ThatGPT

Local ChatGPT-style desktop chat for OpenAI-compatible APIs and Ollama

Features · Screenshots · Installation · Development · Build · Changelog · Trust · Contributing

Landing page · v2.6.3 release notes · Fork builds (GitHub Actions)

---

Description

ThatGPT is an open-source Windows desktop chat client with a ChatGPT-inspired UI. Conversations, prompt presets, and multimodal attachments live in a React UI. A Rust + Tauri backend stores JSON on disk and proxies all model requests so your API key never enters the webview.

Design goals: credentials stay in the desktop shell, no database overhead, any OpenAI-compatible endpoint or local Ollama.

Features

ChatGPT-like UI	Sidebar nav, pill composer, model dropdown, tabbed settings
Providers	Multiple OpenAI-compatible profiles + Ollama (/api/tags)
Streaming	True SSE streaming with smart scroll and stop generation
Agent tools	Web search, knowledge search, analyze image/audio, generate images
Vision & image gen	Separate badges — see images vs create images (gpt-image-1, DALL·E)
RAG	Local knowledge base with embeddings, chunking, citation footnotes
Library	Browse all attachments across conversations
Markdown	GFM, LaTeX (KaTeX), Mermaid diagrams, syntax-highlighted code
Conversations	Pin, archive, folders, Ctrl+K search, ephemeral (temporary) chat
Power user	Branching, fork, templates, bookmarks, command palette, HTML export
Tokens & cost	Usage footer, context bar, editable per-model pricing
Themes	Light / dark with prefers-reduced-motion
Export	Markdown, JSON, HTML share; copy thread as Markdown

Screenshots

Main chat

Models list

Personalization setting

Providers setting

_{Demo GIF}

Installation

Download (recommended)

1. Open GitHub Releases — v2.5.0.
2. Download portable .exe or NSIS setup installer for Windows x64.
3. Verify checksums — SHA256.txt and docs/TRUST.md.
4. (Optional) Review VirusTotal reports (v2.5.0) after maintainers upload scans.

Configure API access

Open Settings in the app (header ··· menu) to set your API key, providers, models, and optional features (web search, knowledge base, dev mode). Settings are saved to:

%APPDATA%\com.thatgpt.desktop\.env

You can also edit that file manually — see src-tauri/.env.example:

AI_API_KEY=your-key-here
AI_BASE_URL=https://api.openai.com/v1
AI_MODEL=gpt-4o-mini
AI_DEFAULT_SYSTEM_PROMPT=You are a helpful assistant.
AI_REQUEST_TIMEOUT_MS=60000
AI_MAX_RETRIES=2

Ollama: Add an Ollama provider in Settings → Providers (http://127.0.0.1:11434).

Build from source

Requires Node.js 20+, Rust, and WebView2 on Windows.

git clone https://github.com/Satan2049/that-gpt.git
cd that-gpt
npm install
npm run build

Portable binary: src-tauri/target/release/bundle/portable/ThatGPT.exe

Development

npm install
npm run dev          # Tauri + Vite hot reload
npm run test:rust    # Rust unit tests
npm run build:client # Frontend only
npm run clean        # Remove build artifacts & temp files

Config template: src-tauri/.env.example

Variable	Description
AI_API_KEY	Provider API key (required for cloud APIs)
AI_BASE_URL	OpenAI-compatible base URL
AI_MODEL	Default model id
AI_IMAGE_MODEL	Image generation model (e.g. gpt-image-1)
AI_DEFAULT_SYSTEM_PROMPT	Fallback system message for new chats
AI_REQUEST_TIMEOUT_MS	Provider request timeout (ms)
AI_MAX_RETRIES	Retry count for transient API errors

See CONTRIBUTING.md for pull request guidelines.

Build

Windows desktop

Command	Output
npm run build	NSIS installer + portable EXE
npm run build:portable	Portable EXE only
npm run release:package	Stage release/ folder with ZIPs
npm run release:hashes	Regenerate SHA256.txt from built artifacts
npm run clean	Delete target/, client/dist/, caches, logs
npm run clean:all	Also remove node_modules and release/

Artifacts:

src-tauri/target/release/bundle/portable/ThatGPT.exe
src-tauri/target/release/bundle/nsis/ThatGPT_*_x64-setup.exe
release/                                    # after release:package
SHA256.txt                                  # after release:hashes

CI & cross-platform release

GitHub Actions builds on Windows, macOS, and Linux (see .github/workflows/ci.yml).

To publish v2.6.3 (or any version), push a version tag — the release workflow builds Windows, macOS, and Linux and uploads SHA256.txt:

git tag v2.6.3
git push origin v2.6.3

To build from a fork, see docs/FORK_BUILDS.md.

Windows-only packaging (local)

npm run build
npm run release:package
npm run release:hashes

Architecture

React UI (Vite)
      │  Tauri invoke
      ▼
Rust services ──► OpenAI-compatible API / Ollama
      │
      └── JSON: %APPDATA%/com.thatgpt.desktop/data/

Details: docs/repository-intelligence.md

Tech stack

Layer	Technology
Desktop shell	Tauri 2
Backend	Rust, Tokio, Reqwest, Serde
Frontend	React 18, Vite, TypeScript, Zustand
Storage	File-based JSON
AI	OpenAI-compatible /chat/completions, /images/generations, embeddings

Trust

• Verify downloads: docs/TRUST.md
• Release notes: docs/RELEASE_v2.6.3.md
• Report security issues: SECURITY.md
• Checksums: SHA256.txt

License

MIT © ThatGPT contributors