cognition Documentation

Welcome to cognition 🧠, an open-source Python framework developed by Polycog, Inc. designed to facilitate the creation of trustworthy cognitive agents via a modular API that handles complex decision-making, natural language processing, structural knowledge representation, and automated reasoning.

Key affordances:

• Enforce (hierarchical) agent orchestration via a knowledge-augmented DecisionProcess - a declarative, state-driven execution machine that continually matches against policy directives.

• Declare domain knowledge using validated data models (Entity, BinaryRelation), related & queried within a dynamic WorldGraph.

• Integrate the cognitive agent loop with custom environmental sensors and/or actuators via a Cogent (similar to a harness in LLM-based systems).

• Use your choice of language model(s) for classification (e.g., intent recognition via EnumClassifier), description (e.g., human-readable situational awareness via describe_facts), and population (e.g., extract structured data via ModelPopulator) tasks, where context is automatically infused with expert knowledge.

• Integrate symbolic reasoning to efficiently & verifiably solve sub-problems, such as using a SearchPlanner to find an optimal sequence of steps to achieve a goal.

Note

This project is under active development!