overlooked

A native, Rust-built desktop chat client for local Ollama and any OpenAI-compatible API.

OpenAI · Anthropic · DeepSeek · Gemini · Kimi · GLM · Qwen · Yi · Mistral · xAI · Groq · OpenRouter · Together · Fireworks · Cerebras · Perplexity · NVIDIA NIM · DeepInfra · Ollama · LM Studio — all from one window.

---

GPT-style desktop chats faded from the conversation. Rust has the toolchain, the safety story, and the runtime — and almost nobody is shipping a serious native chat client for it. This is that client.

Why overlooked

• One binary, ~11 MB. No Electron, no Node, no Python sidecar.
• Local first, cloud ready. Talk to Ollama or LM Studio on localhost, or drop in any OpenAI-compatible API key.
• Real streaming for both Ollama (NDJSON) and OpenAI-format (SSE) — tokens appear as they arrive, with a real Stop button.
• 20+ provider presets. Pick one, the base URL and per-provider chat path are filled in for you.
• Built-in web search (Tavily) as a function-calling round-trip. Toggle it per session.
• Multi-user with per-user chats, settings, themes, and avatars. Argon2 password hashing.
• Light + dark themes plus a custom accent color.
• Local-first storage. Everything (chats, keys, avatars) lives in a single chat.db SQLite file on your machine. No telemetry.

Download

The fastest way to try it: grab the latest pre-built binary from the Releases page for your OS. Unzip and run — no installer needed.

If a release isn't up for your platform yet, see Build from source below.

Screenshots

Light theme

Dark theme

More views (login, full settings panel) live in screenshots/.

Supported providers

Provider	Auth	Notes
Ollama (local)	none	Default http://localhost:11434
LM Studio (local)	none	OpenAI-compatible at http://localhost:1234
OpenAI	API key	gpt-4o, gpt-4o-mini, o1, o3-mini
Anthropic Claude	API key	claude-opus-4-7, claude-sonnet-4-6, claude-haiku-4-5
DeepSeek	API key	deepseek-chat, deepseek-reasoner
Google Gemini	API key	gemini-2.0-flash, gemini-1.5-pro
xAI Grok	API key	grok-4, grok-2-latest
Mistral	API key	mistral-large-latest, codestral-latest
Moonshot Kimi	API key	kimi-k2, moonshot-v1-128k
Zhipu GLM	API key	glm-4.5, glm-4-plus, glm-4-air
Alibaba Qwen	API key	qwen-max, qwen-plus, qwen-turbo
01.AI Yi	API key	yi-large, yi-medium
Groq	API key	llama-3.3-70b, mixtral-8x7b
Cerebras	API key	llama3.1-70b, llama-3.3-70b
Perplexity	API key	sonar, sonar-pro, sonar-reasoning
OpenRouter	API key	Any model in the OpenRouter catalog
Together AI	API key	Llama, Mistral, DeepSeek, Qwen, ...
Fireworks AI	API key	accounts/fireworks/models/*
DeepInfra	API key	Llama, DeepSeek, Mistral, ...
NVIDIA NIM	API key	nvidia/* and meta/* hosted models
Custom	optional	Any OpenAI-compatible endpoint

Quick start

1. Open the app.
2. Click Settings in the bottom-left of the sidebar.
3. Provider → pick yours. The base URL fills in automatically.
4. Paste your API key (Ollama and LM Studio need none).
5. Model → pick from the live-fetched dropdown, or type your own.
6. Apply, hit + New chat, and start typing.

Enable web search

1. Sign up at tavily.com (free tier covers 1,000 searches/month).
2. Settings → Tools → Tavily API key → paste.
3. Click the globe button in the chat input pill before sending — the assistant will decide when to call web_search.

Hotkeys

Key	Action
Enter	Send
Shift+Enter	Newline
Ctrl+N	New chat
Ctrl+B	Toggle sidebar
Ctrl+,	Open / close settings
Esc	Close the open modal

Build from source

If there's no pre-built release for your platform, or you want to hack on it:

Requirements

• Rust (stable)
• Dioxus CLI: cargo install dioxus-cli
• Linux: sudo apt install libwayland-dev libgtk-3-dev libwebkit2gtk-4.1-dev libayatana-appindicator3-dev librsvg2-dev libxdo-dev libssl-dev pkg-config

Build & run

git clone https://github.com/KPCOFGS/overlooked.git
cd overlooked
dx build --release
./target/dx/overlooked/release/linux/app/overlooked

If you hit a black window on Linux + NVIDIA, run with:

WEBKIT_DISABLE_DMABUF_RENDERER=1 WEBKIT_DISABLE_COMPOSITING_MODE=1 ./overlooked

Privacy and security

• API keys, chats, and avatars are stored only in a local SQLite file (chat.db).
• No telemetry, no analytics, no remote logging.
• Passwords are hashed with Argon2 before being written to disk.
• Every text setting is sanitized at the input layer and again on save (control characters stripped, lengths capped on UTF-8 char boundaries).
• All numeric parameters are clamped to safe ranges before they hit any API.
• Avatar uploads validate magic bytes (PNG / JPEG / WEBP / GIF only) and cap at 1 MB.

Roadmap

• Multi-provider streaming chat
• Light / dark themes + custom accent
• Per-user accounts (Argon2) with guest default
• Web search via Tavily
• Native file picker for avatar upload
• MCP support — connect to local Model Context Protocol servers (filesystem, git, GitHub, etc.) you already configured for Claude Desktop or Cursor
• Markdown / code-block rendering
• Conversation search

Contributing

Issues and PRs welcome. Bug reports especially welcome if they include the provider, model, and a reproducible message.

Support

overlooked is built and maintained by @KPCOFGS in spare time, with no ads, telemetry, or paid tier. If it's useful to you, here are ways to keep it going:

• ☕ Buy me a coffee — one-time tip
• 💖 GitHub Sponsors — recurring support
• ⭐ Star the repo — costs nothing, helps a lot
• 🐛 Open an issue or PR — bug reports and patches always welcome

License

MIT. See LICENSE.