Skip to content

oasios-observatory/oasios-generator

Repository files navigation

Open Artificial SuperIntelligence Scenario Generator v0.1

OASIOS Generator v0.1

Experimental generator producing single-ASI and multi-ASI foresight scenarios using local LLM models.

The OASIOS Generator is an open-source module for creating narrative simulations of Artificial Superintelligence (ASI) emergence and interaction. It generates structured scenarios defined by the schema, and stores them in a local SQLite database for later mapping or analysis. It can generate both single-agent (single-ASI) and multi-agent (muli-ASI) scenarios, each accompanied by interpretive narratives and thematic indicators. It is an experimental tool for producing foresight-style reports on potential futures of Artificial Superintelligence, with potential application in research, educational and creative projects.

This module represents the first step toward the larger OASIOS Observatory ecosystem.

Status: Alpha — open for testing and conceptual feedback.

✳️ Focus: narrative coherence, scenario diversity, and epistemic transparency.

✨ Features

  • Generate single-ASI or multi-ASI scenario narratives
  • Plug in your preferred LLM backend (Ollama, OpenRouter, Claude, etc.)
  • Store and query scenarios using SQLite3
  • Extensible: connect to precursor data or visualization tools

📦 Installation

git clone https://github.com/OASIOS-observatory/generator.git
cd generator
pip install -r requirements.txt

Requirements:

Python 3.8+

SQLite3

(Optional) Ollama or compatible LLM backend

🚀 Usage

Generate a single scenario:

python OASIOS_generator/single_asi_scenario.py

Generate a batch:

python OASIOS_generator/single_asi_batch.py

Output database:

data/asi_scenarios.db

## 📁 Project Structure

generator/
├── config/
│   └── asi_scenario_schema.json       # JSON schema definition for validating scenario structure
├── utils/
│   └── abbreviator.py                 # Utility to generate shortened ASI scenario titles
├── generate_batch.py                  # Batch scenario generation utility
├── multi_asi_scenario.py              # Generator logic for multi-agent (multi-ASI) scenario narratives based on the single-ASI scenarios from asi_scenarios.db
├── multi_asi_ollama_client.py         # Connects to local Ollama LLM for multi-ASI scenarios
├── multi_asi_database.py              # Handles SQLite operations for multu-ASI scenarios storage at multi_asi_scenarios.db
├── parameter_sampler.py               # Defines how scenario parameters are randomly or manually sampled
├── single_asi_scenario.py             # Main script to generate a single ASI scenario
├── single_asi_ollama_client.py        # Connects to local Ollama LLM for single-ASI scenarios
├── single_asi_database.py             # Handles SQLite operations for sinle-ASI scenarios storage at asi_scenarios.db
└── ...

🧭 Roadmap

  • Validate results using JSON Schema
  • Add multiple LLM agent evaluators
  • Connect scenarios with precursor signals
  • Add narrative quality metrics (Coherence, Novelty, Diversity)

🪪 License

Licensed under the MIT License. © 2025 OASIOS Observatory Project. Open for research and educational use.

About

Experimental generator producing single-ASI and multi-ASI foresight scenarios using local LLM models

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages