feat: RAG engine plugin architecture * feat: RAG engine plugin architecture (#29) - Add RAG engine component type with full lifecycle support - Implement embedding, vector storage, and retrieval proxy APIs - Add RAG-related host service methods for plugins - Update SDK templates and CLI for RAG engine scaffolding - Bump version to 0.3.0-alpha.1 * feat: normalize the names of rag related actions * refactor(rag): remove redundant builtin fields, unify settings via schema Remove RetrievalConfig class and redundant IngestionContext fields (chunking_strategy, chunk_size, chunk_overlap) since all parameters can be covered by the retrieval_settings_schema / creation_schema mechanism through custom_settings dicts. - Delete RetrievalConfig class and its export - Remove config field and get_top_k() from RetrievalContext, use top_k directly - Add retrieval_settings field to RetrievalContext (aligned with host) - Rename IngestionContext.custom_settings to creation_settings (aligned with host) - Update template to use context.top_k instead of context.get_top_k() Co-Authored-By: Claude Opus 4.6 * refactor(rag): remove standalone top_k from RetrievalContext top_k is now passed via retrieval_settings dict, no longer a dedicated field on RetrievalContext. Plugins should use context.retrieval_settings.get('top_k', 5) instead. Co-Authored-By: Claude Opus 4.6 * refactor(rag): remove unused capability enums CHUNKING_CONFIG, RERANK, HYBRID_SEARCH have no consumers — these behaviors are controlled by the plugin's creation_schema / retrieval_schema. Only DOC_INGESTION is used by the frontend to decide whether to render the document upload UI. Co-Authored-By: Claude Opus 4.6 * feat(rag): add SearchType enum, documents param, and search_type/query_text support - Add SearchType enum (vector, full_text, hybrid) to RAG entities - Add documents parameter to vector_upsert for full-text/hybrid search - Add search_type and query_text parameters to vector_search Co-Authored-By: Claude Opus 4.6 * fix(rag): replace base64 with chunked file transfer for get_rag_file_stream Use send_file/read_local_file pattern instead of inline base64 encoding for large binary file transfers in the GET_RAG_FILE_STREAM action. Co-Authored-By: Claude Opus 4.6 * feat(parser): add Parser component and DOC_PARSING engine capability (#37) * feat(parser): add Parser component and DOC_PARSING engine capability Add Parser as a new component type for external document parsing plugins. Includes SDK entities (ParseContext, ParseResult, TextSection), CLI scaffolding, runtime handlers for parse_document action, and plugin manager methods. Also add DOC_PARSING capability flag to RAGEngineCapability so engines can declare native parsing support — used by the frontend to decide whether to show the RAG engine as a parser option. Co-Authored-By: Claude Opus 4.6 * fix(parser): replace base64 with chunked file transfer and add invoke_parser API - Replace inline base64 file content with send_file/read_local_file chunked transfer pattern across the full parse_document chain - Add invoke_parser() method to LangBotAPIProxy for plugin developers Co-Authored-By: Claude Opus 4.6 --------- Co-authored-by: Claude Opus 4.6 * refactor: rename RAGEngine to KnowledgeEngine in SDK API * remove rag enigine * refactor: change rag engine to knowledge engine * fix(renderer): update error messages for Knowledge Engine validation in multiple languages --------- Co-authored-by: Junyan Qin Co-authored-by: Claude Opus 4.6

[Feature] RAG engine component #29

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions