LangChain4j Features LangChain4j 特性:
Integration with 15+ LLM providers 集成15个以上的LLM
Integration with 20+ embedding (vector) stores 集成20个以上的向量存储
Integration with 15+ embedding models 向量模型
Integration with 5 image generation models 图片集成模型
Integration with 2 scoring (re-ranking) models 评分模型
Integration with one moderation model (OpenAI) 集成Open AI
Support for texts and images as inputs (multimodality) 支持文本和图片作为输入(多模态)
AI Services (high-level LLM API)
Prompt templates 提示词模板
Implementation of persistent and in-memory chat memory algorithms: message window and token window
Streaming of responses from LLMs 基于LLM的流式输出,就是一个字一个字输出
Output parsers for common Java types and custom POJOs
Tools (function calling) 函数调用TOOLS
Dynamic Tools (execution of dynamically generated LLM code)
RAG (Retrieval-Augmented-Generation):增强检索生成
Ingestion: 嵌入集成
Importing various types of documents (TXT, PDFs, DOC, PPT, XLS etc.) from multiple sources (file system, URL, GitHub, Azure Blob Storage, Amazon S3, etc.)
从多个来源(文件系统、URL、GitHub、Azure Blob Storage、Amazon S3等)导入各种类型的文档(TXT、PDF、DOC、PPT、XLS等)
Splitting documents into smaller segments using multiple splitting algorithms
使用多种分割算法将文档分割成更小的片段
Post-processing of documents and segments
文档和片段的Post处理
Embedding segments using embedding models
使用嵌入模型对片段进行嵌入
Storing embeddings in embedding (vector) store
将嵌入向量存储在嵌入(向量)存储中
Retrieval (simple and advanced):
检索(简单与高级):
Transformation of queries (expansion, compression) 查询转换(扩展、压缩)
Routing of queries 查询路由
Retrieving from vector store and/or any custom sources 从向量存储和/或任何自定义源中检索
Re-ranking 重新排序
Reciprocal Rank Fusion 互信息秩融合
Customization of each step in the RAG flow RAG流程中每个步骤的定制
Text classification 文本分类
Tools for tokenization and estimation of token counts 分词工具和词数估算工具
Kotlin Extensions: Asynchronous non-blocking handling of chat interactions using Kotlin's coroutine capabilities.
Kotlin扩展:利用Kotlin的协程功能,