Changelog

Version 5.14.0

September 6, 2026

Tool Permission Modes: Added three clear permission modes for AI tools: Ask, Auto, and Full Access. Auto runs recognized read-only actions without interruption while still asking about risky or unknown actions. Choose a default in Settings or switch modes directly for the current chat.
Built-in Tools Setup: New installations now automatically configure Local Files for the Desktop and Swipeer Expertise folders and connect the built-in Web Browser when a supported browser is detected. Local Files also accepts manually entered paths such as hidden folders under ~/.config.
Markdown Rendering: Replaced the message parser with a standards-based Markdown engine for more reliable nested formatting, lists, tables, links, escaped characters, code spans, and incomplete streamed responses. Raw or unsafe HTML and executable links remain blocked.
Code Highlighting: Added language-aware syntax highlighting for completed fenced code blocks in both light and dark themes, while preserving the original source exactly when using the copy button.
Image Attachments: Image thumbnails in the message composer can now be opened in the full-size viewer by click or keyboard. Editing and resending a multimodal message no longer removes its attached images.
Reliable Usage Billing: Credit System and BYOK costs now remain associated with the correct response when a stream is cancelled, interrupted, or followed by a steered turn. Pending usage is reconciled after reopening a chat, and idempotent request handling prevents duplicate credit deductions.
Long-running Task Stability: Fixed long tool output expanding or appearing to freeze the chat window. The cancel control remains reachable, and tool-permission dialogs stay scrollable and clickable in compact windows.
New AI Models: Added OpenAI GPT-6 Astra, Google Gemini 3.8 Flash, and Meta Muse Spark 1.3 to the built-in OpenRouter model selection.
Model Catalog Update: Refreshed pricing and access tiers across the catalog, made Mercury 2.5 and MiMo-V2.5 available in the free tier, and removed the superseded Qwen3.5-9B, Inkling, and Inkling Small entries.

Version 5.13.0

September 1, 2026

New AI Models: Added Inception Mercury 2.5 Preview, a blazing-fast diffusion large language model (dLLM) for text and reasoning tasks; Tencent Hy4 Preview, a long-context mixture-of-experts model for coding agents and complex tool-use workflows; and Meta Muse Image for generating and editing images from text or reference images.
Model Pricing: Refreshed the built-in pricing for DeepSeek V4 Flash Latest, Google Gemini 3.7 Flash, NVIDIA Nemotron 3.5 Lightning, and Thinking Machines Inkling.
More Reliable Tool Runs: MCP tool runs now automatically retry a transient OpenRouter 504 timeout on both BYOK and the Credit System. Completed tools are preserved instead of running again, and Try Again resumes from the saved checkpoint when possible.
Smarter Window Placement: New chat, notes, chat history, settings, credits, and HTTP request windows now open near the main window on the same display. Valid saved positions are preserved, while off-screen windows are automatically moved back into view.
Windows License Activation: Fixed first-run activation so a newly entered license initializes the AI and credit systems immediately without requiring an app restart. Activation failures now show the actual error instead of being mistaken for an inactive subscription.
Code Copying: Fixed code-block copy buttons sometimes copying content from the wrong block, with more reliable copying through Electron's native clipboard.

Version 5.12.1

August 30, 2026

Fixed Bug: Removed the redundant Steer button next to Send. Steering remains available directly on queued messages while a response is running.

Version 5.12.0

August 30, 2026

Message Queue & Steering: Messages sent while the AI is responding now remain visible in a compact queue and are delivered automatically when the active response finishes. Queued messages can be redirected into the active task with Steer or removed before they are sent.
Built-in Tools: Local Files and Web Browser are now included with Swipeer. They use a managed runtime, require no separate Node.js or MCP package installation, and remain disabled until explicitly connected.
Controlled Local File Access: Local Files can be granted access to one or more selected folders directly from Settings. Swipeer limits the AI to those approved locations, and folder access can be reviewed or revoked at any time.
Persistent Agent Browser: Browser automation now uses a separate persistent agent profile, allowing logins and website state to survive between tasks without using the everyday browser profile. Chrome, Edge, and Chromium are supported, with Firefox available in beta.
AI-Created Expertises: Swipeer can now create, validate, install, and connect custom local Expertises for specialized tasks. Installation requires explicit tool approval and shows a dedicated warning before generated code is written or executed.
Integrations UI: Reorganized Settings into clear sections for built-in Swipeer tools, custom Expertises, and external MCP servers. Added safer deletion confirmations and more actionable startup errors with troubleshooting details.

