序
本文主要研究一下Spring AI Alibaba的YuQueDocumentReader
YuQueDocumentReader
community/document-readers/spring-ai-alibaba-starter-document-reader-yuque/src/main/java/com/alibaba/cloud/ai/reader/yuque/YuQueDocumentReader.java
public class YuQueDocumentReader implements DocumentReader {private final DocumentParser parser;private final YuQueResource yuQueResource;public YuQueDocumentReader(YuQueResource yuQueResource, DocumentParser parser) {this.yuQueResource = yuQueResource;this.parser = parser;}@Overridepublic List<Document> get() {try {List<Document> documents = parser.parse(yuQueResource.getInputStream());String source = yuQueResource.getResourcePath();for (Document doc : documents) {doc.getMetadata().put(YuQueResource.SOURCE, source);}return documents;}catch (IOException ioException) {throw new RuntimeException("Failed to load document from yuque: {}", ioException);}}}
YuQueDocumentReader構造器要求輸入YuQueResource、DocumentParser，其get方法通過DocumentParser解析，最后在其metadata追加一個SOURCE
YuQueResource
community/document-readers/spring-ai-alibaba-starter-document-reader-yuque/src/main/java/com/alibaba/cloud/ai/reader/yuque/YuQueResource.java
public class YuQueResource implements Resource {private static final String BASE_URL = "https://www.yuque.com";private static final String INFO_PATH = "/api/v2/hello";private static final String DOC_DETAIL_PATH = "/api/v2/repos/%s/%s/docs/%s";public static final String SOURCE = "source";public static final String SUPPORT_TYPE = "Doc";private final HttpClient httpClient;private final InputStream inputStream;private final URI uri;private final String resourcePath;private String groupLogin;private String bookSlug;private String id;public YuQueResource(String yuQueToken, String resourcePath) {this.resourcePath = resourcePath;this.httpClient = HttpClient.newBuilder().version(HttpClient.Version.HTTP_2).build();judgePathRule(resourcePath);judgeToken(yuQueToken);URI baseUri = URI.create(BASE_URL + DOC_DETAIL_PATH.formatted(groupLogin, bookSlug, id));HttpRequest httpRequest = HttpRequest.newBuilder().header("X-Auth-Token", yuQueToken).uri(baseUri).GET().build();try {HttpResponse<String> response = this.httpClient.send(httpRequest, HttpResponse.BodyHandlers.ofString());String body = response.body();// Parse the JSON response using JacksonObjectMapper objectMapper = new ObjectMapper();JsonNode jsonObject = objectMapper.readTree(body);JsonNode dataObject = jsonObject.get("data");if (dataObject == null || !dataObject.isObject()) {throw new RuntimeException("Invalid response format: 'data' is not an object");}if (!Objects.equals(dataObject.get("type").asText(), SUPPORT_TYPE)) {throw new RuntimeException("Unsupported resource type, only support " + SUPPORT_TYPE);}inputStream = new ByteArrayInputStream(dataObject.get("body_html").asText().getBytes());uri = URI.create(resourcePath);}catch (Exception e) {throw new RuntimeException(e);}}/*** Judge resource path rule Official online doc* https://www.yuque.com/yuque/developer/openapi* @param resourcePath*/private void judgePathRule(String resourcePath) {// Determine if the path conforms to this format： https://xx.xxx.com/aa/bb/ccString regex = "^https://[a-zA-Z0-9.-]+/([a-zA-Z0-9.-]+)/([a-zA-Z0-9.-]+)/([a-zA-Z0-9.-]+)$";Pattern pattern = Pattern.compile(regex);Matcher matcher = pattern.matcher(resourcePath);Assert.isTrue(matcher.matches(), "Invalid resource path");// Extract the captured groupsthis.groupLogin = matcher.group(1);this.bookSlug = matcher.group(2);this.id = matcher.group(3);Assert.isTrue(StringUtils.hasText(this.groupLogin), "Invalid resource path");Assert.isTrue(StringUtils.hasText(this.bookSlug), "Invalid resource path");Assert.isTrue(StringUtils.hasText(this.id), "Invalid resource path");}/*** judge yuQue token* @param yuQueToken User/Team token*/private void judgeToken(String yuQueToken) {URI uri = URI.create(BASE_URL + INFO_PATH);HttpRequest httpRequest = HttpRequest.newBuilder().header("X-Auth-Token", yuQueToken).uri(uri).GET().build();try {HttpResponse<String> response = this.httpClient.send(httpRequest, HttpResponse.BodyHandlers.ofString());int statusCode = response.statusCode();Assert.isTrue(statusCode == 200, "Failed to auth YuQueToken");}catch (Exception e) {throw new RuntimeException(e);}}//......
}
YuQueResource的構造器要求輸入yuQueToken和resourcePath，它通過httpClient請求
https://www.yuque.com/api/v2/repos/{groupLogin}/{bookSlug}/docs/{id}，之后解析body_html到inputStream；其中groupLogin、bookSlug、id是judgePathRule通過解析resourcePath提取出來
示例
community/document-readers/spring-ai-alibaba-starter-document-reader-yuque/src/test/java/com/alibaba/cloud/ai/reader/yuque/YuQueDocumentLoaderIT.java
@EnabledIfEnvironmentVariable(named = "YUQUE_TOKEN", matches = ".+")
@EnabledIfEnvironmentVariable(named = "YUQUE_RESOURCE_PATH", matches = ".+")
class YuQueDocumentLoaderIT {private static final String YU_QUE_TOKEN = System.getenv("YUQUE_TOKEN");private static final String RESOURCE_PATH = System.getenv("YUQUE_RESOURCE_PATH");YuQueDocumentReader reader;YuQueResource source;static {if (YU_QUE_TOKEN == null || RESOURCE_PATH == null) {System.out.println("YUQUE_TOKEN or YUQUE_RESOURCE_PATH environment variable is not set. Tests will be skipped.");}}@BeforeEachpublic void beforeEach() {// Skip test if environment variables are not setAssumptions.assumeTrue(YU_QUE_TOKEN != null && !YU_QUE_TOKEN.isEmpty(),"Skipping test because YUQUE_TOKEN is not set");Assumptions.assumeTrue(RESOURCE_PATH != null && !RESOURCE_PATH.isEmpty(),"Skipping test because YUQUE_RESOURCE_PATH is not set");source = YuQueResource.builder().yuQueToken(YU_QUE_TOKEN).resourcePath(RESOURCE_PATH).build();reader = new YuQueDocumentReader(source, new TikaDocumentParser());}@Testpublic void should_load_file() {// Skip test if reader is not initializedAssumptions.assumeTrue(reader != null, "Skipping test because reader is not initialized");List<Document> document = reader.get();String content = document.get(0).getText();System.out.println(content);}}
小結
spring-ai-alibaba-starter-document-reader-yuque提供了YuQueDocumentReader，它通過YuQueResource去請求資源，再通過DocumentParser解析(比如TikaDocumentParser)為Document，最后追加一個SOURCE的metadata。
doc
- java2ai
