Unleash Local AI with AetherAI Studio

A private, single-file, zero-dependency local Chat Console, Onboarding checker, and Autonomous command executor. Run completely private models using Ollama, LM Studio, and vLLM APIs locally on your Windows PC.

Windows PowerShell Setup
PS > irm https://hankyle.com/run.ps1 | iex

Rich Utilities. 100% Offline.

Take absolute control over your local intelligence stack. Zero data collection, telemetry-free, runs entirely in sandboxed directories.

Autonomous Command Agent

Features an autonomous shell loop. Your local models write standard XML tags (read_file, write_file, list_dir, run_command) that are executed inside PowerShell to build or fix workspaces automatically.

Provider Hub Auto-Discovery

Integrated connectors query model metadata routes dynamically. Open LM Studio on port 1234, vLLM on port 8000, or Ollama on port 11434 and let the interface pull active model lists automatically.

Telegram & Discord Bridges

Sync background bridges safely. Long-poll incoming prompts via a dedicated Telegram Command Bot or stream status reports and metrics to Discord webhooks from the background thread.

Scientific LaTeX typesetting

Features a custom mathematical parser mapping. Instantly converts math tokens (both inline and block formula symbols) into beautiful HTML layouts with scroll bars.

Bilingual Voice Synthesis

Equipped with browser speech synthesis API. Transcribe input voice prompts dynamically using browser speech dictation or hear replies spoken aloud with localized voices.

Zero-Install Execution

Extracts assets completely on demand. The installer loads in memory, generates wrappers, sets environment variables, creates Desktop shortcuts, and starts the server with one click.

How to Setup in 3 Easy Steps

No zip extraction, no manual path editing, completely automated compiler delivery.

1

Run Command in Terminal

Open Windows PowerShell (Run as User), paste the installer pipeline, and press enter:

PS > irm https://hankyle.com/run.ps1 | iex
2

Select Language & Setup Options

The script downloads the compiled package, initializes the localized files inside $env:UserProfile\AetherAI-Studio, and runs diagnostic tests to detect your active local model provider.

3

Launch Console UI

The local web server will boot at port 8080 and automatically launch the chat HUD in your browser. A desktop shortcut named "AetherAI Studio" will be ready for subsequent launches!

Interactive Console Simulator

Test drive the chat interface and rendering engine directly inside this official website. Click a preset prompt on the left to watch AetherAI stream responses, format LaTeX math equations, and highlight code.

Local Engine Online
Hello! Select one of the preset prompts on the left or type your query in the box below to see how AetherAI Studio renders markdown text, LaTeX block math formulas, and code snippets in real-time.

Overview 概述

AetherAI Studio is a high-performance local AI client designed to run completely private intelligence processes. Unlike traditional web clients that require complex configurations or cloud connections, AetherAI serves as an interactive console loaded and extracted directly via Windows PowerShell.

Once deployed, it executes locally on port 8080 and registers a shell shortcut path so you can run it inside any CMD or terminal window at any time.

Prerequisites 準備工作

To connect the application to actual local models, you must have an active local inference backend running on your computer. AetherAI Studio has native support for the following engines:

  • Ollama (Default): The installer can download and configure this for you. Ensures OLLAMA_ORIGINS is configured as * to permit browser Cross-Origin connections.
  • LM Studio: Turn on the "Developer/Local Server" option inside LM Studio, start the API server on port 1234, and ensure CORS features are enabled.
  • vLLM: Start your server API using the argument command flag --allowed-origins "*".

Messaging Bridges 社群軟體串接

The messaging bridges system allows you to sync AetherAI Studio to background processes. This creates a multi-threaded daemon thread that communicates with Telegram or Discord.

Telegram Command Bot Setup:

  1. Talk to Telegram BotFather to generate a new Bot Token.
  2. Obtain your Chat ID (using ID bots or webhook tools).
  3. Paste both keys inside the Settings Modal in AetherAI Studio and click "Save Settings".
  4. The server starts an asynchronous background thread. You can prompt AetherAI Studio via Telegram. If the prompt contains file or command tags, the agent will execute the shell actions locally on your machine and pipe results back to you!

Discord Webhook Broadcast:

  • Paste your Discord Webhook URL inside the settings pane. The server will broadcast color-coded status reports, loading indicators, and boot diagnostics directly to your channel.

