TensorLogic Meta Crate

Unified access to all TensorLogic components

This is the top-level umbrella crate that re-exports all TensorLogic components for convenient access. Instead of importing individual crates, you can use this meta crate to access the entire TensorLogic ecosystem.

Overview

TensorLogic compiles logical rules (predicates, quantifiers, implications) into tensor equations (einsum graphs) with a minimal DSL + IR, enabling neural/symbolic/probabilistic models within a unified tensor computation framework.

Quick Start

Add to your Cargo.toml:

[dependencies]
tensorlogic = "0.1.0-beta.1"

Basic Usage

use tensorlogic::prelude::*;

// Define logical expressions
let x = Term::var("x");
let y = Term::var("y");
let knows = TLExpr::pred("knows", vec![x.clone(), y.clone()]);

// Compile to tensor graph
let graph = compile_to_einsum(&knows)?;

// Execute with backend
let mut executor = Scirs2Exec::new();
let result = executor.forward(&graph)?;

Architecture

The meta crate provides organized access to three layers:

Planning Layer (Engine-Agnostic)

use tensorlogic::ir::*;        // AST and IR types
use tensorlogic::compiler::*;  // Logic → tensor compilation
use tensorlogic::infer::*;     // Execution traits
use tensorlogic::adapters::*;  // Symbol tables, domains

Components:

• tensorlogic::ir - Core IR types (Term, TLExpr, EinsumGraph)
• tensorlogic::compiler - Logic-to-tensor mapping with static analysis
• tensorlogic::infer - Execution/autodiff traits (TlExecutor, TlAutodiff)
• tensorlogic::adapters - Symbol tables, axis metadata, domain masks

Execution Layer (SciRS2-Powered)

use tensorlogic::scirs_backend::*;  // SciRS2 runtime executor
use tensorlogic::train::*;           // Training infrastructure

Components:

• tensorlogic::scirs_backend - Runtime executor with CPU/SIMD/GPU features
• tensorlogic::train - Training loops, loss wiring, schedules, callbacks

Integration Layer

use tensorlogic::oxirs_bridge::*;      // RDF*/SHACL integration
use tensorlogic::sklears_kernels::*;   // ML kernels
use tensorlogic::quantrs_hooks::*;     // PGM integration
use tensorlogic::trustformers::*;      // Transformer components

Components:

• tensorlogic::oxirs_bridge - RDF*/GraphQL/SHACL → TL rules; provenance binding
• tensorlogic::sklears_kernels - Logic-derived similarity kernels for SkleaRS
• tensorlogic::quantrs_hooks - PGM/message-passing interop for QuantrS2
• tensorlogic::trustformers - Transformer-as-rules (attention/FFN as einsum)

Prelude Module

For convenience, commonly used types are available through the prelude:

use tensorlogic::prelude::*;

This imports:

• Core types: Term, TLExpr, EinsumGraph, EinsumNode
• Compilation: compile_to_einsum, CompilerContext, CompilationConfig
• Execution: TlExecutor, TlAutodiff, Scirs2Exec
• Errors: IrError, CompilerError

Examples

This crate includes 5 comprehensive examples demonstrating all features:

# Basic predicate and compilation
cargo run --example 00_minimal_rule

# Existential quantifier with reduction
cargo run --example 01_exists_reduce

# Full execution with SciRS2 backend
cargo run --example 02_scirs2_execution

# OxiRS bridge with RDF* data
cargo run --example 03_rdf_integration

# Comparing 6 compilation strategy presets
cargo run --example 04_compilation_strategies

Features

Compilation Strategies

TensorLogic supports 6 preset compilation strategies:

1. soft_differentiable - Neural network training (smooth gradients)
2. hard_boolean - Discrete Boolean logic (exact semantics)
3. fuzzy_godel - Gödel fuzzy logic (min/max operations)
4. fuzzy_product - Product fuzzy logic (probabilistic)
5. fuzzy_lukasiewicz - Łukasiewicz fuzzy logic (bounded)
6. probabilistic - Probabilistic interpretation

use tensorlogic::compiler::CompilationConfig;

let config = CompilationConfig::soft_differentiable();
let graph = compile_with_config(&expr, config)?;

Logic-to-Tensor Mapping

Default mappings (configurable per use case):

Feature Flags

Control which components are included:

[dependencies]
tensorlogic = { version = "0.1.0-beta.1", features = ["simd"] }

Available features:

• simd - Enable SIMD acceleration in SciRS2 backend (2-4x speedup)
• gpu - Enable GPU support (future)

Documentation

• Project Guide: CLAUDE.md
• API Reference: docs.rs/tensorlogic
• Main README: README.md
• Tutorial: Check individual crate READMEs for detailed guides

Component Documentation

Each component has comprehensive documentation:

• tensorlogic-ir - IR and AST types
• tensorlogic-compiler - Compilation
• tensorlogic-infer - Execution traits
• tensorlogic-scirs-backend - SciRS2 backend
• tensorlogic-train - Training
• tensorlogic-adapters - Symbol tables
• tensorlogic-oxirs-bridge - RDF* integration
• tensorlogic-sklears-kernels - ML kernels
• tensorlogic-quantrs-hooks - PGM integration
• tensorlogic-trustformers - Transformers

Development

Building

# Build the meta crate
cargo build -p tensorlogic

# Build with SIMD support
cargo build -p tensorlogic --features simd

# Run tests
cargo nextest run -p tensorlogic

# Run examples
cargo run -p tensorlogic --example 00_minimal_rule

Testing

The meta crate includes all component tests:

# Run all tests
cargo test -p tensorlogic --all-features

# Run with nextest (faster)
cargo nextest run -p tensorlogic --all-features

Version Compatibility

This meta crate version 0.1.0-beta.1 includes:

All components are synchronized to version 0.1.0-beta.1.

Migration from Individual Crates

If you were using individual crates:

Before:

[dependencies]
tensorlogic-ir = "0.1.0-beta.1"
tensorlogic-compiler = "0.1.0-beta.1"
tensorlogic-scirs-backend = "0.1.0-beta.1"

After:

[dependencies]
tensorlogic = "0.1.0-beta.1"

Your code remains the same, just update imports:

Before:

use tensorlogic_ir::{Term, TLExpr};
use tensorlogic_compiler::compile_to_einsum;

After:

use tensorlogic::ir::{Term, TLExpr};
use tensorlogic::compiler::compile_to_einsum;
// Or use prelude for common types
use tensorlogic::prelude::*;

Contributing

Contributions are welcome! Please see CONTRIBUTING.md for guidelines.

License

Licensed under Apache 2.0 License. See LICENSE for details.

References

• Tensor Logic Paper: https://arxiv.org/abs/2510.12269
• Project Repository: https://github.com/cool-japan/tensorlogic
• Documentation: https://docs.rs/tensorlogic

Part of the COOLJAPAN Ecosystem

For questions and support, please open an issue on GitHub.