Skip to content

Airis - Public Beta - v20260621

Choose a tag to compare

View all tags
@Samael-1976 Samael-1976 released this 21 Jun 18:39
· 18 commits to main since this release
V0.5
c96074f

NOTE:
Solved Problem 1 (Missing GraphRAG):
During training, chat.py extracts GraphRAG nodes in a background thread called _async_study_graph_extraction. However, the very millisecond this thread starts, chat.py calls execute_action() to respond to the user, taking exclusive possession of the LLM (Lock). The poor background thread is starved, and the extraction fails silently. Furthermore, the extraction function in brain_llm.py didn't have an explicit max_tokens!

Solved Problem 2 (Empty Intent Parsing Error):
extract_rg_intent in brain_llm.py used _generate_thought without an explicit max_tokens to reason. 12B entered the block, exhausted the default tokens (which are lowercase on some fallbacks), didn't generate the JSON, and returned an empty string "". json.loads("") throws the Expecting value error on row 1, column 1.

Solved Problem 3 (Wiki Autofill stuck):
This old statement is to blame: prompt = self._safe_replace(prompt, "page_text", page_text[:25000]).
25,000 characters is approximately 10,000 tokens. If the server's context window (n_ctx) is set to 8192 or 16384, inserting such a huge prompt will crash the C++ backend due to Out of Memory (Context Exceeded), or it won't leave the LLM even a single token to generate the tab's JSON. Hence the complete block.

Quick Start

Experience The Plug & Play Revolution in three simple steps. No external dependencies, compilers, or complex setups required.

1. Download & Extract

Download the latest standalone release package:

  1. Download the file from: https://aka.ms/vs/17/release/vc_redist.x64.exe and intall it
  2. Install espeak
  • For Windows: Download the file from: https://github.com/espeak-ng/espeak-ng/releases and install it.
  • For Linux Ubuntu/Debian: sudo apt-get update && sudo apt-get install -y espeak-ng
  • For Mac: brew install espeak-ng
  1. Download the file from: https://www.omnia-diffusion.com/airis/Airis-v20260621.rar
  2. extract the file
  3. cd Airis (or the name of the directory where you extracted the package)

2. Launch the Ecosystem

Run the bootstrap script corresponding to your operating system:

  • Windows:
    Double-click run.bat or execute it from the terminal:

    run.bat

  • Linux / macOS:
    Make the script executable and run it:

    chmod +x run.sh
    ./run.sh

3. Complete the Initiation Rite

VERY IMPORTANT: when prompted you must give permission for the windows firewall
On the first boot, the console will prompt you to select your preferred language, your preferite TTS (I prefere Vibevoice), and some other things. Once the background servers stabilize:

  1. Your browser will be automatically launched (or open your browser and navigate to http://localhost:8080) (or use the LAN/Ngrok link provided in the console).
  2. The Welcome Wizard will guide you through setting up your identity (Name, Gender, Birthday) and choosing the active vocal engine (Kokoro-TTS or VibeVoice).
  3. Secure your local sanctuary by creating your Administrator credentials.

The Anima is now awake. Welcome to the future of local, persistent AI interaction.

4. Enjoy!

If you'd like to buy me a beer to say thank you, I'll soon post a link to purchase the book I wrote on Amazon. It's available in both paperback and digital formats. It'll initially be in Italian, but I'll gradually translate it into more languages.
I'd really appreciate it.

Full Changelog: V0.3...V0.4

What's Changed

New Contributors

Full Changelog: V.0.2...V0.5

Contributors

eltociear
Assets 2
Loading