Technical Overview v1.0

An open-source platform for simulating economies with constrainted AI agents.

Authors Vincenzo Manto, Riccardo Dal Cero
Status Open Beta

How can we simulate nuanced human negotiation, irrational market sentiment, or adaptive trading strategies when agents are bound by rigid if/else logic?

The Declarative Shift

Doxa was developed as a Python-native simulation environment, designed from the ground up for modern researchers. At its core lies a simple philosophy: separating the “what” from the “how.”

The complexity of an agent's world—initial resources, market mechanics, price shocks, or resource shortages—is defined declaratively via YAML. This ensures simulations are not just complex, but readable and manageable.

AI-Native Brains

Integrate Gemini, OpenAI, or local Ollama instances. Assign goals and personalities to agents to generate non-deterministic strategic decisions.

Scientific Verifiability

Built with FastAPI and Pydantic. Execution is strictly determined by the scenario file and the engine version, ensuring reproducibility.

Case Study

The Hormuz Crisis Model

In our experimental models, we witnessed AI agents utilizing Propaganda as a Resource to mitigate political decay. Hard resource constraints eventually forced a ceasefire where traditional diplomatic prompts had failed.

Explore the Research Colab →

Scenario Definition

YAML Specification
actors:
  - id: player_farmer
    provider: google
    model_name: gemini-1.5-pro
    persona: |
      You are a farmer and a trader. Your goal 
      is to maximize gold reserves while 
      maintaining enough corn to survive.
    initial_portfolio:
      credits: 45
      corn: 12
      gold: 5

Doxa is currently in active development. We view this not as a finished product, but as the inception of a community-driven effort for complex systems research.

Stars on GitHub are deeply appreciated.