🤖 agent-logic

Modular Symbolic Logic for Agent Reasoning, SAT Solving, and LLM-Enhanced Proof Systems

🚀 Overview

agent-logic is a modular Python library for constructing, evaluating, transforming, and proving logical expressions — designed to act as a lightweight symbolic logic and SAT solving engine for LLM-driven agent systems.

It enables:

• ✅ Propositional Logic (AND, OR, IMPLIES, IFF, NOT)
• ✅ Predicate Logic (Terms, Predicates, Quantifiers: FORALL ∀, EXISTS ∃)
• ✅ Formal Proof Validation (Inference Rules, Structured Derivations)
• ✅ Truth Table Generation and Logical Satisfiability Checking
• ✅ Logical Transformations (Equivalences, Normal Forms)
• ✅ Recursive Abstract Syntax Tree (AST) Parsing
• ✅ Native Pydantic Models and Strong Typing for Safe Structured Outputs

🔗 Structured. Serializable. Reasonable. Agent-Ready.

💪 Current Status

• Core symbolic logic (propositions, connectives, predicates, quantifiers) is fully implemented.
• Truth tables, tautology/contradiction checking, and core proof validation are working and tested.
• Inference rules are largely implemented (Modus Ponens, Modus Tollens, Hypothetical Syllogism, Dilemmas, Biconditional Elimination, etc).
• AST parsing, SAT-based search, and deeper quantifier handling are in progress.

⚠️ Note:

• Some features (e.g., deeper quantifier transformations, large proof automation) are actively being debugged.
• Basic and intermediate logical operations are stable; complex proof search under refinement.

📊 Motivation

Large Language Models can predict, generate, and reflect — but they struggle with formal, structured, symbolic reasoning.

agent-logic empowers:

• Agents that perform valid, step-by-step derivations.
• LLMs that validate, transform, and construct proofs.
• Systems that reason explicitly over symbolic structures, not just language.

By combining a SAT-solving core, formal proof system, and structured Pydantic output models, it provides the foundation for autonomous, interpretable reasoning agents.

"Prediction ends where true reasoning begins."

💡 Key Features

💡 Example Usage

from agent_logic.core.operations import Proposition, BinaryOp
from agent_logic.evaluation.truth_table import TruthTable

# Define propositions
p = Proposition(name="P")
q = Proposition(name="Q")

# Create an expression: (P AND Q)
expr = BinaryOp(left=p, right=q, operator="AND")

# Generate a truth table
table = TruthTable(expression=expr)
for row in table.generate():
    print(row)

# Check logical properties
print("Is tautology:", table.is_tautology())
print("Is contradiction:", table.is_contradiction())

📙 LLM and Agent Toolkit Use Cases

• Formal proof verification of LLM-generated outputs
• Autonomous deduction chains in multi-agent debates
• Structured symbolic output parsing for LangChain tools / OpenAI functions
• Hypothetical reasoning, consequence checking, and goal validation
• Safe, introspectable logical reasoning pipelines for AI agents

All models use Pydantic v2, meaning:

• JSON-serializable and function-call ready
• Validatable against strict schemas
• Compatible with LangChain Structured Tools, OpenAI Tools, JSON mode parsing

"Not just token prediction. Formal reasoning."

🌟 Roadmap

• Propositional and Predicate Logic Core
• Truth Tables and Tautology Checking
• Structured Proof Validation Engine
• Advanced SAT Solving and Forward/Backward Proof Search
• Quantifier Manipulation (Skolemization, Unification)
• Natural Language to Formal Logic Parsing (Experimental)
• Web Visualizer Playground

🚀 Getting Started

pip install agent-logic

or from git:

git clone https://github.com/pr1m8/agent-logic.git
cd agent-logic
poetry install

👤 Authors

Built by developers passionate about combining symbolic logic, autonomous reasoning, and practical agentic AI design.

Contributions, ideas, and PRs are welcome!

🎉 License

MIT License.

Empower Your Agents with True Reason.

💡 "Teach your models to reason, not just predict."