前面我们基于Agent继承了Skill，这一期讲讲直接用ChatClient如何集成。 1、增加依赖，增加skill、修改Skill.md、简历上传到minio、发起请求、输出结果等步骤和上一节课程一样，不重复了。
开发LLM相关代码
package cn.hollis.llm.HelloLlm.agent.react.controller;
import cn.hollis.llm.HelloLlm.agent.react.tools.FileReaderTool;
import com.alibaba.cloud.ai.graph.RunnableConfig;
import com.alibaba.cloud.ai.graph.agent.ReactAgent;
import com.alibaba.cloud.ai.graph.agent.hook.shelltool.ShellToolAgentHook;
import com.alibaba.cloud.ai.graph.agent.hook.skills.SkillsAgentHook;
import com.alibaba.cloud.ai.graph.agent.tools.ShellTool2;
import com.alibaba.cloud.ai.graph.checkpoint.savers.MemorySaver;
import com.alibaba.cloud.ai.graph.exception.GraphRunnerException;
import com.alibaba.cloud.ai.graph.skills.registry.SkillRegistry;
import com.alibaba.cloud.ai.graph.skills.registry.classpath.ClasspathSkillRegistry;
import org.springframework.ai.chat.messages.AssistantMessage;
import org.springframework.ai.chat.model.ChatModel;
import org.springframework.ai.support.ToolCallbacks;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.web.bind.annotation.RequestMapping;
import org.springframework.web.bind.annotation.RestController;
import java.util.List;
@RestController
@RequestMapping("/skill")
public class SkillController {
@Autowired
private ChatModel dashScopeChatModel;
@PostConstruct
public void init() {
/*
1. 创建 ClasspathSkillRegistry，从 classpath:skills/ 下加载 Skill
*/
SkillRegistry skillRegistry = ClasspathSkillRegistry.builder()
.classpathPath("skills")
.build();
/*
2. 创建 read_skill 工具的 ToolCallback
这是 LLM 在推理时主动调用的工具，用于按需读取某个 Skill 的完整 SKILL.md 内容
*/
ToolCallback readSkillToolCallback = ReadSkillTool.createReadSkillToolCallback(skillRegistry, null);
/*
3. 创建文件读取工具的 ToolCallback
*/
ToolCallback[] fileReaderToolCallback = ToolCallbacks.from(FileReaderTool.class);
/*
4. 创建 SpringAiSkillAdvisor，把 SkillRegistry 注入进去
Advisor 会在每次对话的 before() 阶段将 Skill 列表追加到 System Prompt
*/
SpringAiSkillAdvisor skillAdvisor = SpringAiSkillAdvisor.builder()
.skillRegistry(skillRegistry)
.build();
/*
5. 创建 ChatClient，并注入 SkillAdvisor 和 read_skill 工具
*/
this.chatClient = ChatClient.builder(chatModel)
.defaultAdvisors(skillAdvisor)
.defaultToolCallbacks(readSkillToolCallback, fileReaderToolCallback[0])
.build();
}
@RequestMapping("/resumeCheck")
public String resumeCheck(String message) throws GraphRunnerException {
return chatClient.prompt()
.user(message)
.call()
.content();
}
}
我们一共给这个ChatClient提供了一个Advisor和一个Tool。 SpringAiSkillAdvisor，把 SkillRegistry 注入进去，这个Advisor 会在每次对话的 before() 阶段将 Skill 列表追加到 System Prompt。 ReadSkillTool，read_skill 工具，这是 LLM 在推理时主动调用的工具，用于按需读取某个 Skill 的完整 SKILL.md 内容。
工具定义
接着，我们需要提供工具，提供一个可以读取文件的工具，包括PDF、MD的读取：
package cn.hollis.llm.mentor.agent.tools;
import org.apache.pdfbox.Loader;
import org.apache.pdfbox.pdmodel.PDDocument;
import org.apache.pdfbox.text.PDFTextStripper;
import org.apache.poi.hssf.usermodel.HSSFWorkbook;
import org.apache.poi.hwpf.HWPFDocument;
import org.apache.poi.hwpf.extractor.WordExtractor;
import org.apache.poi.sl.usermodel.TextShape;
import org.apache.poi.ss.usermodel.*;
import org.apache.poi.xslf.usermodel.XMLSlideShow;
import org.apache.poi.xssf.usermodel.XSSFWorkbook;
import org.apache.poi.xwpf.usermodel.XWPFDocument;
import org.apache.poi.xwpf.usermodel.XWPFParagraph;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.springframework.ai.tool.annotation.Tool;
import org.springframework.ai.tool.annotation.ToolParam;
import org.springframework.core.io.ClassPathResource;
import java.io.File;
import java.io.FileInputStream;
import java.io.IOException;
import java.io.InputStream;
import java.net.HttpURLConnection;
import java.net.URI;
import java.net.URL;
import java.nio.charset.StandardCharsets;
import java.nio.file.Files;
/**
* 通用文件读取工具，供 ReactAgent 调用。
* 支持格式：PDF、Word（docx/doc）、Excel（xlsx/xls）、PowerPoint（pptx）、
