Kernel Memory 5.2 코드 분석2
웹 서비스로 RAG 서비스를 제공하는 것이나 이것과 연동되어 서비스를 제공하는 것이나, 할 일로 보면 같습니다.
RAG를 위한 기능들을 제공하는 것을 MemoryService라고 하면, MemoryWebClient는 MemoryService와 연동되어 RAG를 위한 기능들을 제공하는 것입니다. 실제적으로 하는 일은 다를 수 있지만, 외부에서 보면 둘 다 같은 일을 하는 같은 것으로 보일 겁니다. 같은 역할을 하는 것으로 보이는 겁니다 . 커널 메모리에서는 이 역할을 IKernelMemory라고 합니다.
MemoryService, MemoryServerless, MemoryWebClient 모두 IKernelMemory 인터페이스를 구현하고 있습니다. 커널 메모리는 독립 서버에서 실행되는 웹 서비스를 기본으로 하지만, 이것과 연동되는 AI 애플리케이션에서 개발 과정 중에 쉽게 사용할 수 있도록 서버가 아닌 형태도 지원합니다. 둘 을 구분해서 각각 MemoryService, MemoryServerless라고 합니다.
IKernelMemory 문서화 합니다.
뭐든 첫 출발이 중요합니다. 아래 내용이 “완전히 이해 되었다.”할 때 까지 반복해서 여러 번 읽기를 추천합니다. 특히 개발자이신 분들은 더.
This code defines an interface called IKernelMemory that allows you to interact with a system that stores and processes documents, files, and memories (more on that later). Think of it like a library where you can import, search, and manage documents and their contents.
**Key Components:**
1. **Documents**: These are the files or content you want to import and store in the system. Documents can have tags, user IDs, and other metadata.
2. **Indexes**: An index is like a catalog or a folder that organizes and stores related documents. You can think of it as a library shelf where similar documents are stored together.
3. **Memories**: These are the individual pieces of information extracted from the documents, like keywords, phrases, or concepts. Memories are generated by processing the documents.
**Methods (Functions) Explained:**
1. **ImportDocumentAsync**: This method allows you to import a document (file or content) into the system. You can specify the document’s details, tags, and index where it should be stored. It returns a document ID.
2. **ListIndexesAsync**: This method lists all the available indexes in the system.
3. **DeleteIndexAsync**: This method deletes an entire index, removing all documents and memories associated with it.
4. **DeleteDocumentAsync**: This method deletes a specific document and its related memories from the system.
5. **IsDocumentReadyAsync**: This method checks if a document has been successfully uploaded and processed, returning a boolean value.
6. **GetDocumentStatusAsync**: This method retrieves information about an uploaded document, such as its processing status.
7. **ExportFileAsync**: This method exports a file from the document storage, allowing you to retrieve the file content.
8. **SearchAsync**: This method searches the specified index for documents that match a query, using filters and relevance scoring. It returns a list of relevant documents.
9. **AskAsync**: This method searches the specified index for an answer to a question, using filters and relevance scoring. It returns an answer to the question, if possible.
**In Simple Terms:**
Imagine a digital library where you can store and manage documents, files, and memories. This interface allows you to:
* Import documents and files into the library
* Organize documents into indexes (like folders)
* Search for specific documents or answers to questions
* Delete documents or entire indexes
* Check the status of uploaded documents
* Export files from the library
* And more!