Version 5.11.0

August 30, 2026

New Feature: Added transparent cost details for every AI response. Credit System requests show the charged Swipeer Credits, while BYOK requests show the OpenRouter cost in USD; token usage and model details are available directly from the response.
New Feature: Added a live total cost for each chat, with Swipeer Credit and BYOK spending kept separate. Voice transcription costs are included so the total reflects the complete conversation.
Provider Settings: Unified OpenRouter access into one clear choice between Swipeer Credits and your own API key (BYOK), with clearer setup guidance and direct links when configuration is required.
Pricing Accuracy: Refreshed the bundled OpenRouter model prices so estimates and BYOK cost displays use the latest available pricing.
UI Improvement: Refined themed scrollbars for a cleaner and more consistent appearance throughout the app.

Version 5.10.2

August 27, 2026

Fixed Bug: The Thinking Process remains scrollable while a model is actively reasoning. Incoming reasoning text no longer interrupts scrollbar dragging or forces the panel away from the position you are reading.
Streaming Reliability: Long-running reasoning and agentic tasks no longer stop at the previous 10-minute request limit. BYOK and Credit System streams now handle cancellation, timeouts, and interrupted connections more reliably while preserving partial responses when a genuine failure occurs.

Version 5.10.1

August 27, 2026

New Model: Added Qwen3.8 Flash, a fast multimodal reasoning model for coding, agentic workflows, visual understanding, and long-context analysis.
Fixed Bug: Fixed links opening unexpectedly. Swipeer now opens supported web and email links externally while blocking invalid or unsupported child windows.
MCP Reliability: Stabilized the connection lifecycle for integrations and expertises, including cleaner cancellation and shutdown, safer reconnects, request timeouts, and protection against duplicate connection actions.

Version 5.10.0

August 26, 2026

New Feature: Added a dedicated Integrations toggle to quickly enable or disable expertises and external tools independently from AI Memory. The preference is synchronized across Swipeer windows and saved between sessions.
New Model: Added Z.ai GLM 5.3 Flash, a fast multimodal model for coding, long-context, and agentic tasks.
Fixed Bug: Standalone LaTeX arrow notation in AI responses now renders as clean arrow symbols instead of showing raw commands such as $\rightarrow$ .

Version 5.9.2

August 19, 2026

Fixed Bug: Credit-based requests now only go through when you have enough credits — no more requests slipping through with an empty balance.
Improved Error Message: When a model returns an empty response, the message now names the model and explains how to fix it.

Version 5.9.1

August 19, 2026

Fixed Bug: The last selected model is now saved correctly and stays selected after restarting the app.

Version 5.9.0

August 19, 2026

UI Improvement: Redesigned the model selection dropdown with a cleaner layout and clearer grouping of models.
New Feature: Added model price tips in the dropdown and model info tooltip, showing input and output costs per 1M tokens — displayed as credits for Credit System users and in USD for BYOK.
New Models: Added Anthropic Claude Fable 5, Meta Muse Spark 1.2, NVIDIA Nemotron 3.5 Lightning, Qwen3.8 27B, Grok 4.6, and Google Gemini 3.7 Flash.
Updated Model: Upgraded DeepSeek V4 Pro to the 0813 GA revision.
Model Clean-up: Removed superseded entries, including Gemini 2.5 Flash Image, Gemini 3 Flash Preview, Gemini 3.1 Pro Preview, Gemini 3.5 Flash, Mistral Small 4, gpt-oss-120b, Qwen3.7 Max, Qwen3.7 Plus, and Grok 4.5.
Improvement: Expanded the reasoning effort levels to Minimal, Medium, Very High, and Maximum for finer control over reasoning-capable models.
Fixed Bug: Fixed a false credit error message that could appear under certain usage conditions.

