evocuriosity 0.2.2

A local-first Python SDK implementing a cognitive architecture for curiosity-driven AI.

EvoCuriosity SDK

A production-grade, local-first cognitive AI framework implementing a modular curiosity-driven architecture.

Overview

EvoCuriosity is designed to provide a robust foundation for building self-evolving AI systems that prioritize local-first execution and curiosity-driven discovery. The architecture facilitates complex reasoning through a multi-agent orchestration layer, ensuring high adaptability and efficient knowledge management.

Key Features

• Multi-Agent Reasoning: Specialized Researcher, Reasoner, Critic, and Planner agents collaborating within a unified cognitive loop.
• Database Integration: Comprehensive support for SQL, MongoDB, and Vector databases through a standardized adapter interface.
• Curiosity-Driven Logic: Advanced gap detection and hierarchical question generation based on uncertainty and novelty thresholds.
• Local-First Architecture: Optimized for offline execution with pluggable LLM adapters for varied computational environments.

Installation

The SDK can be installed via pip:

pip install evocuriosity

Quick Start

from evocuriosity.core import CuriosityEngine
from evocuriosity.connectors.mongo import MongoConnector

# 1. Initialize Engine and Database
db = MongoConnector(uri="mongodb://localhost:27017")
ai = CuriosityEngine(db_connector=db)

# 2. Configure Intelligence Adapter
ai.attach_llm("rule-based")

# 3. Cognitive Observation and Execution
ai.observe("Theoretical implications of quantum entanglement")
ai.think()

# 4. Retrieve Structured Output
results = ai.get_output()
print(results)

Architecture

• core/: High-level orchestration and execution loops.
• agents/: Collaborative cognitive agents for research, reasoning, and planning.
• adapters/: Abstraction layers for Large Language Models and Databases.
• connectors/: Specialized database interface implementations.
• memory/: Multi-tiered storage for semantic, episodic, and relational data.

Author

Daksha Dubey

License

This project is licensed under the MIT License. See the LICENSE file for details.