LoFS v1.2.3

Feature

Auto pulling models when adding Aliyun Bailian API-KEY or Kimi API-KEY, ordering them by updating time. You can select a model just in quick without adding them by hand. Therefore, the model list also displays in model library when adding Aliyun Bailian API-KEY or Kimi API-KEY

Performance, Refactor, Fixes

None.

LoFS v1.2.2

Features

Integrated Mem0 long-term memory: backend adds validation and retrieval endpoints, chat flow now queries and persists memories when enabled, and the renderer introduces a dedicated toggle plus API key handling.
Added a macOS-specific application icon to the build pipeline for better desktop branding.

Refactor

Moved optional asset loading logic into a standalone module and aligned index markup with the new CSP configuration.

Performance

No changes.

Fixes

Suppressed repeated optional dependency warnings and prune missing assets proactively to quiet the renderer.
Re-enabled Electron’s default web security settings, resolving prior security warnings.

LoFS v1.2.1

Features

• Adds conversation summarization with background generation of titles and summaries that does not block chat responses.
• Integrates web search into the chat flow to enrich answers and retrieval context.
• Improves chat UI to surface summaries and streamline file interactions.

Refactor

• Introduces shared text utilities for stripping reasoning tags and generating consistent summary previews across endpoints.
• Unifies summary handling in conversation APIs to return concise, UI-friendly previews.
• Reduces duplicated logic by centralizing text processing behavior.

Performance

• Offloads summarization to asynchronous background tasks, reducing perceived latency in chat.
• Trims summary payloads for listings and details, lowering bandwidth and rendering overhead.

Fixes

• Cleans up image attachments on send to prevent stale or unintended images from persisting between messages.
• Sanitizes summary content to avoid exposing raw reasoning segments, ensuring safer UI display.

LoFS v1.2.0

Features

• Added chat file tree to browse and select context files.
• Integrated document retrieval and reference management into conversations.
• Image uploads in chat with responsive thumbnails and click-to-expand.
• Streaming output support for smoother interaction.

Refactor

• Extracted the chat file panel into a standalone component to simplify main entry logic.
• Restructured chat module and stylesheet organization to improve maintainability.
• Unified file-tree related markers and event bindings in index.html.

Performance

• Enabled SSE streaming to reduce time-to-first-byte and average response latency.
• Minor UI/rendering tweaks around the file tree to avoid unnecessary reflows.

Fixes

• No explicit bug-fix commits detected; robustness and edge-case handling in chat API improved alongside feature changes.

LoFS v1.1.9

Merged Version1.1.8 and Version1.1.9

Features

• Added image upload capability and updated documentation (4a195e3).
• Integrated Tongyi Qwen model API into chat flow (907cd14).
• Added ModelScope API integration with server-side wiring (e4acc74).
• Introduced async upload workflow across renderer and server (228386e).

Refactor

• Reorganised model, config, and settings modules to support new providers and the upload pipeline (907cd14, e4acc74, 228386e).

Performance

• Tightened chat input styles and layout to reduce rendering cost and jitter (11cc1cc).

Fixes

• No user-facing fixes in this commit range.

LoFS v1.1.7

Features

Added rich document-detail overlay in the file tree, powered by a new /api/document/info endpoint that surfaces size, hashes, timestamps, summaries, and chunk counts (77c8839).
Enhanced chat UI assets and search dialog styling for clearer modal layouts and cross-theme polish (30267698d).

Refactor

Modularised the chat frontend into dedicated utilities and components (utils, search-dialog, model-selector, reference-manager), simplifying ChatModule responsibilities and loading order (72e1ed59d).

Performance

Optimised file-tree interactions and retrieval workflows, including cached path resolution, preview formatting helpers, and updated SQLite document metadata handling (77c8839).
Streamlined chat-renderer asset pipeline and stylesheet structure to reduce duplication and tighten UI rendering costs (30267698d).

Fixes

Eliminated duplicate reference evidence in chat replies by deduplicating metadata entries and summary cards within the new reference manager (f741ed7).

LoFS v1.1.6

Features

• Ollama Model Selector
  • Adds a runtime model picker so users can switch LLMs (Ollama) without restarting the app.
  • Integrates with the chat flow to route requests to the selected model.
• Document Topic Summary
  • Generates concise summaries for document-derived topics to improve navigation and discovery.
  • Persists topic metadata for consistent retrieval and display across sessions.
• Topic Retrieval
  • Implements topic-based retrieval to surface contextually relevant content during chats and searches.
  • Aligns retrieval with app settings to make behavior configurable by users.

Refactor

• Settings and Configuration Flow
  • Consolidates settings management across UI and backend to reduce coupling and improve maintainability.
  • Streamlines how chat and retrieval features consume configuration, making toggles and defaults more predictable.
• Topic Card Theme
  • Updates the visual theme of topic cards for clearer hierarchy, readability, and consistency with the chat interface.

Performance

• No performance-specific changes identified in the last four commits.

Fixes

• No bug fixes explicitly indicated in the last four commits.

LoFS v1.1.5

Features

• GitHub Module
  • Introduces a GitHub feature surface with a themed UI and dedicated workflows.
  • Integrates with splash, settings, and view state for consistent startup and navigation.
• CLIP ViT-B/32 (Multilingual)
  • Adds multilingual CLIP ViT-B/32 support to power cross-language image/text embeddings and retrieval.
  • Hooks into model download and management pipeline with tests to validate embedding behavior.

Refactor

• Database Management Module
  • Restructures database management to simplify APIs, reduce coupling, and improve maintainability across app and server.
  • Removes legacy cleanup pathways and aligns FAISS/data access patterns with a clearer model lifecycle.

Performance

• Reference Card
  • Optimizes render and interaction paths for the reference card, improving responsiveness in chat and explorer views.
• Image Retrieval Card
  • Speeds up image retrieval with leaner query and storage access, reducing latency and increasing throughput.

Fixes

• No bug fixes explicitly indicated in this commit range.

LoFS 1.1.4

Release Notes

• Introduces a PPTX loading spinner to eliminate the black screen when opening PPTX files; visuals match the splash screen.
• Adds drag-and-drop import to the file tree, improving import experience and efficiency.
• Includes multiple retrieval performance optimizations, significantly reducing response times and improving stability.

Added

• Drag-and-drop import in the file tree (drop files/folders onto the target directory for quick import).

Improved

• PPTX opening flow now shows a “Loading” spinner, avoiding black screens and providing consistent feedback.

Performance

• Optimized text retrieval to reduce latency.
• Optimized the search retrieval pipeline to improve query efficiency.
• Optimized chat information retrieval and prompt processing to improve message generation and contextual response speed.

LoFS v1.1.3

• Added real‑time document and folder upload progress streaming over WebSocket, including detailed PDF parse stages.
• Updated the renderer to track WebSocket status events and drive determinate loading overlays for single files, folder operations, and deep PDF parsing.
• Prevented transient progress broadcasts from overriding the backend “ready” status to keep the splash screen responsive after reloads.