Agent Mode Guide 智能代理操作

When selecting **Agent Code** inside the provider panel, the application guides the local LLM to follow a recursive reasoning loop. The model outputs the following specific XML tool tags in its chat completions:

<list_dir></list_dir> # Lists files recursively in the workspace <read_file>path/to/file.txt</read_file> # Reads file contents <write_file path="path/to/file.txt">content</write_file> # Writes or edits files <run_command>npm run build</run_command> # Executes a shell command

The client intercepts the stream output, locks the UI, executes the command in local PowerShell, dumps the terminal output into the chat bubble, and recursively posts the output back to the model for the next step, capping execution loops at 12 steps for security.

Troubleshooting 疑難排解

Cloudflare Managed Challenge Intercepts CLI Queries

If running the one-liner command irm https://hankyle.com/run.ps1 | iex returns a Cloudflare warning page, CLI anti-bot measures are blocking PowerShell. Solve the challenge by loading the script page in your browser at https://hankyle.com/run.ps1, save the script on disk, and run the downloaded file directly: powershell -ExecutionPolicy Bypass -File .\run.ps1.

CORS / Origin Connection Blocked

If model indicators display 🔴 Offline, your browser is blocking requests due to missing CORS headers. Ensure Ollama is started with environment variables set: set $env:OLLAMA_ORIGINS="*" in PowerShell, or run the repair flag aetherai --fix which resolves this automatically.

CLI Maintenance References

AetherAI Studio registers global system wrapper paths. You can execute diagnostic, onboarding, or clean uninstall tasks directly from any terminal window using the flags listed below.

Flag Options / 旗標參數 Description & Behavior / 執行行為與說明
--onboard
Runs localized onboarding checks. It checks local environment PATH variables for active Ollama setups, scans current active background bridges configuration in config.json, and polls thread statuses.
語法：aetherai --onboard (執行系統引導檢測)
--fix
Automatically diagnoses and repairs common configuration roadblocks. It stops hung Ollama applications and restarts them served with CORS origins enabled (OLLAMA_ORIGINS=*), stops hung Telegram bridge background jobs, and checks if server port 8080 is free.
語法：aetherai --fix (診斷並修復環境設定)
--uninstall
A clean system uninstall. It terminates background bridges, deletes desktop shortcuts ("AetherAI Studio.lnk"), removes global commands, and cleans up workspace path references inside the User Env Path variables.
語法：aetherai --uninstall (移除系統指令與捷徑)
--help / -h / /?
Renders the bilingual CLI Usage Guide in the active console. Explains syntax rules and details all maintenance command references.
語法：aetherai --help (顯示說明選單)

Frequently Asked Questions

Find answers to common inquiries regarding installation, data handling, and hosting configurations.

Is AetherAI Studio telemetry-free?
Yes! AetherAI Studio contains zero tracking scripts, cloud collectors, analytics engines, or external reporting channels. All inputs, prompts, file reads, and bridge operations occur 100% locally on your computer.
What should I do if Windows blocks running the installer script (Execution Policy Error)?
By default, Windows restricts script execution for security. If you receive an execution policy error, you can easily solve it:
Option 1 (Quick Process Bypass): Run the script with an explicit bypass flag: 
powershell -ExecutionPolicy Bypass -File .\run.ps1
Option 2 (Enable System-wide Execution): Open a standard PowerShell window and run: 
Set-ExecutionPolicy RemoteSigned -Scope CurrentUser -Force
This registers execution permissions securely for your active user account, allowing AetherAI Studio to install and run smoothly.
How do I set up Cloudflare WAF Bypass if Cloudflare blocks `irm`?
If you are hosting `run.ps1` behind Cloudflare, you can add a Custom Rule in the WAF settings:
1. Go to WAF -> Custom Rules -> Create Rule.
2. Select Field: URI Path, Operator: equals, Value: /run.ps1.
3. Action: Skip / Bypass security policies (WAF, rate limiting, and security levels).
This lets script queries like `irm` fetch the script without getting blocked by Cloudflare's JS challenges.
Can I run Agent Mode without Ollama?
Absolutely. You can select LM Studio or vLLM inside the provider options, input their API endpoints (e.g. port 1234 or 8000), save settings, and toggle "Agent Code". Ensure that the local model you load (like Llama 3 or Mistral) is capable of generating clean XML syntax.
What files does the installer create on my PC?
The installer sets up a permanent directory at %USERPROFILE%\AetherAI-Studio. It extracts index.html (the portal UI), server.ps1 (the static bypass server), and updates configurations inside config.json. No keys are written to the registry, making cleanup simple using the --uninstall flag.

MIT License

Command copied to clipboard!