Version 5.8.2

August 5, 2026

Voice Transcription: Long recordings are now encoded as compact Opus audio and automatically split into segments before they reach the upload limit. Each segment is transcribed in order while recording continues, preventing multi-minute dictations from being lost.
Streaming Reliability: Improved OpenRouter streaming error handling so provider timeouts and interrupted responses are reported clearly, partial responses are preserved, and empty responses no longer fail silently.
Image Generation: Fixed OpenRouter image models being sent to the chat completions endpoint. BYOK and credit-based image generation now use the dedicated Images API for Krea, FLUX, Gemini Image, GPT Image, Qwen Image, Seedream, and other supported image models.
Streaming Performance: Batched small response chunks and scrolling updates for smoother rendering during long answers and reasoning streams.
List Rendering: Improved Markdown list handling so numbered lists, repeated 1. markers, and explanatory paragraphs render with the correct structure and numbering.
New Models: Added Qwen Image 3, Qwen Image 3 Pro, Qwen 3.8 Max, and Thinking Machines Inkling Small.

Version 5.8.1

July 31, 2026

Fixed Deployment Issue: Fixed an issue affecting the release deployment.

Version 5.8.0

July 31, 2026

New Models: Added Anthropic Claude Opus 5 and Claude Sonnet 5, Google Gemini 3.5 Flash Lite, MoonshotAI Kimi K3, Poolside Laguna S 2.1, Qwen 3.7 Flash, and Thinking Machines Inkling.
Updated Model: Upgraded DeepSeek V4 Flash to the latest 0731 revision for improved coding, reasoning, and agentic workflows.
New Image Models: Added Krea 2 Large, Krea 2 Medium, and Krea 2 Medium Turbo for a broader choice of image quality, speed, and cost profiles.
Model Catalog Update: Refreshed model descriptions, context-window information, pricing, and license-tier assignments for the current OpenRouter selection.
Model Clean-up: Removed or replaced older and superseded entries, including Claude Opus 4.7, Claude Sonnet 4.6, Gemini 3.1 Flash-Lite Preview, MiniMax M2.7, Kimi K2.6, GPT-5.4, GPT-5.4 Mini, GPT-5.4 Nano, GPT-5.5, Grok 4.20 Beta, MiMo V2 Flash, MiMo V2 Omni, and GLM 5.1.

Version 5.7.0

July 19, 2026

New Feature: Added native right-click menus to chat messages and the message input with localized actions for undo, redo, cut, copy, paste, delete, and select all.
Improvement: Select All now targets only the clicked user or assistant message. Internal thinking content is excluded from assistant selections and copied text.
New Feature: Added window-local voice recording shortcuts: Ctrl+M on Windows/Linux and Option+M on macOS. Pressing Enter finishes an active recording and starts transcription without sending the message.
New Feature: Added shortcuts to pin or unpin the active Swipeer window above other windows: Ctrl+Shift+P on Windows/Linux and Option+Shift+P on macOS.
New Feature: Added a configurable context window for downloaded Ollama models, with Automatic, 8K, 16K, 32K, 64K, and 128K options. Ollama Cloud continues to manage its context size automatically.
UI Improvement: Improved the local-provider settings layout and context-window guidance across all supported languages.
Fixed Bug: Fixed image-generation requests that could fail or produce incorrect prompts when MCP tools or web search were enabled.

Version 5.6.0

July 9, 2026

New Feature: Redesigned tool permissions with clearer one-time, per-tool chat, and persistent approvals, an explicit all-tools mode for the current chat, and controls to review or revoke saved permissions in Settings.
New Models: Added OpenAI GPT-5.6 Sol, Terra, and Luna, xAI Grok 4.5, and Tencent Hy3 to the OpenRouter model selection.
Improvement: Improved MCP tool calling for local Ollama models, including more reliable note-related tool selection and preservation of reasoning during tool calls.
Fixed Bug: Fixed the recommended-tools setup flow when Node.js is unavailable, with clearer guidance to restart Swipeer after installing it.

Version 5.5.2

June 29, 2026

Fixed Bug: Fixed local Ollama model prompting by removing empty assistant placeholders from Ollama requests, so downloaded local models now receive the latest user message reliably.
Improvement: Kept Ollama tool-call messages intact while cleaning only placeholder messages with no real content.

Version 5.5.1

June 29, 2026

Fixed Bug: Fixed missing streaming behavior in tool-call and reasoning flows so responses continue to appear live instead of seeming stalled.
Improvement: Simplified the AI Memory settings display by removing unnecessary preference and note-reference fields from the visible memory overview.
Improvement: Improved internal AI Memory tool handling and coverage around memory display behavior.

