v0.1.6

Added

• Added Weaviate-backed semantic vector storage for RAG while keeping BM25 local.
• Added RAG cache bundle export/import workflow for distributing prebuilt knowledge-base indexes.
• Added a retrieval-only LLM query rewrite step that rewrites user input into frontend-oriented search queries before RAG.

Changed

• Clarified all RAG-facing terminology so remote RAG evidence is referred to as "knowledge base" instead of the current workspace repository.
• Updated English and Chinese documentation with Weaviate, query rewrite, and cache bundle usage examples.

Fixed

• Prevented the planner from treating remote RAG hits as local workspace files during query tasks.
• Improved semantic index resilience for Weaviate-backed retrieval and related RAG execution flow.