\n \"Project Name\": \"language_model_search\",\n \"Product Goals\": [\n \"提供基于大语言模型的高效搜索功能\",\n \"整合私有知识库，实现个性化搜索结果\",\n \"实现搜索结果的自动总结功能\"\n ],\n \"User Stories\": [\n \"作为用户，我希望能够快速找到我需要的信息\",\n \"作为用户，我希望搜索结果能够根据我的偏好进行个性化排序\",\n \"作为用户，我希望搜索结果能够自动总结，方便我快速了解内容\"\n ],\n \"Competitive Analysis\": [\n \"搜索引擎A: 提供基于大语言模型的搜索功能，但个性化程度较低\",\n \"知识库B: 整合私有知识库，但搜索速度较慢\",\n \"语言模型搜索C: 提供搜索结果自动总结功能，但搜索准确度有待提高\"\n ],\n \"Competitive Quadrant Chart\": \"quadrantChart\\n title \\\"搜索引擎竞争分析\\\"\\n x-axis \\\"低搜索准确度\\\" --> \\\"高搜索准确度\\\"\\n y-axis \\\"低个性化程度\\\" --> \\\"高个性化程度\\\"\\n quadrant-1 \\\"需改进\\\"\\n quadrant-2 \\\"需提升\\\"\\n quadrant-3 \\\"重新评估\\\"\\n quadrant-4 \\\"扩展发展\\\"\\n \\\"搜索引擎A\\\": [0.6, 0.3]\\n \\\"知识库B\\\": [0.4, 0.2]\\n \\\"语言模型搜索C\\\": [0.7, 0.5]\\n \\\"我们的目标产品\\\": [0.8, 0.7]\",\n \"Requirement Analysis\": \"\",\n \"Requirement Pool\": [\n [\n \"P0\",\n \"基于大语言模型的高效搜索功能\"\n ],\n [\n \"P1\",\n \"整合私有知识库，实现个性化搜索结果\"\n ],\n [\n \"P2\",\n \"实现搜索结果的自动总结功能\"\n ]\n ],\n \"UI Design draft\": \"搜索页面简洁明了，搜索结果清晰展示，提供个性化排序和自动总结功能。\",\n \"Anything UNCLEAR\": \"\"\n}\n[/CONTENT]",
+ "\n## context\nNone\n\n-----\n\n## format example\n[CONTENT]\n{\n \"Project Name\": \"game_2048\"\n}\n[/CONTENT]\n\n## nodes: \": # \"\n- Project Name: # According to the content of \"Original Requirements,\" name the project using snake case style with underline, like 'game_2048' or 'simple_crm.\n\n\n## constraint\nLanguage: Please use the same language as Human INPUT.\nFormat: output wrapped inside [CONTENT][/CONTENT] like format example, nothing else.\n\n## action\nFollow instructions of nodes, generate output and make sure it follows the format example.\n": "[CONTENT]\n{\n \"Project Name\": \"game_2048\"\n}\n[/CONTENT]",
+ "\n## context\nCompare the key's value of nodes_output and the corresponding requirements one by one. If a key's value that does not match the requirement is found, provide the comment content on how to modify it. No output is required for matching keys.\n\n### nodes_output\n{\"Project Name\": {\"value\": \"game snake\", \"requirement\": \"According to the content of \\\"Original Requirements,\\\" name the project using snake case style with underline, like 'game_2048' or 'simple_crm.\"}}\n\n-----\n\n## format example\n[CONTENT]\n{\n \"key1\": \"comment1\",\n \"key2\": \"comment2\",\n \"keyn\": \"commentn\"\n}\n[/CONTENT]\n\n## nodes: \": # \"\n- key1: # the first key name of mismatch key\n- key2: # the second key name of mismatch key\n- keyn: # the last key name of mismatch key\n\n## constraint\nFormat: output wrapped inside [CONTENT][/CONTENT] like format example, nothing else.\n\n## action\nFollow format example's json format, generate output and make sure it follows the format example.\n": "[CONTENT]\n{\n \"Project Name\": \"Use snake case style with underline, like 'game_2048' or 'simple_crm.'\"\n}\n[/CONTENT]",
+ "\n## context\nCompare the key's value of nodes_output and the corresponding requirements one by one. If a key's value that does not match the requirement is found, provide the comment content on how to modify it. No output is required for matching keys.\n\n### nodes_output\n{\"Project Name\": {\"value\": \"game_2048\", \"requirement\": \"According to the content of \\\"Original Requirements,\\\" name the project using snake case style with underline, like 'game_2048' or 'simple_crm.\"}}\n\n-----\n\n## format example\n[CONTENT]\n{\n \"key1\": \"comment1\",\n \"key2\": \"comment2\",\n \"keyn\": \"commentn\"\n}\n[/CONTENT]\n\n## nodes: \": # \"\n- key1: # the first key name of mismatch key\n- key2: # the second key name of mismatch key\n- keyn: # the last key name of mismatch key\n\n## constraint\nFormat: output wrapped inside [CONTENT][/CONTENT] like format example, nothing else.\n\n## action\nFollow format example's json format, generate output and make sure it follows the format example.\n": "[CONTENT]\n{\n \"Project Name\": \"According to the content of \\\"Original Requirements,\\\" name the project using snake case style with underline, like 'game_2048' or 'simple_crm.\"\n}\n[/CONTENT]",
+ "\n## context\nchange the nodes_output key's value to meet its comment and no need to add extra comment.\n\n### nodes_output\n{\"Project Name\": {\"value\": \"game snake\", \"comment\": \"Use snake case style with underline, like 'game_2048' or 'simple_crm.'\"}}\n\n-----\n\n## format example\n[CONTENT]\n{\n \"Project Name\": \"game_2048\"\n}\n[/CONTENT]\n\n## nodes: \": # \"\n- Project Name: # According to the content of \"Original Requirements,\" name the project using snake case style with underline, like 'game_2048' or 'simple_crm.\n\n\n## constraint\nFormat: output wrapped inside [CONTENT][/CONTENT] like format example, nothing else.\n\n## action\nFollow format example's json format, generate output and make sure it follows the format example.\n": "[CONTENT]\n{\n \"Project Name\": \"game_snake\"\n}\n[/CONTENT]",
