Skills resources

apiVersion: skill.ooder.net/v1 kind: Skill

metadata: id: skill-document-processor name: 文档处理器 version: 2.3 description: 文档处理服务 - 支持多格式文档解析(TXT/MD/PDF/DOCX)、智能分块、元数据提取、Token估算 author: ooder Team type: utility-service license: Apache-2.0 homepage: https://gitee.com/ooderCN/skills keywords: - document - parsing - chunking - rag - knowledge

spec: skillForm: PROVIDER type: service-skill

capability: category: knowledge

ownership: independent

supportedSceneTypes: - document-assistant - knowledge-qa - meeting-minutes - project-knowledge

dynamicSceneTypes: true

autoStart: enabled: true delay: 3s

autoJoin: enabled: true matchSceneTypes: true maxSceneGroups: 15

dependencies: []

providedInterfaces: - id: document-processing version: "1.0" description: "文档处理接口" - id: text-chunking version: "1.0" description: "文本分块接口"

supportedFormats: - extension: .txt mimeType: text/plain description: 纯文�? - extension: .md mimeType: text/markdown description: Markdown文档 - extension: .pdf mimeType: application/pdf description: PDF文档 - extension: .docx mimeType: application/vnd.openxmlformats-officedocument.wordprocessingml.document description: Word文档 - extension: .html mimeType: text/html description: HTML文档

capabilities: - id: document-parsing name: 文档解析 description: 支持TXT/MD/PDF/DOCX/HTML多种格式文档解析 category: document - id: text-chunking name: 文本分块 description: 智能分块，支持固定大�?语义/句子边界分块策略 category: document - id: metadata-extraction name: 元数据提�? description: 提取标题、关键词、摘要、作者等元数�? category: document - id: token-estimation name: Token估算 description: 估算文本Token数量，支持多种模�? category: document - id: encoding-detection name: 编码检�? description: 自动检测文档编码格�? category: document

endpoints: - path: /api/v1/document/parse method: POST description: 解析文档 capability: document-parsing - path: /api/v1/document/chunk method: POST description: 文本分块 capability: text-chunking - path: /api/v1/document/metadata method: POST description: 提取元数�? capability: metadata-extraction - path: /api/v1/document/token-count method: POST description: 计算Token�? capability: token-estimation - path: /api/v1/document/formats method: GET description: 获取支持的格�? capability: document-parsing

chunkingStrategies: - FIXED_SIZE - SEMANTIC - SENTENCE_BOUNDARY - PARAGRAPH

runtime: language: java javaVersion: "8" framework: spring-boot

llmConfig: required: false defaultProvider: "deepseek" defaultModel: "deepseek-chat" capabilities: - chat - streaming - function-calling modelSelection: allowUserOverride: true availableProviders: - deepseek - openai - qianwen - volcengine - ollama functionCalling: enabled: true tools: - name: query_skill_capability description: "查询当前技能的能力和使用方�? parameters: type: object properties: capability: type: string description: "能力名称" detail: type: string enum: [brief, detailed, examples] default: "brief" - name: execute_mvel_action description: "通过MVEL表达式执行后台操�? parameters: type: object properties: expression: type: string description: "MVEL表达�? context: type: object - name: generate_ui_form description: "生成UI表单供用户填�? parameters: type: object properties: formType: type: string fields: type: array items: type: object - name: execute_batch_operation description: "执行批量操作" parameters: type: object properties: operation: type: string items: type: array items: type: object - name: convert_to_javascript description: "转换为JavaScript代码供用户使�? parameters: type: object properties: action: type: string parameters: type: object toolChoice: auto rateLimits: requestsPerMinute: 60 tokensPerMinute: 100000

config: optional: - name: CHUNK_SIZE type: integer default: 500 description: 分块大小(字符�? - name: CHUNK_OVERLAP type: integer default: 50 description: 分块重叠(字符�? - name: MAX_FILE_SIZE type: integer default: 10485760 description: 最大文件大�?字节) - name: DEFAULT_ENCODING type: string default: "UTF-8" description: 默认编码 - name: CHUNKING_STRATEGY type: string default: "FIXED_SIZE" description: 分块策略

resources: cpu: "200m" memory: "256Mi" storage: "100Mi"

offline: enabled: true cacheStrategy: local syncOnReconnect: true

knowledge: documents: - id: overview name: 功能概述 path: README.md type: guide language: zh priority: high ragConfig: enabled: true indexName: "skill-document-processor-knowledge" embeddingModel: text-embedding-3-small chunkSize: 1000 chunkOverlap: 200 searchStrategy: hybrid topK: 5 threshold: 0.7