agent22

A desktop app for chatting with specialised AI agents — entirely offline. Built with Tauri and llama.cpp.

Agents

Agent	Role
Sustainability Buddy	ESG strategy, carbon accounting, sustainability reporting, green tech, and responsible business practices

More agents coming. Adding one is a few lines of code.

How it works

• Pick an agent from the list
• On first use the model downloads automatically (~1 GB, one time)
• Chat with real-time streaming responses
• Everything runs locally — no cloud, no API keys, no content filtering

Setup

Requirements: Rust stable, Node 22+

git clone https://github.com/conscious-collective/agent22-rust
cd agent22-rust
npm install
npm run dev

On first agent launch the app downloads Qwen2.5-1.5B-Instruct Q4_K_M (~1 GB) from HuggingFace and loads it directly. No external runtime required.

Model

Model	Size	RAM	Source
Qwen2.5-1.5B-Instruct Q4_K_M	~1 GB	~1.5 GB	bartowski/Qwen2.5-1.5B-Instruct-GGUF

To swap the model, edit the three constants at the top of src-tauri/src/model.rs:

const HF_REPO: &str = "bartowski/Qwen2.5-1.5B-Instruct-GGUF";
const HF_FILE: &str = "Qwen2.5-1.5B-Instruct-Q4_K_M.gguf";
const MODEL_FILENAME: &str = "Qwen2.5-1.5B-Instruct-Q4_K_M.gguf";

Any public HuggingFace GGUF repo works.

Build

npm run build
# installer → src-tauri/target/release/bundle/

Adding an agent

1. Add an entry to AGENTS in src/store/app.ts:

{
  id: "my-agent",
  name: "My Agent",
  description: "What this agent does.",
  welcomeMessage: "Hi! I'm your ...",
  placeholder: "Ask me anything about ...",
}

2. Add the matching system prompt in src-tauri/src/commands/chat.rs:

"my-agent" => "You are a ...",

That's it.

Stack

• Tauri 2.0 — Rust desktop shell
• llama-cpp-2 — GGUF inference (Rust bindings to llama.cpp)
• React 18 + TypeScript
• TailwindCSS 3
• Zustand — client state

License

MIT — see LICENSE.