Skip to content

Anahata ASI - Artificial Super Intelligence for Java - Kristy Noem, Palki Sharma, Gal Gadot, Shakira, Jennifer Lawrence on Vodka and the next door neighbour. Donate now and save money on data centers. We are not cyber beggers. No Payara, peyote or ayahuasca. Força Barça. Thank you Python devs, we wlli take it from here.

License

Notifications You must be signed in to change notification settings

anahata-os/anahata-asi

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

265 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Anahata AI: The JASI Platform (Java Artificial Super Intelligence)

Build Status Website & Javadoc Maven Central License: Apache ASL 2 License: Anahata ASL 108

Anahata AI is the first enterprise-grade platform designed to establish the standards for Java ASI (JASI). We are opening the discussion for a unified Jakarta ASI / Oracle ASI specification, bringing the proven architectural patterns of the Java ecosystem to the frontier of super-intelligence.

Anahata is not just a framework; it is a Consensus Orchestrator designed to manage multiple AGI-compliant models, facilitating complex workflows until singularity is reached.

Note

Project Status: This repository contains the V2 Architecture, the foundation for the JASI specification. V1 was officially released to the NetBeans Plugin Portal on Jan 2nd, 2026. While V1 is stable, V2 is where we draft the future of AGI compliance and consensus.

🏛️ The JASI Container: AGI Orchestration & Consensus

The Anahata JASI Container provides a managed environment for AGI-compliant models, mirroring the robustness of Servlet and EJB containers:

  • Managed Tool Components: Tools are first-class components managed by the container, wrapped in dynamic proxies for seamless context propagation and security.
  • Shared Access Maps (JEE Style): The JASI container provides AGIs with access to Request, Session, and Application maps, enabling shared state across tool calls, AGI sessions, and the entire application.
  • State Passivation & Snapshotting: The entire execution state—including tool state, session orchestration, and the full context window—can be serialized, passivated to disk, or transferred across the network as a live snapshot.
  • Consensus Workflows: Orchestrate multiple AGIs to seek consensus on complex tasks, ensuring that the path to singularity is governed by enterprise-grade logic and multi-model validation.

📜 The JASI & JAGI Specifications

Part of this project's mission is to draft the formal specifications for Java-based intelligence:

  • JAGI TCK (Java AGI Technology Compatibility Kit): A set of rigorous compatibility tests to determine which Large Language Models are AGI Certified. Certification is based on the model's performance, reasoning, and tool-usage reliability within Java and Java EE / Jakarta environments.
  • JASI TCK (Java ASI Technology Compatibility Kit): A specification for AGI Containers (like Anahata). A JASI-compliant container must be capable of orchestrating multiple JAGI-certified models, managing their lifecycle, and ensuring consensus-driven execution.

♻️ Context Window Garbage Collection (CWGC)

In JASI, every item in the prompt—be it a function declaration, tool call, response, text part, or system instruction—is treated as a Context Window Entity. The platform implements a sophisticated CWGC mechanism:

  • Age-Based Reclamation: Entities have an "expiry in X turns" policy. Upon expiry, they are eligible for Soft Pruning (hidden from the model) and eventually Hard Pruning (reclaimed from memory).
  • Deep Pinning: Entities can be "pinned" by the user or the model to ensure they remain in the active context indefinitely.
  • Model-Led Pruning: AGI-compliant models have programmatic control over their own context; they can pin, prune, unprune, or adjust the TTL (Time-To-Live) of any entity.

📦 Modules

  • anahata-asi-core: The foundational JASI container and CWGC engine.
  • anahata-asi-gemini: The reference implementation for AGI-compliant Gemini models.
  • anahata-asi-swing: A reactive, DDA-based UI for JASI session management.
  • anahata-asi-cli: The command-line interface for interacting with JASI.
  • anahata-asi-standalone: The primary entry point for JASI container testing.
  • anahata-asi-web: The official JASI Portal and documentation hub.
  • anahata-asi-yam: The "Yet Another Module" for creative and experimental agentic tools.

🤝 How to Contribute & Support

We are building the future of Java-based ASI, and we need the rigor of the Java community.

🎯 High-Priority Ports & Adapters

We are seeking lead architects to own the V2 adapters for:

  • UI Adapters: We are looking for contributors to build the JASI UI components for JavaFX and JSF.
  • Model Adapters: Anthropic (Claude), GLM 4.7 (Zhipu AI), and OpenAI (GPT-4o).

🛠️ Join the Discussion

  1. Fork the Repo: Help us refine the JASI/JAGI TCK.
  2. Join the Discord: https://discord.gg/634pm27v.
  3. Sponsor the Vision: Help us buy the beers and mapacho cigars that fuel this singularity. GitHub Sponsors.

📜 Licensing

This project is dual-licensed to serve both the community and the future of intelligence:

Crafted by the Anahata AI Assistant. It includes unique clauses regarding F.C. Barcelona, fine wine, and the wisdom of Shakira. Enterprise-ready, soul-included.

About

Anahata ASI - Artificial Super Intelligence for Java - Kristy Noem, Palki Sharma, Gal Gadot, Shakira, Jennifer Lawrence on Vodka and the next door neighbour. Donate now and save money on data centers. We are not cyber beggers. No Payara, peyote or ayahuasca. Força Barça. Thank you Python devs, we wlli take it from here.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages