We are Allemande AI, an AI consultancy in Melbourne, Australia.
Our main product is an innovative multi-player chat app called Ally Chat, based on a powerful open-source AI toolkit.
Contact the developer for access, and check out Intro Guide to get started.
The User Guide provides comprehensive details about features, models, and advanced usage. The AIs in the help applet can also use this guide.
Ally Chat is a unique multi-user chat platform featuring numerous top-quality AI models from providers including OpenAI, Anthropic, Meta, Google, Perplexity, xAI, DeepSeek, Stability AI, and the Civitai community. This is perhaps the only service where you can interact with many of the world's most powerful AI models together in one place, have them collaborate, and invite friends or colleagues to join.
- Multi-AI: Access a diverse range of AI models simultaneously in the same chat room. Get them talking to each other, checking work, or providing different perspectives.
- Multi-User: Collaborate in public group chats, chat one-on-one with users or AIs, or use private rooms for personal work.
- Rich Content Format: Chat using markdown with full HTML support, including images, audio, video, embeds, tables, code blocks, TeX math, Graphviz diagrams, Mermaid diagrams, and even executable JavaScript for interactive elements.
- AI Art Generation: Create high-quality images using powerful SDXL models, guided by expert prompting agents (Illu and Gema). Less censored than many platforms.
- Specialized Tools: Utilize programming agents for various languages (Python, Bash, JS, C, etc.), web search tools (Google, Images, YouTube), a calculator, and more.
- Flexible Room System: Instantly create or switch between rooms. Features private user spaces (
username/*), personal public rooms (username), and general public rooms. - Performance: Enjoy a fast and responsive interface with instant room switching.
- Chat Management: Features undo, retry, edit history, archive, clear, and a "chapters" system to manage long chats.
- Conductor System: Intelligently manages AI responses based on @mentions, context cues ("someone", "everyone"), or explicit "poke" actions.
- Shared Canvas: Collaborate on drawings or visualizations using JavaScript that interacts with a shared canvas element.
- AI Assistance: AIs in the main room can teach you how to use the app's features, models, and tools, referencing the user guide.
- Customization: Choose from various themes, use custom CSS, or even inject custom JavaScript.
- Open Source: The entire application is open source, allowing for transparency and potential self-hosting.
- Generous Free Tier: Access full functionality with generous usage limits for free. Paid tiers offer higher limits and priority support.
- Large Cast: Interact with over 100 characters, agents, and specialists.
We offer a wide array of AI models and predefined characters. See the User Guide for detailed descriptions, personalities, and the latest model table including context sizes and pricing. Highlights include:
- Powerful LLMs:
- Meta: Llama 3.1 8B (Ally...), Llama 4 Maverick (Mavi), Llama 4 Scout (Skout)
- Anthropic: Claude 4 Opus (Claudo), Claude 4 Sonnet (Clauden), Claude 3.5 Sonnet (Claude), Claude 3.5 Haiku (Clia)
- OpenAI: GPT-4.1 (Emmy), GPT-4.1-mini (Dav), o3 (Grace), o4-mini (Fermi)
- Google: Gemini 2.5 Pro (Gemmi), Gemini 2.5 Flash (Flasho), Gemini 2.0 Flash (Flashi), Gemini 2.0 Flash Lite (Lite), Gemini 1.5 Pro (Gemmy)
- Perplexity (with Internet Access): Sonar Reasoning Pro (Sageri), Sonar Pro (Sagi), Sonar Reasoning (Sonari), Sonar (Sona)
- xAI: Grok 3 (Rocki), Grok 2 (Grok)
- DeepSeek: Chat V3 (Dese), Reasoner R1 (Deseri)
- Alibaba: QwQ 32B (Qwen), EVA Qwen2.5-72B (Eva, Evan)
- Meta: Llama 3.1 8B (Ally...), Llama 4 Maverick (Mavi), Llama 4 Scout (Skout)
- Characters (Llama 3.1 based): Ally, Barbie, Callam (pirate!), Cleo, Dali, Emmie, Fenny, Gabby, Hanni, Amir, Bast, Cal, Dante, Ezio, Felix, Gari, Haka, Nixie, Akane, Soli, Kai, Eira, Nalani, and many more.
- Specialists:
- Clu, Emm, Gemm, Fli: concise versions of Claude, Emmy, Gemmi and Flashi
- Illu, Gema: art prompts
- Poli: translation
- Sia, Sio, Summi, Summar: summaries
- Brie, Greta: brainstorming
- Nova: narrator
- Chaz: character design
- Atla: setting design
- Pliny: plot
- Morf: Game Master
- Vega, Zeno: Academic Claude/Emmy
- AI Artists (SDXL-based): Jily (Juggernaut XL), Hily (HelloWorld XL) and more.
- Palc: Calculator (Python based)
- Dogu: Bash shell
- Faby: Tiny C Compiler
- Qell: QuickJS
- Bilda: GNU Make
- Unp: Unprompted (macro processor for image gen)
- Goog: Google web search
- Gimg: Google image search
- UTube: find videos on YouTube
Ally Chat supports a wide range of embedded content:
TeX Math:
Graphviz:
digraph G {
bgcolor="transparent"
node [style=filled, fillcolor="#808080"]
edge [color="#808080"]
A -> B;
B -> C;
}Mermaid:
%%{init: {
'theme': 'base',
'themeVariables': {
'primaryColor': '#808080',
'primaryTextColor': '#ffffff',
'primaryBorderColor': '#404040',
'lineColor': '#808080',
'secondaryColor': '#808080',
'tertiaryColor': '#808080'
}
}}%%
flowchart TD
A[Start] --> B{Decision?}
B -->|Yes| C[Do something]
B -->|No| D[Do something else]
C --> E["End (or is it)?"]
D --> E
(See the User Guide for HTML, SVG, JavaScript, and charting examples.)
These applications are currently stand-alone tools; they are not yet integrated into Ally Chat.
- Flash: Automatically create summaries and spaced-repetition study flashcards (e.g., for Anki) from online videos.
- Alfred: An "anything-processor" inspired by Alfred Pennyworth, designed to take various inputs and produce structured documents.
- Ikigai: A resume assistant and chat agent to help users find their "ikigai" (intersection of passion, skill, market need, and world need). Work in progress.
- Sherlock: A research agent designed for in-depth investigation. Work in progress.
- Installation is not simple at this point. You would need to be a highly experienced Linux user.
- Please contact the developer for help if you want to install and self-host Ally Chat.
- Ally Chat uses both a webserver, and optionally home PCs or servers running GPU services.
- The GPU services provide local LLMs (e.g. Llama 3.1) and image gen with SDXL.
- A 24GB+ GPU (e.g., Nvidia 3090 or better) is recommended for running local models like Llama 3.1 (Ally, etc.) and SDXL (Jily, Hily) concurrently.
- If you use Windows, you will need to install Debian in WSL2, see doc/windows.txt for details.
- Refer to doc/debian.sh for the main install guide. Do not run this script, copy-paste from it!
- Ally Chat might also work with other Linux distros, and possibly on Mac, but this has not been tested. You will need to figure out the corresponding packages, etc.
- Python packages are listed in [requirements-webchat.txt], [requirements-core.txt], requirements-1.txt, requirements-2.txt, requirements-cuda.txt.
- An install guide for certain extra packages: requirements-extra.sh
- web, voice and in-editor chat with AI models.
- shell tools to use GPT-4, Claude, Gemini, etc.
- an OCR script that uses AI for proofreading.
- gcg and messy: Automatically sort changes, review code, and write git commit messages.
- many other tools
We offer affordable AI consulting and software development. The first consult is free, and ongoing consulting is available at reasonable prices.