2024-01-12 16:24:14 +08:00
|
|
|
|
# Agents-Flex ChangeLog
|
2024-05-12 13:35:55 +08:00
|
|
|
|
|
2024-06-17 09:08:27 +08:00
|
|
|
|
|
2024-07-14 17:40:33 +08:00
|
|
|
|
## v1.0.0-beta.8 20240714
|
|
|
|
|
- feat: add "async" flag for the ChainNode
|
|
|
|
|
- feat: add Ollama LLM
|
|
|
|
|
- feat: add DnjsonClient for OllamaLlm
|
|
|
|
|
- refactor: refactor ChainCondition.java
|
|
|
|
|
- refactor: add throw LlmException if LLMs has error
|
|
|
|
|
- refactor: refactor DocumentParser
|
|
|
|
|
- refactor: refactor chain module
|
|
|
|
|
- refactor: rename GroovyExecNode.java and QLExpressExecNode.java
|
|
|
|
|
- refactor: add children property in Parameter
|
|
|
|
|
- refactor: remove unused code AsyncHttpClient.java
|
|
|
|
|
- refactor: use LlmException to replace LLMClientException
|
|
|
|
|
fix: Milvus type mismatch for filed 'id'
|
|
|
|
|
- test: add LoopChain test
|
|
|
|
|
- test: add ollama test use openai instance
|
|
|
|
|
- docs: add Japanese README
|
|
|
|
|
|
|
|
|
|
---
|
|
|
|
|
- 新增:为 ChainNode 添加 "async" 属性标识的设置
|
|
|
|
|
- 新增:添加基于 Ollama 大语言模型的对接,非 openai 适配模式
|
|
|
|
|
- 新增:新增 DnjsonClient 用于和 Ollama 的 stream 模型对接
|
|
|
|
|
- 优化:重构 ChainCondition
|
|
|
|
|
- 优化:chat 时当大语言模型发生错误时抛出异常,之前返回 null
|
|
|
|
|
- 优化:重构 DocumentParser
|
|
|
|
|
- 优化:Parameter 支持子参数的配置能力
|
|
|
|
|
- 修复:Milvus 向量数据库当传入 number 类型是出错的问题
|
|
|
|
|
- 测试:添加对 LoopChain 的测试
|
|
|
|
|
- 测试:添加文使用 openai 兼容 api 对 Ollama 对接的测试
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-07-05 08:57:20 +08:00
|
|
|
|
## v1.0.0-beta.7 20240705
|
|
|
|
|
- feat: add image models support
|
|
|
|
|
- feat: add SimpleTokenizeSplitter
|
|
|
|
|
- feat: add OmniParseDocumentParser
|
|
|
|
|
- feat: add openai,stability AI and gitee-sd3 AI support
|
|
|
|
|
- feat: add moonshot support
|
|
|
|
|
- feat: add chain dsl support
|
|
|
|
|
- refactor: optimize llm clients
|
|
|
|
|
- refactor: optimize SparkLLM
|
|
|
|
|
- refactor: optimize slf4j dependencies
|
|
|
|
|
- refactor: optimize Agent define
|
|
|
|
|
- refactor: optimize chain
|
|
|
|
|
- test: add .pdf and .doc parse test
|
|
|
|
|
- test: add SimpleDocumentSplitterTest.java
|
|
|
|
|
|
|
|
|
|
---
|
|
|
|
|
- 新增:新增图片模型的支持
|
|
|
|
|
- 新增:新增 SimpleTokenizeSplitter 分割器
|
|
|
|
|
- 新增:新增 OmniParseDocumentParser 文档解析器
|
|
|
|
|
- 新增:新增 openai、stability ai 以及 gitee ai 对图片生成的支持
|
|
|
|
|
- 新增:新增月之暗面的支持
|
|
|
|
|
- 优化:优化 llm 客户端的细节
|
|
|
|
|
- 优化:优化星火大模型的细节
|
|
|
|
|
- 优化:优化 slf4j 依赖的细节
|
|
|
|
|
- 优化:优化 Agent 和 Chain 的定义细节
|
|
|
|
|
- 测试:添加 .pdf 和 .doc 的解析测试
|
|
|
|
|
- 测试:添加文档分割器的测试
|
|
|
|
|
- 测试:添加 token 文档分割器的测试
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-06-17 09:08:27 +08:00
|
|
|
|
## v1.0.0-beta.5 20240617
|
|
|
|
|
- feat: add ImagePrompt to send image to LLM
|
|
|
|
|
