ai-agent-book 精选快照(<2MB 代码与文档,来自 github.com/bojieli/ai-agent-book)
Build latest book artifacts / build (push) Canceled after 0s
dependency resolution / resolve (3.11) (push) Canceled after 0s
dependency resolution / resolve (3.13) (push) Canceled after 0s
deploy-pages / build (push) Canceled after 0s
deploy-pages / deploy (push) Canceled after 0s
i18n consistency check / check (push) Canceled after 0s
provider adoption tests / test (chapter2/context-compression) (push) Canceled after 0s
provider adoption tests / test (chapter2/prompt-injection) (push) Canceled after 0s
provider adoption tests / test (chapter2/system-hint) (push) Canceled after 0s
provider adoption tests / test (chapter3/log-sanitization) (push) Canceled after 0s
web-search-agent tests / test (push) Canceled after 0s
web-search-agent tests / agentbook (push) Canceled after 0s

This commit is contained in:
2026-08-20 13:12:50 +00:00
commit b119135836
10275 changed files with 3284984 additions and 0 deletions
+197
View File
@@ -0,0 +1,197 @@
"""实验 5-11:对话式界面定制 Agent。
职责:接收一条自然语言 UI 定制需求(如"把发送按钮改成蓝色"),读取前端源码,
调用 OpenAI 让模型定位并改写相应源文件(颜色 / 字体 / 文案 / 布局 / 组件)。
设计要点
--------
- 只暴露少量"可定制文件"给模型(frontend/src 下的 App.jsx 与 theme.css),
降低模型改错文件的概率,也让改动可控、可验证。
- 通过 function calling 的 `apply_edits` 工具,让模型返回"要整体改写的文件全文"
相比零散的 search/replace,整文件改写对小文件更稳定、更少破坏语法。
- 修改前先把原文件内容快照下来,改后可计算 diff、读回断言,并跑构建验证。
环境变量:
OPENAI_API_KEY (必填,本实验读取此项)
OPENAI_BASE_URL (可选,切换到兼容 OpenAI 协议的服务端点)
MODEL (可选,默认 gpt-5.6-luna
OPENROUTER_API_KEY(可选,无直连 key 时自动改走 OpenRouter 兜底)
"""
import os
import json
from pathlib import Path
from openai import OpenAI
try:
from dotenv import load_dotenv
load_dotenv()
except Exception: # dotenv 是可选依赖
pass
# 可被 Agent 定制的前端源文件(相对 frontend/ 的路径)。
EDITABLE_FILES = [
"src/App.jsx",
"src/theme.css",
]
OPENROUTER_BASE_URL = "https://openrouter.ai/api/v1"
def map_model_to_openrouter(model: str) -> str:
"""把直连模型名映射为 OpenRouter 上的 id(非可映射 id 统一兜底到当前廉价旗舰)。"""
if not model or "/" in model:
return model or "openai/gpt-5.6-luna"
m = model.lower()
if m.startswith(("gpt-", "o1", "o3", "o4")):
return "openai/" + model
if m.startswith("claude"):
if "haiku" in m:
return "anthropic/claude-haiku-4.5"
if "sonnet" in m:
return "anthropic/claude-sonnet-4.6"
return "anthropic/claude-opus-4.8"
if m.startswith("gemini"):
return "google/" + model
return "openai/gpt-5.6-luna"
def build_client_and_model():
model = os.getenv("MODEL", "gpt-5.6-luna")
api_key = os.getenv("OPENAI_API_KEY")
base_url = os.getenv("OPENAI_BASE_URL")
orkey = os.getenv("OPENROUTER_API_KEY")
# 通用 OpenRouter 兜底:无直连 key,或默认 gpt-5.x(直连需组织实名认证)时改走 OpenRouter。
prefer_or = bool(orkey) and (model or "").lower().startswith("gpt-5")
if prefer_or or (not api_key and orkey):
api_key, base_url, model = orkey, OPENROUTER_BASE_URL, map_model_to_openrouter(model)
if not api_key:
raise SystemExit("未找到 OPENAI_API_KEY(或 OPENROUTER_API_KEY 兜底),请先在环境变量或 .env 中设置。")
