A multi-AI theorem prover integrating SMT solving, type theory, homotopy type theory, and geometric reasoning.
GPVE combines multiple reasoning paradigms:
┌─────────────────────────────────────────────────────┐
│ Layer 6: AI Exploration (Gemini, DeepSeek, GPT) │
├─────────────────────────────────────────────────────┤
│ Layer 5: Geometric Proof Space │
├─────────────────────────────────────────────────────┤
│ Layer 4: Higher Category Theory │
├─────────────────────────────────────────────────────┤
│ Layer 3: Sheaf-based Modular Reasoning │
├─────────────────────────────────────────────────────┤
│ Layer 2: Type-theoretic Logical Core │
├─────────────────────────────────────────────────────┤
│ Layer 1: SMT / Symbolic Solving (Z3) │
└─────────────────────────────────────────────────────┘
- SMT Solving — Z3 integration for constraint solving
- Proof Graphs — Directed acyclic graphs representing proofs
- Type Theory — Dependent types and proof kernel (HoTT)
- Geometric Reasoning — Simplicial complexes, homotopy detection
- Category Theory — Morphisms, 2-morphisms, higher categories
- Sheaf Theory — Modular proof reasoning
- Multi-AI Orchestration — Gemini, DeepSeek, ChatGPT
- Visualization — DOT, Mermaid, JSON, Matplotlib
# Clone the repo
git clone https://github.com/chrispusnjoroge15/gpve.git
cd gpve
# Install dependencies
pip install -r requirements.txt
# Optional: Install matplotlib for advanced visualization
pip install matplotlibSet environment variables for AI features:
export OPENAI_API_KEY="sk-..."
export GEMINI_API_KEY="..."
export DEEPSEEK_API_KEY="..."The core works without API keys — SMT solving, proof graphs, topology, etc.
from gpve import create_engine
from gpve.core import ProofGraph
# Create engine
engine = create_engine()
# Build a proof
graph = ProofGraph("demo")
v1 = graph.add_vertex("x > 0")
v2 = graph.add_vertex("y > 0")
v3 = graph.add_vertex("x + y > 0")
graph.add_edge(v1.id, v3.id, rule_name="add_positive")
graph.add_edge(v2.id, v3.id, rule_name="add_positive")
# Verify
result = engine.full_verify("x + y > 0", graph)
print(f"Valid: {result.is_valid}")# Run full demo
python -m gpve.cli all
# Run SMT solver
python -m gpve.cli smt
# Run tests
python -m gpve.cli test
# Run examples
python -m gpve.cli examples
# Query AI
python -m gpve.cli ai "Prove that sqrt(2) is irrational"
# Visualize proof
python -m gpve.cli visualize
# Check AI status
python -m gpve.cli ai-status| Command | Description |
|---|---|
gpve all |
Run full demo |
gpve verify |
Verify a proof |
gpve smt |
Run SMT solver |
gpve test |
Run unit tests |
gpve examples |
Run example proofs |
gpve ai <prompt> |
Query AI providers |
gpve ai-status |
Check API keys |
gpve visualize |
Visualize proof graph |
gpve sheaf |
Sheaf reasoning demo |
gpve category |
Category theory demo |
gpve topology |
Topology analysis |
gpve/
├── gpve/
│ ├── core/ # SMT, proof graph, kernel
│ │ ├── proof_graph.py
│ │ ├── smt_solver.py
│ │ ├── proof_kernel.py
│ │ ├── functional_analysis.py
│ │ └── reverse_eng.py
│ ├── geometric/ # Topology, homotopy
│ │ ├── simplicial_complex.py
│ │ ├── homotopy.py
│ │ └── topological_proof.py
│ ├── category/ # Category theory
│ │ ├── morphisms.py
│ │ └── higher_morphisms.py
│ ├── sheaf/ # Modular reasoning
│ │ ├── sheaf_structure.py
│ │ └── gluing.py
│ ├── ai/ # Multi-AI orchestration
│ │ └── orchestrator.py
│ ├── engine.py # Unified engine
│ ├── cli.py # CLI
│ ├── examples.py # Examples
│ ├── visualize.py # Visualization
│ └── config.py # Configuration
├── tests/ # Unit tests
├── pyproject.toml
├── requirements.txt
└── README.md
from gpve.core import SMTSolver
solver = SMTSolver()
x = solver.Int("x")
y = solver.Int("y")
solver.assert_formula(x + y == 10)
solver.assert_formula(x > 0)
result = solver.check()
print(result.model) # {'x': '1', 'y': '9'}from gpve import create_engine
engine = create_engine()
engine.add_proposition("P")
engine.add_proof_morphism("P", "Q", "f")from gpve import create_engine
engine = create_engine()
results = engine.ai_explore("Prove Fermat's Last Theorem")# Run all tests
python -m gpve.cli test
# Or with pytest
pytest tests/ -vGPVE supports multiple output formats:
- Mermaid —
gpve visualize -f mermaid - DOT/GraphViz —
gpve visualize -f dot - JSON —
gpve visualize -f json - Matplotlib — Requires
pip install matplotlib
- Python 3.10+
- z3-solver
- openai
- google-generativeai
- pydantic
- rich
MIT License
Chrispus Njoroge
If you use GPVE in research, please cite:
@software{gpve,
title = {GPVE - Geometric Proof \& Verification Engine},
author = {Chrispus Njoroge},
year = {2026},
url = {https://github.com/chrispusnjoroge15/gpve}
}