+ "\n## context\nCompare the key's value of nodes_output and the corresponding requirements one by one. If a key's value that does not match the requirement is found, provide the comment content on how to modify it. No output is required for matching keys.\n\n### nodes_output\n{\"Project Name\": {\"value\": \"game_snake\", \"requirement\": \"According to the content of \\\"Original Requirements,\\\" name the project using snake case style with underline, like 'game_2048' or 'simple_crm.\"}}\n\n-----\n\n## format example\n[CONTENT]\n{\n \"key1\": \"comment1\",\n \"key2\": \"comment2\",\n \"keyn\": \"commentn\"\n}\n[/CONTENT]\n\n## nodes: \": # \"\n- key1: # the first key name of mismatch key\n- key2: # the second key name of mismatch key\n- keyn: # the last key name of mismatch key\n\n## constraint\nFormat: output wrapped inside [CONTENT][/CONTENT] like format example, nothing else.\n\n## action\nFollow format example's json format, generate output and make sure it follows the format example.\n": "[CONTENT]\n{\n \"Project Name\": \"The project name should be in snake case style with underscores, such as 'game_2048' or 'simple_crm'.\"\n}\n[/CONTENT]",
+ "\n## context\nchange the nodes_output key's value to meet its comment and no need to add extra comment.\n\n### nodes_output\n{\"Project Name\": {\"value\": \"game_snake\", \"comment\": \"The project name should be in snake case style with underscores, such as 'game_2048' or 'simple_crm'.\"}}\n\n-----\n\n## format example\n[CONTENT]\n{\n \"Project Name\": \"game_2048\"\n}\n[/CONTENT]\n\n## nodes: \": # \"\n- Project Name: # According to the content of \"Original Requirements,\" name the project using snake case style with underline, like 'game_2048' or 'simple_crm.\n\n\n## constraint\nFormat: output wrapped inside [CONTENT][/CONTENT] like format example, nothing else.\n\n## action\nFollow format example's json format, generate output and make sure it follows the format example.\n": "[CONTENT]\n{\n \"Project Name\": \"game_snake\"\n}\n[/CONTENT]",
+ "You are an action classifier#SYSTEM_MSG_END#If the text explicitly want you to Generate a voice file from the input text, text-to-speech, return `[SKILL]: text_to_speech` brief and clear. For instance: [SKILL]: text_to_speech\nIf the text explicitly want you to Create a drawing based on the text., return `[SKILL]: text_to_image` brief and clear. For instance: [SKILL]: text_to_image\nIf the text explicitly want you to Perform Google searches to provide real-time information., return `[SKILL]: web_search` brief and clear. For instance: [SKILL]: web_search\nOtherwise, return `[TALK]: {talk}` brief and clear. For instance: if {talk} is \"xxxx\" return [TALK]: xxxx\n\nNow what specific action is explicitly mentioned in the text: No, I do not have a poison apple. Do you have a poison apple?\n": "[TALK]: No, I do not have a poison apple. Do you have a poison apple?",
+ "You are an action classifier#SYSTEM_MSG_END#If the text explicitly want you to Generate a voice file from the input text, text-to-speech, return `[SKILL]: text_to_speech` brief and clear. For instance: [SKILL]: text_to_speech\nIf the text explicitly want you to Create a drawing based on the text., return `[SKILL]: text_to_image` brief and clear. For instance: [SKILL]: text_to_image\nIf the text explicitly want you to Perform Google searches to provide real-time information., return `[SKILL]: web_search` brief and clear. For instance: [SKILL]: web_search\nOtherwise, return `[TALK]: {talk}` brief and clear. For instance: if {talk} is \"xxxx\" return [TALK]: xxxx\n\nNow what specific action is explicitly mentioned in the text: Sure, I can draw you an apple. Draw me an apple.\n": "[SKILL]: text_to_image",
+ "You are a function parser.#MSG_SEP#You can convert spoken words into function parameters.#SYSTEM_MSG_END#text_to_image function parameters description:\nparameter `text`: The text used for image conversion.\nparameter `size_type`: size type\n\n---\nExamples:\nIf want you to do `Draw a girl`, return `text_to_image(text=\"Draw a girl\", size_type=\"512x512\")` brief and clear.\nIf want you to do `Draw an apple`, return `text_to_image(text=\"Draw an apple\", size_type=\"512x512\")` brief and clear.\n\n---\n\nRefer to the `text_to_image` function description, and fill in the function parameters according to the example \"I want you to do xx\" in the Examples section.\nNow I want you to do `Sure, I can draw you an apple. Draw me an apple.`, return function parameters in Examples format above, brief and clear.": "`text_to_image(text=\"Sure, I can draw you an apple. Draw me an apple.\", size_type=\"512x512\")`",
+ "You are an action classifier#SYSTEM_MSG_END#Otherwise, return `[TALK]: {talk}` brief and clear. For instance: if {talk} is \"xxxx\" return [TALK]: xxxx\n\nNow what specific action is explicitly mentioned in the text: Sure, I can draw you an apple. Draw me an apple.\n": "[DRAW]: draw an apple"
}
\ No newline at end of file