- feat: chatOptions add topP/topK and stop config
|
|
|
|
|
- refactor: rename TextMessage.java to AbstractTextMessage.java
|
|
|
|
|
- refactor: refactor llm methods
|
|
|
|
|
- refactor: refactor FunctionMessageResponse.java
|
|
|
|
|
- refactor: optimize HttpClient.java And SseClient.java
|
|
|
|
|
- fix: fix function calling error in QwenLLM
|
|
|
|
|
- test: add chat with image test
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
---
|
|
|
|
|
- 新增:新增 ImagePrompt 用于发送图片对话的场景
|
|
|
|
|
- 新增:对话模型下的 ChatOptions 添加 topK 和 topP 配置的支持
|
|
|
|
|
- 优化:重命名 TextMessage 为 AbstractTextMessage
|
|
|
|
|
- 优化:重构 LLM 的方法定义,使之更加简单易用
|
|
|
|
|
- 优化:优化 HttpClient.java 和 SseClient.java 的相关代码
|
|
|
|
|
- 修复:通义千问 QwenLLM 在 function calling 下无法正常调用的问题
|
|
|
|
|
- 测试:添加发送图片相关的测试内容
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-05-31 16:10:34 +08:00
|
|
|
|
## v1.0.0-beta.4 20240531
|
|
|
|
|
- feat: OpenAiLlm support embedding model config
|
|
|
|
|
- feat: add get dimensions method in EmbeddingModel
|
|
|
|
|
- feat: SparkLlm support embedding
|
|
|
|
|
- feat: optimize MilvusVectorStore
|
|
|
|
|
- feat: MilvusVectorStore add username and password config
|
|
|
|
|
- refactor: optimize HttpClient.java
|
|
|
|
|
- refactor: optimize AliyunVectorStore
|
|
|
|
|
- refactor: update StoreOptions to extends Metadata
|
|
|
|
|
- refactor: optimize StoreResult and Metadata
|
|
|
|
|
- fix: fix AIMessage tokens parse
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
---
|
|
|
|
|
- 新增:OpenAiLlm 添加自定义 embedding 模型的支持
|
|
|
|
|
- 新增:EmbeddingModel 添加获取向量维度的支持
|
|
|
|
|
- 新增:SparkLlm 星火大模型添加对 embedding 的支持
|
|
|
|
|
- 新增:添加 Milvus 向量数据库的支持
|
|
|
|
|
- 优化:优化 HttpClient 的代码
|
|
|
|
|
- 优化:优化 AliyunVectorStore 的代码
|
|
|
|
|
- 优化:优化 StoreOptions 和 Metadata 的代码
|
|
|
|
|
- 修复:AIMessage 的 tokens 消耗解析不正确的问题
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-05-16 17:39:02 +08:00
|
|
|
|
## v1.0.0-beta.3 20240516
|
|
|
|
|
- feat:add "description" to agent for automatic arrangement by LLM
|
|
|
|
|
- feat: StoreResult can return ids if document store success
|
|
|
|
|
- feat: StoreOptions support set multi partitionNames
|
|
|
|
|
- feat: add DocumentIdGenerator for Document and Store
|
|
|
|
|
- feat: add ChainException for Chain.executeForResult
|
|
|
|
|
- refactor: rename "SimplePrompt" to "TextPrompt"
|
|
|
|
|
|
|
|
|
|
---
|
|
|
|
|
- 新增:为 Agent 添加 description 属性,方便用于 AI 自动编排的场景
|
|
|
|
|
- 新增:Agent 添加对 outputKeys 的定义支持
|
|
|
|
|
- 新增:添加 DocumentIdGenerator 用于在对文档存储时自动生成 id 的功能
|
|
|
|
|
- 新增:StoreOptions 添加多个 partitionName 配置的支持
|
|
|
|
|
- 新增:当 Document 保存成功时,自动返回保存的 id
|
|
|
|
|
- 优化:Chain.executeForResult 会抛出异常 ChainException
|
|
|
|
|
- 修复:ChatGLM 的 Chat JSON 解析错误的问题
|
|
|
|
|
- 测试:优化 SparkLlmTest 的测试代码
|
2024-05-12 13:35:55 +08:00
|
|
|
|
- 文档:完善基础文档
|