From bfdc3f1e497867db59714f23b05bfed9c5fe08c4 Mon Sep 17 00:00:00 2001 From: camera-2018 <2907618001@qq.com> Date: Thu, 20 Apr 2023 12:49:21 +0800 Subject: [PATCH] =?UTF-8?q?fix:=204.10=20=E5=B0=96=E6=8B=AC=E5=8F=B7?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- ... 智能系统 —— 从 LLMs 到 Agents.md | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/4.人工智能/4.10从 AI 到 智能系统 —— 从 LLMs 到 Agents.md b/4.人工智能/4.10从 AI 到 智能系统 —— 从 LLMs 到 Agents.md index 7935c6d..8119327 100644 --- a/4.人工智能/4.10从 AI 到 智能系统 —— 从 LLMs 到 Agents.md +++ b/4.人工智能/4.10从 AI 到 智能系统 —— 从 LLMs 到 Agents.md @@ -737,18 +737,18 @@ AutoGPT 所使用的 Commands 接口中,就有很大一部分接口用于实 - Google Search: "google", - - args: "input": "" + - args: "input": `""` - 返回网页关键信息与网址的匹配列表 - Browse Website: "browse_website", - args: - - "url": "", - - "question": "" + - "url": `""`, + - "question": `""` - 返回网址中与 question 相关的匹配文段(通过 embedding 搜索) - Read file: "read_file", - - args: "file": "" + - args: "file": `""` - 读取文件并解析文本数据 这些访问接口由程序集暴露给 GPT ,将知觉系统中实际使用的微观处理器隐藏在了系统框架之下 @@ -836,10 +836,10 @@ Generative Agents 的知觉设计:关联性难题 对于如上数据库信息,我们对于以视图为单位进行整理,将其转化为 prompt 输入 'Commands:\n' -'1. Google Search: "google", args: "input": ""\n' -'2. Browse Website: "browse_website", args: "url": "", "question": ""\n' -'3. Start GPT Agent: "start_agent", args: "name": "", "task": "", "prompt": ""\n' -'4. Message GPT Agent: "message_agent", args: "key": "", "message": ""\n' +'1. Google Search: "google", args: "input": `""`\n' +'2. Browse Website: "browse_website", args: "url": `""`, "question": `""`\n' +'3. Start GPT Agent: "start_agent", args: "name": `""`, "task": `""`, "prompt": `""`\n' +'4. Message GPT Agent: "message_agent", args: "key": `""`, "message": `""`\n' '5. List GPT Agents: "list_agents", args: \n' - 对于对数据库进行感知的任务,我们需要进一步设计 Pormpt 驱动 Agent 的行为