v1.1.0

[中文]
BUG FIX :

1. 增加上下文總結，修正當歷史對話變得過長時，RAG 系統的 Prompt 可能會超出模型的上下文視窗，或者因為包含過多無關資訊而降低回 答品質。
2. 新增錯誤處理，在 ask 和 stream_and_save 中加入了 try...except 區塊，在處理 stream_and_save 中，可以將錯誤訊息透過 EventSource 流回傳給前端，使前端可以向使用者顯示錯誤，而不是讓連線直接中斷。
3. 新增日誌，在總結和錯誤處理中加入了新的 print 日誌，加強後續的除錯和追蹤。
4. 修正功能開關在點擊文字標籤時可能無反應或行為異常的 UI 錯誤。
5. 重構 HTML 結構和 JavaScript ，提升前端介面的穩定性和渲染邏輯的健壯性。

新增核心功能 :

1. 多格式文件上傳，現在可以直接在「資料庫管理」分頁上傳本地文件，包括 .pdf, .docx, .pptx, .txt, .md 等多種格式 (但我只測試了PDF)。
2. 上傳的文件，可以被RecursiveCharacterTextSplitter切割成小區塊，以方便LLM搜尋。
3. 增加歷史檢索來源引用與追溯。
4. 增加歷史檢索時間戳。

[ENG]
Bug Fixes:

1. Added context summarization to address issues where the RAG system's prompt might exceed the model's context window due to excessively long conversation history, or where irrelevant information degrades answer quality.
2. Implemented error handling by adding try...except blocks in ask and stream_and_save. In stream_and_save, error messages can now be streamed back to the frontend via EventSource, allowing the frontend to display errors to the user instead of abruptly disconnecting.
3. Added new print logs in summarization and error handling to improve subsequent debugging and tracking.
4. Fixed a UI bug where clicking the text label of a feature toggle might result in no response or abnormal behavior.
5. Refactored the HTML structure and JavaScript to improve the stability of the frontend interface and the robustness of the rendering logic.

New Core Features:

1. Multi-format file upload: Users can now upload local files directly on the "Database Management" page, including various formats such as .pdf, .docx, .pptx, .txt, and .md (although I only tested PDF).
2. The uploaded files can be split into smaller chunks using RecursiveCharacterTextSplitter to facilitate LLM searching.
3. Added historical retrieval source citation and traceability.
4. Added timestamps for historical retrievals.