This project provides a framework for building and simulating large-scale populations of autonomous digital agents, designed to exhibit emergent social and collaborative behaviors. It targets researchers and developers interested in artificial societies, multi-agent systems, and AI, offering a platform to study unprogrammed, civilization-level properties arising from local agent interactions, moving beyond simple chatbots.

How It Works

Ava simulates thousands of agents, each possessing a distinct memory, belief system, and social model, within a shared environment. Agents operate on a perceive-deliberate-act loop. The framework's core architectural choices prioritize scalability and emergent behavior: concurrency is structural, leveraging asyncio.TaskGroup and asyncio.Semaphore for parallel agent execution; memory is tiered (episodic, semantic, procedural) with distinct retrieval and decay mechanisms; and emergence is measured statistically via change-point detection on population metrics, rather than subjective observation.

Quick Start & Requirements

• Primary install: Clone the repository (git clone https://github.com/TianhangZhuzth/Fundamental-Ava.git), navigate into the directory, and run pip install -e ".[dev]".
• Prerequisites: Python 3.11+ is required due to the use of asyncio.TaskGroup.
• Links: Repository: https://github.com/TianhangZhuzth/Fundamental-Ava

Highlighted Details

• Tiered Memory: Agents utilize separate stores for episodic (events), semantic (facts), and procedural (skills) memory, each with unique learning and retrieval dynamics.
• Scalable Governance: A GovernanceSystem allows agents to propose, vote on, and ratify laws, with thresholds dynamically adjusting to population size.
• Fault-Tolerant Communication: Implements a PBFT-style (PROPOSE → PREPARE → COMMIT) consensus protocol for reliable inter-agent communication.
• Statistical Emergence Detection: The EmergenceDetector employs statistical methods to rigorously identify and quantify emergent phenomena like role specialization within the agent population.
• Performance: Benchmarks indicate throughput of approximately 25,000 agents per second on commodity hardware for no-op agents, with graceful degradation at population sizes up to 5000.

Maintenance & Community

The project is described as "research infrastructure under active development." No specific community channels (e.g., Discord, Slack) or notable contributors/sponsorships are mentioned in the README.

Licensing & Compatibility

The project is licensed under the Apache 2.0 license, which is permissive for commercial use and integration into closed-source projects.

Limitations & Caveats

This is research infrastructure currently under active development. APIs, particularly within the ava.analysis module, are subject to evolution as new emergence detectors are added.