Version 5.5.0

June 28, 2026

New Feature: Added AI Memory, a local persistent memory layer for user preferences, project facts, recurring rules, and note references, including a settings view to review and edit saved memories.
New Feature: Added AI Note-Editing, allowing you to edit notes with AI assistance.
Removed: Removed the old Add All Knowledge Base flow. Use AI Memory by default or select individual notes when specific context is needed.
Improvement: Improved note saving, note previews, and knowledge-base context handling.
New Models: Added GLM 5.2 and GPT Image 2.
Platform: Added Windows as a beta download path.

Version 5.4.3

June 27, 2026

New Feature: Added Speech-to-Text input, allowing you to dictate messages directly into Swipeer.
New Feature: Added transcription support for both BYOK users and the Credit System backend path.
Improvement: Added credit validation and charging for transcription requests in the Credit System.
Fixed Bug: Fixed Speech-to-Text when using your own OpenRouter API key.
Fixed Bug: Improved native audio dependency handling and app packaging for Speech-to-Text builds.
Fixed Bug: Fixed an issue where the Send button could stay disabled after inserting dictated text.

Version 5.3.0

June 5, 2026

New Model: Integrated the latest Qwen 3.7 Plus model.
New Model: Integrated the latest StepFun 3.7 Flash model.
New Model: Integrated the latest Minimax M3 model.
Model Clean-up: Removed outdated Qwen 3.6 and Qwen 3.5 (397B-A17B) models from the selection interface.
New Feature: Added Auto-Install Recommended Functions to automatically configure essential MCP functions during OAuth registration.
UI Improvement: Optimized the rendering, styling, and layout of the AI's thinking process in the chat interface.
Improvement: Increased maximum working/execution time limits for complex, longer-running agentic tasks.
Fixed Bug: Resolved empty or blank AI answers occurring during tool-call reasoning with certain models.
Fixed Bug: Hardened MCP server stability, fixing OAuth registration, SSE transport errors, and reconnection logs.
Fixed Bug: Fixed sudden terminations or premature stops in streaming AI responses.

Version 5.2.0

May 29, 2026

New Feature: Permissioned Tool Calling – Added security options to prompt for user approval before executing local MCP tools or system commands, ensuring complete control over actions performed on your machine.
New Feature: Reasoning Effort Selector – Introduced controls (Off, Low, Medium, High) to adjust the reasoning depth of reasoning-capable models before they generate responses.
New Model: Integrated the latest Qwen 3.7 Max model.
Deprecation: Removed Perplexity models and integration from the interface.
UI Improvement: Added a direct shortcut link to the settings panel from the main window.
Fixed Bug: Fixed startup crashes and loading failures for tool-calling settings in renderer contexts.
Fixed Bug: Resolved a bug where the tool call confirmation modal would occasionally disappear automatically.
Fixed Bug: Fixed thinking process layout and placeholder display issues.

Version 5.1.3

May 20, 2026

New Model: Added Gemini 3.5 Flash.

Version 5.1.2

May 17, 2026

Fixed Bug: Fixed starting issues.

Version 5.1.1

May 17, 2026

Fixed Bug: Fixed Windows starting issues.

Version 5.1.0

May 17, 2026

New Feature: Parallel MCP Tool Calling – Tools marked as parallel-safe now execute concurrently, dramatically speeding up multi-tool tasks while sequential tools still maintain their order.
Improvement: Live Reasoning Streaming – AI reasoning now streams live during tool detection, eliminating the 30-60 second UI freeze during long-thinking models.
UI Improvement: Collapsible Thinking Process – The thinking process visualization is now a collapsible box with a clean header, keeping the chat area tidy while preserving full reasoning visibility.
Improvement: Streaming Tag Handling – Consecutive thought_process tags from streaming chunks are now properly merged, and unclosed tag fragments are cleaned up automatically.
Improvement: Request Cancellation – The Cancel button now properly aborts all in-flight requests including active MCP tool calls, leaving no hanging processes behind.
Fixed Bug: Windows Startup Crash – Fixed crashes on Windows startup caused by missing null/undefined guards across multiple modules (license, settings, theme, i18n).
Fixed Bug: Sending Message Error – Fixed a reference error that could prevent messages from sending in browser contexts.
Fixed Bug: Empty Response Handling – Empty LLM responses are now properly treated as errors with a clear error message instead of showing a blank placeholder.