* 纯文本（txt/md/csv/log/xml/json/yaml/yml/properties）
*/
public class FileReaderTool {
private static final Logger log = LoggerFactory.getLogger(FileReaderTool.class);
/**
* 单次返回文本的最大字符数，防止超出上下文窗口
*/
private static final int MAX_CHARS = 8000;
@Tool(description = """
读取文件并返回其文本内容。支持从绝对路径、classpath 或网络地址读取。
支持格式：
- PDF (.pdf)
- Word (.docx / .doc)
- Excel (.xlsx / .xls) —— 返回所有 Sheet 的表格文本
- PowerPoint (.pptx) —— 返回所有幻灯片的文字
- 纯文本 (.txt / .md / .csv / .log / .xml / .json / .yaml / .yml / .properties)
参数 path 可以是：
- 文件的绝对路径，如：/Users/hollis/data/sample.txt
- classpath 路径，以 classpath: 开头，如：classpath:templates/sample.txt
- 网络地址，以 http:// 或 https:// 开头，如：http://localhost:9001/api/v1/download-shared-object/xxx
对于 PDF 文件，可选参数 startPage / endPage 指定读取页码范围（从 1 开始）；其他格式忽略此参数。
若文本过长会自动截断并在末尾注明剩余字符数。
""")
public String read_file(
@ToolParam(description = "文件路径，可以是绝对路径、classpath: 开头或 http(s):// 开头的 URL，如 /Users/hollis/data/sample.txt、classpath:templates/sample.txt 或 http://example.com/file.pdf") String path,
@ToolParam(required = false, description = "【仅 PDF 有效】起始页码（从 1 开始，默认第 1 页）") Integer startPage,
@ToolParam(required = false, description = "【仅 PDF 有效】结束页码（含，默认最后一页）") Integer endPage) {
log.info("Starting execution of tool: read_file, path: {}, startPage: {}, endPage: {}", path, startPage, endPage);
if (path == null || path.isBlank()) {
return "Error: path 不能为空";
}
// 判断路径类型
if (path.startsWith("classpath:")) {
String resourcePath = path.substring("classpath:".length());
return readFromClasspath(resourcePath, startPage, endPage);
} else if (path.startsWith("http://") || path.startsWith("https://")) {
return readFromUrl(path, startPage, endPage);
} else {
return readFromFilesystem(path, startPage, endPage);
}
}
private String readFromClasspath(String resourcePath, Integer startPage, Integer endPage) {
ClassPathResource resource = new ClassPathResource(resourcePath);
if (!resource.exists()) {
return "Error: classpath 资源不存在 -> " + resourcePath;
}
String ext = getExtension(resourcePath).toLowerCase();
try (InputStream is = resource.getInputStream()) {
return switch (ext) {
case "pdf" -> readPdfFromStream(is, startPage, endPage);
case "docx" -> readDocxFromStream(is);
case "doc" -> readDocFromStream(is);
case "xlsx" -> readExcelFromStream(is, false);
case "xls" -> readExcelFromStream(is, true);
case "pptx" -> readPptxFromStream(is);
default -> readTextFromStream(is, ext);
};
} catch (Exception e) {
log.error("读取 classpath 文件失败: {}", resourcePath, e);
return "Error: 读取 classpath 文件失败 -> " + e.getMessage();
}
}
private String readFromFilesystem(String filePath, Integer startPage, Integer endPage) {
File file = new File(filePath);
if (!file.exists()) {
return "Error: 文件不存在 -> " + filePath;
}
if (!file.canRead()) {
return "Error: 文件无读取权限 -> " + filePath;
}
String ext = getExtension(file.getName()).toLowerCase();
try {
return switch (ext) {
case "pdf" -> readPdf(file, startPage, endPage);
case "docx" -> readDocx(file);
case "doc" -> readDoc(file);
case "xlsx" -> readExcel(file, false);
case "xls" -> readExcel(file, true);
case "pptx" -> readPptx(file);
default -> readText(file);
};
} catch (Exception e) {
log.error("读取文件失败: {}", filePath, e);
return "Error: 读取文件失败 -> " + e.getMessage();
}
}
private String readFromUrl(String urlString, Integer startPage, Integer endPage) {
try {
URL url = URI.create(urlString).toURL();
HttpURLConnection connection = (HttpURLConnection) url.openConnection();
connection.setRequestMethod("GET");
connection.setConnectTimeout(30000); // 30秒连接超时
connection.setReadTimeout(60000); // 60秒读取超时
connection.setInstanceFollowRedirects(true);
int responseCode = connection.getResponseCode();
if (responseCode != HttpURLConnection.HTTP_OK) {
return "Error: HTTP 请求失败，状态码 -> " + responseCode;
}
// 尝试从 Content-Disposition 头获取文件名
