Nexus Chat v1.3.0 — Desktop Pet

New: Petdex Desktop Pet

A companion lives in your chat area and reacts to what the AI is doing.

• Petdex gallery — browse and search 1400+ animated pets from petdex.crafter.run directly inside Settings → Appearance → Desktop Pet
• Animated sprites — pets use the Codex spritesheet format with distinct animations for idle, walking, jumping, thinking, and more
• Wandering — the pet occasionally walks around the chat area on its own; can be disabled to fix it in place
• AI state linkage — the pet reacts to the agent in real time:
• Thinking (before first token) → review animation
• Generating (streaming) → running animation
• Done → waves, then returns to idle
• Draggable — drag to any position; saved across sessions; clamped within the chat area
• Click to interact — pet jumps and shows a random message on click
• Gallery pagination — 24 pets per page with Prev/Next and live search

What's Changed

• Added PetFloat in-window floating component (no separate Electron window)
• Added CodexSprite for Codex/Petdex spritesheet rendering
• Added PetdexGallery settings panel
• Petdex manifest fetched via IPC to bypass CORS
• New settings: Desktop Pet selection, wandering toggle

Nexus Chat v1.2.0

Image OCR

Smarter image handling: upload a screenshot and the AI reads it

When you attach an image (PNG, JPG, WEBP, etc.) to a message, Nexus Chat now automatically extracts the text from it using on-device OCR before sending it to the model.

How it works:

• Attach any image via the paperclip button or drag-and-drop
• A spinner appears on the thumbnail while OCR runs locally via tesseract.js (no network call)
• On success, the image is converted to a text attachment and included in your message — the AI sees the extracted content directly
• If OCR fails, the image falls back to the original image attachment with an error indicator

Language support: English and Simplified Chinese (eng+chi_sim) out of the box.

Note: First use downloads Tesseract language data (~10 MB) and requires tesseract.js to be installed (npm install).

---

SAP UI5 Visualization Cards

Rich data cards now powered by SAP UI5 React components

All built-in visualization cards have been migrated from Recharts to @ui5/webcomponents-react and @ui5/webcomponents-react-charts, bringing a consistent SAP Horizon Dark theme to all in-chat data visualizations.

What changed:

Card type	Component
Bar chart	ColumnChart (vertical) / BarChart (horizontal)
Line chart	LineChart
Area chart	ComposedChart with area series
Pie / Donut	PieChart / DonutChart
Data table	AnalyticalTable (built-in column sort)
Key metrics	Card + FlexBox + ObjectStatus (trend indicators)
Progress bars	ProgressIndicator (color-coded by completion: Positive / Information / Warning / Critical)

The entire app is wrapped in SAP UI5's ThemeProvider using the sap_horizon_dark theme.

Installation: requires npm install to pull @ui5/webcomponents-react, @ui5/webcomponents-react-charts, and the peer dependency @ui5/webcomponents-fiori.

Nexus Chat v1.1.1 — Anthropic Tool Support

What's New

Anthropic Tool Use

Web search and Skills are now fully functional when using the Anthropic API provider. Previously, these agentic features were only available with LiteLLM and OpenAI.

Under the hood, the main process now handles the full translation between OpenAI's tool format (used by the renderer) and Anthropic's native format:

• Tool definitions are converted from function.parameters → input_schema
• Assistant messages with tool calls are converted to Anthropic tool_use content blocks
• Tool results are grouped into Anthropic tool_result user messages
• Anthropic's tool_use stop reason is normalized back to the OpenAI shape so the agentic loop works identically across all providers

No configuration changes are needed — web search and Skills work as-is when Anthropic is selected as your provider.

Nexus Chat v1.1.0 — Multi-Provider Support

What's New

Multi-Provider API Support

You can now connect directly to OpenAI or Anthropic without a LiteLLM proxy,
in addition to the existing LiteLLM support.

Select your provider in Settings → Provider:

• LiteLLM — connect to any LiteLLM proxy (unchanged)
• OpenAI — direct connection to OpenAI API; defaults to api.openai.com
• Anthropic — direct connection to Anthropic API; defaults to api.anthropic.com

All three providers support a custom Base URL, so you can point them at
any compatible proxy or self-hosted endpoint.

Other Improvements

• App window and taskbar now display the Nexus Chat logo
• Windows build ships as a portable .zip (no installer required)

v1.0.1: Update app icon and switch win build target to portable

Replace default Electron icon with custom SVG-derived icon (512x512).
Change Windows build target from nsis to portable to avoid EPERM when
electron-builder tries to execute the generated uninstaller on restricted
Windows systems.

Nexus Chat v1.0.0 — Initial Release

Nexus Chat v1.0.0

First public release of Nexus Chat — an AI chat desktop app with rich data
visualization, powered by LiteLLM.

Features

Model Flexibility

• Connect to any LLM via LiteLLM: OpenAI, Claude, Gemini, Ollama, and more
• Configurable Base URL, API Key, and model name in Settings

Rich Visualizations

• AI renders inline charts (bar, line, area, pie), tables, metric cards, and
  progress bars directly in the conversation via a structured card syntax

File Uploads

• Attach images, PDFs, Word documents, Excel sheets, CSVs, and plain text / code files

Web Search

• Optional real-time search via Tavily, injected into the AI context automatically

Skills

• Define custom instruction sets and invoke them with /skill-name slash commands

Conversation Management

• Full history grouped by date
• Auto-generated titles from the first message of each conversation

UI

• Dark / Light / System theme
• Streaming responses with a stop button
• Clean, minimal interface built with React, Tailwind CSS, and Radix UI

Installation

Download NexusChat-win-portable.zip, extract, and run Nexus Chat.exe.
No installation required.

Requirements

• A running LiteLLM proxy, or any OpenAI-compatible endpoint
• Windows 10 / 11 (x64)