Version 5.0.2

May 13, 2026

Fixed Bug: Resolved License version display issues to ensure correct information is shown.

Version 5.0.1

May 10, 2026

New Guide: Added Note-Template Guide for Expertise usage, providing pre-configured templates for AI assistants to easily integrate local MCP scripts.

Version 5.0.0

May 10, 2026

Major Feature: Expertise (Beta) – Introduced Local MCP (Model Context Protocol) support. You can now extend Swipeer's capabilities by adding your own local scripts and tools directly to the AI's expertise.
Technical Upgrade: Increased Max Tool Steps from 20 to 80, enabling the AI to handle much more complex and multi-step agentic tasks.
New Feature: Improved Image Handling for better processing and presentation of generated and uploaded images.
Fixed Bug: Resolved Dropdown Stability issues on Linux where weblinks and automation menus would occasionally close prematurely.
Fixed Bug: Improved Keyboard Shortcut Logic – shortcuts are now correctly prioritized and handled even when input fields are active.
Fixed Bug: Fixed Chat Context Persistence issues, ensuring that Knowledge Base and Expertise context remain active throughout the session.

Version 4.5.4

April 29, 2026

New Feature: Added support for 7 additional languages including Arabic, Hindi, Italian, Japanese, Korean, Portuguese, and Turkish.
Improvement: Enhanced Keyboard-Shortcut Modal with a more polished and intuitive design.
Fixed Bug: Resolved an issue where the New Window Shortcut (Cmd/Ctrl + N) would not work correctly.
Fixed Bug: Improved Credit Error Handling for clearer user feedback on payment and usage issues.
Fixed Bug: Fixed a redundant event listener causing unexpected Error Pop-ups during chat sessions.
Fixed Bug: Resolved a bug when Deactivating MCP servers in settings.

Version 4.5.3

April 26, 2026

New Models: Integrated GPT 5.5, GPT Image 5.4, GPT Image 2, Mimo v2.5 (Pro & Multimodal), and DeepSeek v4 (Pro & Flash).
Improvement: Enhanced HTTP Request window with a more polished and intuitive design.
Improvement: Cleaned up Credit Management history by removing redundant device listings.
Fixed Bug: Resolved an issue where MCP servers would incorrectly disconnect when creating new ones.
Fixed Bug: Fixed a bug where Web-links and HTTP dropdowns would occasionally disappear.
Fixed Bug: Corrected MCP usage tracking within the credit system.

Version 4.5.2

April 21, 2026

New Model: Integrated Kimi 2.6 from Moonshot AI, delivering strong performance in long-horizon coding and agentic tasks.
Update: Refined Model Comparison benchmarks with latest verified data for Kimi 2.6 and Claude Opus 4.7.

Version 4.5.1

April 21, 2026

Fixed Bug: Resolved a critical Deployment Bug affecting the latest release cycle.

Version 4.5.0

April 20, 2026

New Models: Integrated Opus 4.7 and Mimo v2 Omni for superior performance.
New Feature: Introduced Device Management to monitor and manage active devices per license.
New Feature: Added BYOK (Bring Your Own Key) support for the Pro plan.
Improvement: Enhanced Model Distribution across license tiers for better accessibility.
UI Improvement: Optimized Error Message Rendering with specific guidance for vision-unsupported models.
Improvement: Refined application Translations across various interfaces.
Fixed Bug: Resolved an issue where the Model Selection dropdown would not correctly persist.
Fixed Bug: Corrected Next Renewal Date loading from the database in Credit Management.
Fixed Bug: Fixed Device Limit visibility for internal license types.

Version 4.4.1

April 12, 2026

New Models: Integrated Qwen 3.6 Plus and GLM 5.1 for cutting-edge performance and reasoning.
Improvement: Removed the Lock Icon from weblinks for a cleaner interface.
Fixed Bug: Resolved a critical issue where the MCP Connection would occasionally disappear.