String contentDisposition = connection.getHeaderField("Content-Disposition");
String filename = null;
if (contentDisposition != null && contentDisposition.contains("filename=")) {
filename = contentDisposition.replaceAll(".*filename=\"([^\"]+)\".*", "$1");
}
// 如果无法从头获取，从 URL 路径获取
if (filename == null || filename.isBlank()) {
String path = url.getPath();
int lastSlash = path.lastIndexOf('/');
filename = (lastSlash >= 0 && lastSlash < path.length() - 1) ? path.substring(lastSlash + 1) : "download";
}
String ext = getExtension(filename).toLowerCase();
try (InputStream is = connection.getInputStream()) {
return switch (ext) {
case "pdf" -> readPdfFromStream(is, startPage, endPage);
case "docx" -> readDocxFromStream(is);
case "doc" -> readDocFromStream(is);
case "xlsx" -> readExcelFromStream(is, false);
case "xls" -> readExcelFromStream(is, true);
case "pptx" -> readPptxFromStream(is);
default -> readTextFromStream(is, ext);
};
}
} catch (Exception e) {
log.error("从网络地址读取文件失败: {}", urlString, e);
return "Error: 从网络地址读取文件失败 -> " + e.getMessage();
}
}
// -------------------------------------------------------------------------
// PDF
// -------------------------------------------------------------------------
private String readPdf(File file, Integer startPage, Integer endPage) throws IOException {
try (PDDocument doc = Loader.loadPDF(file)) {
return readPdfDocument(doc, startPage, endPage);
}
}
private String readPdfFromStream(InputStream is, Integer startPage, Integer endPage) throws IOException {
try (PDDocument doc = Loader.loadPDF(is.readAllBytes())) {
return readPdfDocument(doc, startPage, endPage);
}
}
private String readPdfDocument(PDDocument doc, Integer startPage, Integer endPage) throws IOException {
int total = doc.getNumberOfPages();
int from = (startPage != null && startPage >= 1) ? startPage : 1;
int to = (endPage != null && endPage >= 1) ? Math.min(endPage, total) : total;
if (from > total) {
return String.format("Error: startPage(%d) 超过文件总页数(%d)", from, total);
}
PDFTextStripper stripper = new PDFTextStripper();
stripper.setStartPage(from);
stripper.setEndPage(to);
stripper.setSortByPosition(true);
String text = stripper.getText(doc);
String header = String.format("[PDF 共 %d 页，本次读取第 %d-%d 页]\n\n", total, from, to);
return header + truncate(text);
}
// -------------------------------------------------------------------------
// Word docx
// -------------------------------------------------------------------------
private String readDocx(File file) throws IOException {
try (XWPFDocument doc = new XWPFDocument(new FileInputStream(file))) {
return readDocxDocument(doc);
}
}
private String readDocxFromStream(InputStream is) throws IOException {
try (XWPFDocument doc = new XWPFDocument(is)) {
return readDocxDocument(doc);
}
}
private String readDocxDocument(XWPFDocument doc) {
StringBuilder sb = new StringBuilder();
for (XWPFParagraph para : doc.getParagraphs()) {
String text = para.getText();
if (text != null && !text.isBlank()) {
sb.append(text).append("\n");
}
}
return "[Word (.docx)]\n\n" + truncate(sb.toString());
}
// -------------------------------------------------------------------------
// Word doc (旧格式)
// -------------------------------------------------------------------------
private String readDoc(File file) throws IOException {
try (HWPFDocument doc = new HWPFDocument(new FileInputStream(file));
WordExtractor extractor = new WordExtractor(doc)) {
return readDocExtractor(extractor);
}
}
private String readDocFromStream(InputStream is) throws IOException {
try (HWPFDocument doc = new HWPFDocument(is);
WordExtractor extractor = new WordExtractor(doc)) {
return readDocExtractor(extractor);
}
}
private String readDocExtractor(WordExtractor extractor) {
String text = String.join("\n", extractor.getParagraphText());