# timeout / max_retries:让偶发的网络/SSL 抖动自动重试,不至于整轮崩溃
client_kwargs = {"api_key": api_key, "timeout": 60.0, "max_retries": 3}
if base_url:
client_kwargs["base_url"] = base_url
client = OpenAI(**client_kwargs)
return client, model
APPLY_EDITS_TOOL = {
"type": "function",
"function": {
"name": "apply_edits",
"description": (
"根据用户的界面定制需求,改写一个或多个前端源文件。"
"只返回真正需要改动的文件;每个文件返回改写后的完整内容。"
),
"parameters": {
"type": "object",
"properties": {
"summary": {
"type": "string",
"description": "用一句话说明本次改了什么(中文)。",
},
"files": {
"type": "array",
"description": "需要改写的文件列表。",
"items": {
"type": "object",
"properties": {
"path": {
"type": "string",
"description": "相对 frontend/ 的文件路径,"
"必须是可编辑文件之一。",
},
"content": {
"type": "string",
"description": "改写后的文件完整内容。",
},
},
"required": ["path", "content"],
},
},
},
"required": ["summary", "files"],
},
},
}
SYSTEM_PROMPT = """你是一个前端界面定制 Agent,负责把用户的自然语言 UI 需求落到 React(Vite) 源码上。
规则:
1. 只能修改用户提供的"可编辑文件",不要新增或删除文件。
2. 优先做最小改动:改颜色/字体/间距等样式,改 theme.css;改文案/组件结构,改 App.jsx。
3. 颜色请使用明确的 CSS 颜色值(如十六进制 #2563eb)。如果用户给了具体色值,就用它。
4. 保持代码可编译:JSX/CSS 语法必须正确,不要破坏原有功能。
5. 必须调用 apply_edits 工具返回结果,files 里给出改写后的完整文件内容。
"""
def read_editable_sources(frontend_dir: Path) -> dict:
"""读取所有可编辑文件当前内容,返回 {相对路径: 内容}。"""
sources = {}
for rel in EDITABLE_FILES:
p = frontend_dir / rel
sources[rel] = p.read_text(encoding="utf-8")
return sources
def customize(client, model, frontend_dir: Path, requirement: str) -> dict:
"""让模型针对一条自然语言需求改写源码,返回 apply_edits 的参数 dict。
仅调用模型并解析工具参数,不落盘(写文件、验证在 demo.py 里做,便于展示 diff)。
"""
sources = read_editable_sources(frontend_dir)
file_blocks = "\n\n".join(
f"===== 文件: {rel} =====\n{content}" for rel, content in sources.items()
)
user_prompt = (
f"可编辑文件当前内容如下:\n\n{file_blocks}\n\n"
f"用户的定制需求:{requirement}\n\n"
f"请调用 apply_edits 返回需要改写的文件全文。"
)
resp = client.chat.completions.create(
model=model,
messages=[
{"role": "system", "content": SYSTEM_PROMPT},
{"role": "user", "content": user_prompt},
],
tools=[APPLY_EDITS_TOOL],
tool_choice={"type": "function", "function": {"name": "apply_edits"}},
temperature=(1 if any(k in (model or "").lower()
for k in ("gpt-5", "o1", "o3", "o4", "thinking", "reasoner", "kimi-k3"))
else 0),
)
msg = resp.choices[0].message
if not msg.tool_calls:
raise RuntimeError("模型没有返回 apply_edits 工具调用。")
raw_args = msg.tool_calls[0].function.arguments or "{}"
try:
args = json.loads(raw_args)
except json.JSONDecodeError:
# Tolerate bad apply_edits JSON; degrade to empty edits.
args = {}
# 安全校验:只允许改写白名单内的文件。
# 要求每个文件项带有字符串 content:否则下游写盘循环会对
# {"path": ...}(缺少 content)抛出 KeyError 而中断整个 demo。
# 形状不合法的项直接丢弃(与非 dict 过滤一致);白名单校验仍会对
# 带 content 但路径非法的项抛错。
files = [
f
for f in (args.get("files") or [])
if isinstance(f, dict) and isinstance(f.get("content"), str)
]
for f in files:
path = f.get("path")
if path not in EDITABLE_FILES:
raise RuntimeError(f"模型试图修改非白名单文件:{path}")
args["files"] = files
return args