Version 4.4.0

April 9, 2026

New Feature: Added support for custom MCP Headers and Authentication Tokens (enabling integration with protected servers like NocoDB).
Improvement: Enhanced cross-platform deployment logic for Windows and Linux to ensure consistent versioning.
Fixed Bug: Resolved "phantom update" issues where Linux and Windows clients would incorrectly prompt for updates.

Version 4.3.0

April 7, 2026

New Feature: Integrated Gemma 4 31B model for advanced reasoning and productivity.
New Feature: Chat History Search now displays the full context of a selected chat (showing all messages instead of just the first one).
New Feature: Added a direct link to the Changelog in update notifications.
Fixed Bug: Resolved an issue with Image Generation related to certain MCP configurations.
Fixed Bug: Corrected Context Injection behavior in the secondary chat window.
Fixed Bug: Fixed the User Upgrade flow for automatic account updates from Free to Paid plans.
Fixed Bug: Resolved a UI bug where Dropdown Menus would auto-close when interacting with context menus.
Fixed Bug: Fixed Auto-Hiding behavior for Web-links and HTTP requests.
Fixed Bug: Resolved general Chat History loading and persistence issues.

Version 4.2.0

March 29, 2026

UI Improvement: Enhanced the User Message Editing interface for a more seamless and intuitive experience.
Fixed Bug: Resolved a rendering issue where Copying Ordered Lists included extra whitespace and incorrect formatting.
Fixed Bug: Fixed a bug where Knowledge Base context was not consistently injected into active chat sessions.

Version 4.1.2

March 28, 2026

New Feature: Knowledge Base can now be activated at any time during an active chat session.
New Feature: Full Chat History is now accessible in the preview window (no more limits).
New Feature: Shortcut Tip (Ctrl/Cmd + Space) added to the welcome screen for quicker access.
UI Improvement: Enhanced Notes Styling with full Markdown-to-HTML rendering for a premium look.
New Models: Integrated GPT-5.4 (Mini/Nano/Full), Minimax-M2.7, Gemini-3.1-Flash-Lite, and Grok-4.20-beta.
Improvement: Added info icons and direct links to Model Comparison throughout the app.
Fixed Bug: Resolved a Linux-specific issue where chat loading could hang under certain conditions.

Version 4.0.1

March 2026

New Feature: Introduced a Free-Tier registration, allowing everyone to get started with Swipeer for free.
New Models: Integrated the latest Qwen 3.5 (9B and 397B-A17B) models for state-of-the-art performance and reasoning.
New Feature: Added Bot-Protection for registration to ensure platform stability.
Fixed Bug: Resolved a critical Save Image bug where generated images were not properly persisted.
Fixed Bug: Corrected License Check logic to prevent incorrect blocking of valid licenses.
Fixed Bug: Resolved an issue where saving chats in the Chat History failed.
Improvement: Fixed a Windows execution bug that prevented the app from starting on some systems.
Improvement: Enhanced release automation and platform stability for Windows builds.

Version 4.0.0

March 2026

New Feature: Added support for Ollama Tool-Calling for MCP integration.
New Feature: Added functionality to Edit MCP Functions directly.
New Feature: Support for ODS file uploads and processing.
New Feature: User messages are now rendered as Plain Text to ensure consistent display.
New Feature: Added Edit and Regenerate capabilities for user messages.
New Feature: Introduced Branch Chat for managing multiple conversation paths.
New Feature: Added Vision Support for Ollama models (e.g., Qwen 3.5).
Improvement: Migrated Chat History storage to a robust SQL database with automated migrations.
UI Improvement: Integrated a Dropdown selector for MCP servers.
UI Improvement: Persistent Model Selection state across application restarts.

Version 3.5.0

February 2026

New Feature: Added MCP Protocol (Model Context Protocol) support for integration with external tools and servers.
New Model: Integrated Google Gemini 3.1 Flash Image Preview (Nano Banana 2) for state-of-the-art image generation and editing at Flash speed.
New Feature: Notes are now beautifully displayed with full Markdown formatting support.
New Feature: Added ability to import and extract text from PDFs directly into notes via drag & drop or file selection.
UI Improvement: Refined application design featuring modern rounded corners for a sleeker look.
Internal optimizations and stability improvements.

Version 3.4.1

February 2026

New Feature: Added option to Export Chat History (text only).
Fixed Bug: Resolved an Image Generation issue with OpenAI.
Minor bug fixes and internal optimizations (default models update, name-migration cleanup).