return "[Word (.doc)]\n\n" + truncate(text);
}
// -------------------------------------------------------------------------
// Excel xlsx / xls
// -------------------------------------------------------------------------
private String readExcel(File file, boolean isOld) throws IOException {
try (InputStream is = new FileInputStream(file);
Workbook wb = isOld ? new HSSFWorkbook(is) : new XSSFWorkbook(is)) {
return readExcelWorkbook(wb, isOld);
}
}
private String readExcelFromStream(InputStream is, boolean isOld) throws IOException {
try (Workbook wb = isOld ? new HSSFWorkbook(is) : new XSSFWorkbook(is)) {
return readExcelWorkbook(wb, isOld);
}
}
private String readExcelWorkbook(Workbook wb, boolean isOld) {
StringBuilder sb = new StringBuilder();
DataFormatter formatter = new DataFormatter();
for (int si = 0; si < wb.getNumberOfSheets(); si++) {
Sheet sheet = wb.getSheetAt(si);
sb.append("=== Sheet: ").append(sheet.getSheetName()).append(" ===\n");
for (Row row : sheet) {
StringBuilder rowSb = new StringBuilder();
for (Cell cell : row) {
if (rowSb.length() > 0) rowSb.append("\t");
rowSb.append(formatter.formatCellValue(cell));
}
String rowStr = rowSb.toString().trim();
if (!rowStr.isEmpty()) {
sb.append(rowStr).append("\n");
}
}
sb.append("\n");
}
String ext = isOld ? ".xls" : ".xlsx";
return "[Excel (" + ext + ") 共 " + wb.getNumberOfSheets() + " 个 Sheet]\n\n" + truncate(sb.toString());
}
// -------------------------------------------------------------------------
// PowerPoint pptx
// -------------------------------------------------------------------------
private String readPptx(File file) throws IOException {
try (XMLSlideShow ppt = new XMLSlideShow(new FileInputStream(file))) {
return readPptxShow(ppt);
}
}
private String readPptxFromStream(InputStream is) throws IOException {
try (XMLSlideShow ppt = new XMLSlideShow(is)) {
return readPptxShow(ppt);
}
}
private String readPptxShow(XMLSlideShow ppt) {
StringBuilder sb = new StringBuilder();
int slideNum = 1;
for (var slide : ppt.getSlides()) {
sb.append("--- 第 ").append(slideNum++).append(" 页 ---\n");
for (var shape : slide.getShapes()) {
if (shape instanceof TextShape<?, ?> ts) {
String text = ts.getText();
if (text != null && !text.isBlank()) {
sb.append(text).append("\n");
}
}
}
sb.append("\n");
}
return "[PowerPoint (.pptx) 共 " + ppt.getSlides().size() + " 页]\n\n" + truncate(sb.toString());
}
// -------------------------------------------------------------------------
// 纯文本（txt/md/csv/log/xml/json/yaml/yml/properties 等）
// -------------------------------------------------------------------------
private String readText(File file) throws IOException {
String content = Files.readString(file.toPath(), StandardCharsets.UTF_8);
String ext = getExtension(file.getName()).toUpperCase();
return "[文本文件 (." + ext.toLowerCase() + ")]\n\n" + truncate(content);
}
private String readTextFromStream(InputStream is, String ext) throws IOException {
String content = new String(is.readAllBytes(), StandardCharsets.UTF_8);
return "[文本文件 (." + ext.toLowerCase() + ")]\n\n" + truncate(content);
}
// -------------------------------------------------------------------------
// 工具方法
// -------------------------------------------------------------------------
private String truncate(String text) {
if (text == null) return "";
if (text.length() <= MAX_CHARS) return text;
int remaining = text.length() - MAX_CHARS;
return text.substring(0, MAX_CHARS)
+ String.format("\n\n...[内容过长，已截断，剩余约 %d 字符未显示，请指定更小的范围继续读取]", remaining);
}
private String getExtension(String filename) {
int dot = filename.lastIndexOf('.');
return (dot >= 0 && dot < filename.length() - 1) ? filename.substring(dot + 1) : "";
}
}
以上是一个新的文件读取工具，我们在之前的工具的基础上，让他同时能够读取classpath下的文件和网络文件，这样能解决他无法直接读取classpath下的jd.md和我们上传后的简历Pdf的问题，因为之前这个工具只能从绝对路径读取文件。