Version 3.4.0

February 2026

New Feature: Introduced Ollama Cloud integration.

Version 3.3.3

February 2026

New Model: Integrated Gemini 3.1 Pro for enhanced multimodal capabilities and reasoning.
Fixed a critical bug in the License Check validation logic.
Resolved an issue with Chat History retrieval and persistence.

Version 3.3.2

February 2026

New Models: Integrated state-of-the-art MiniMax 2.5 and GLM 5 models for advanced reasoning and coding.

Version 3.3.1

February 2026

Fixed Bug: Resolved a 404 error when clicking the "View Changelog" button in update notifications.

Version 3.3.0

February 2026

New Feature: Successfully implemented and integrated Web Search functionality.
Improved model loading and selection reliability across all providers.
Fixed a Knowledge Base bug where non-indexed notes were occasionally included in AI context.
Resolved intermittent feature-blocking issues related to license validation logic.
Localized and corrected translation strings for the credit system across multiple languages.
Standardized legal documentation by transitioning to neutral "Merchant of Record" terminology.

Version 3.2.0

February 2026

Links now open in the default browser.
Refactored email notification service for better deliverability.

Version 3.0.4 - 3.1.0

February 2026

Implementation of systematic backend auditing and event logging.
Resolved cross-platform UI theme synchronization for Linux environments.
Enhanced model pricing fallbacks for high-load scenarios.

Version 3.0.0

February 2026

Swipeer Rebranding & Architecture Migration.
Complete transition of API endpoints to api.swipeer.ai.
Introduced unified license and credit management system.
Major UI overhaul with decoupled components for improved modularity.
Official deployment support for Linux (AppImage).

Version 2.2.1 - 2.2.0

January 2026

Advanced Credit Orchestration and hardened database security.
Introduced automated database recovery and maintenance utilities.
Optimized model routing for improved latency.

Version 2.1.2 - 2.1.0

January 2026

Multi-Window Productivity features.
Introduced detached window support for Notes and Chat History.
New centralized admin dashboard for real-time usage monitoring.
Secure proxy routing for global AI service availability.

Version 2.0.0

December 2025

Added Image Generation functionality with Flux 2 (Flex/Pro/Max) models.
Transition to a high-performance desktop core engine.
Initial introduction of the AI Audio Processing module.

Version 1.10.0

December 2025

Added global proxy support for all AI Provider signatures.
Advanced error handling logic for third-party LLM providers.
Native support for macOS ARM64 (Apple Silicon) architecture.

Version 1.9.0 - v1.8.0

December 2025

Refined chat input focus handling after message transmission.
Improved system-wide shortcut reliability and latency.
Localized update notifications and version check optimizations.

Version 1.7.5 - 1.7.0

December 2025

Transitioned core components to a protected, closed-source architecture.
Added initial support for Grok-series LLM models.
Implemented advanced code protection and obfuscation measures.
Finalized E2E deployment tests for official builds.

Version 1.6.1 - 1.6.0

September - December 2025

Official macOS notarization and application signing.
Introduced Knowledge Base selection for persistent context.
Fixed UI state preservation for model-settings modals.

Version 1.5.4 - 1.5.0

September 2025

Initial database-backed credit tracking and usage monitoring.
Introduced support for Prompt Templates in secondary windows.
Improved translation management and localized error handling.
Automated Vercel deployment workflows for web assets.

Version 1.4.1 - 1.4.0

September 2025

Security Foundations: Introduced BackendKeyManager for secure transmissions.
Internal task management system for improved background processing.
Build pipeline stabilization for cross-platform releases.

Version 1.3.0

September 2025

Refined chat history persistence and retrieval logic.
Optimized message streaming performance.

Version 1.2.18 - 1.2.0

August - September 2025

Introduced system tray menu for non-intrusive application control.
Added autostart functionality and background execution hooks.
Initial implementation of Google Gemini Pro model support.
Knowledge Base foundation: initial PDF and file indexing support.

Version 1.1.0

August 2025

Early UI refinements and foundational model connectivity.
Basic session management and history storage.

Version 1.0.1

August 2025

Initial Public Launch of the Swipeer core platform.
Foundational LLM framework and cross-platform architecture.