ECHIDNA

*E*xtensible *C*ognitive *H*ybrid *I*ntelligence for *D*eductive *N*eural *A*ssistance

[](LICENSES/MIT.txt) [](LICENSES/Palimpsest-0.6.txt) [](https://api.reuse.software/info/gitlab.com/non-initiate/rhodinised/quill) [](https://gitlab.com/non-initiate/rhodinised/quill/-/pipelines)

A neurosymbolic theorem proving platform that transforms Quill (Agda-only neural solver) into a universal multi-prover system supporting 12 theorem provers with aspect tagging, OpenCyc integration, and DeepProbLog probabilistic logic.

Features

• 12 Theorem Prover Support: Universal backend supporting Agda, Coq/Rocq, Lean, Isabelle, Z3, CVC5, Metamath, HOL Light, Mizar, PVS, ACL2, and HOL4

• Neurosymbolic AI: Neural proof synthesis combined with symbolic verification

• Aspect Tagging: Intelligent proof categorization and analysis

• OpenCyc Integration: Common sense reasoning with ontological knowledge

• DeepProbLog: Probabilistic logic programming for uncertain reasoning

• 4-Language Architecture: Rust (core), Julia (ML), ReScript (UI), Mercury/Logtalk (logic)

Quick Start

Prerequisites

• Rust 1.75 or later

• Julia 1.10 or later

• Deno 1.40 or later (for ReScript)

• Podman (NOT Docker)

• Just command runner

Installation

= Clone the repository
git clone https://gitlab.com/non-initiate/rhodinised/quill.git
cd quill

= Build the project
just build

= Run tests
just test

= Run ECHIDNA
just run

Using Podman Container

= Build container
podman build -f Containerfile -t echidna:latest .

= Run container
podman run -it echidna:latest

Architecture

Prover Tiers

Tier 1 (6 provers - Months 2-4): - Agda - Dependent type theory - Coq/Rocq - Interactive theorem prover - Lean - Functional programming and proving - Isabelle - Higher-order logic - Z3 - SMT solver - CVC5 - SMT solver

Tier 2 (3 provers - Months 5-7): - Metamath - Plain text verifier (easiest, start here!) - HOL Light - Classical higher-order logic - Mizar - Mathematical vernacular

Tier 3 (2 provers - Months 8-10): - PVS - Specification and verification - ACL2 - Applicative Common Lisp

Tier 4 (1 prover - Months 11-12): - HOL4 - Higher-order logic

Technology Stack

• Rust: Core logic, FFI, WASM compilation

• Julia: Machine learning components (NO Python allowed!)

• ReScript + Deno: User interface

• Mercury/Logtalk: Optional logic reservoir

Key Components

• echidna_provers.rs: Complete Rust trait system (600+ lines)

• Aspect Tagging System: Proof classification and analysis

• Neural Solver: Integration from original Quill project

• Universal Prover Abstraction: Unified interface for all 12 provers

Usage

Basic Theorem Proving

use echidna::prover::{Prover, AgdaProver};

let prover = AgdaProver::new();
let result = prover.prove("∀ (n : ℕ) → n + 0 ≡ n")?;
println!("Proof: {}", result);

Using Multiple Provers

use echidna::prover::{ProverBackend, lean, coq, z3};

// Try multiple provers
let backends = vec![lean(), coq(), z3()];
for backend in backends {
    if let Ok(proof) = backend.prove(theorem) {
        println!("Proved with {}: {}", backend.name(), proof);
        break;
    }
}

Neural Proof Synthesis

using ECHIDNA

= Train neural model
model = train_proof_synthesizer(training_data)

= Generate proof candidates
candidates = synthesize_proofs(model, theorem)

= Verify with symbolic prover
for candidate in candidates
    if verify_proof(candidate)
        return candidate
    end
end

Development

Building from Source

= Install dependencies
just deps

= Build all components
just build

= Build specific component
just build-rust
just build-julia
just build-rescript

Running Tests

= Run all tests
just test

= Run specific language tests
just test-rust
just test-julia

= Run integration tests
just test-integration

Quality Checks

= Run all quality checks
just check

= Individual checks
just lint          # REUSE, rustfmt, clippy
just security      # Trivy, cargo-audit
just coverage      # Test coverage

Documentation

• [Contributing Guidelines](CONTRIBUTING.md)

• [Code of Conduct](CODE_OF_CONDUCT.md)

• [Security Policy](SECURITY.md)

• [Changelog](CHANGELOG.md)

• [API Documentation](docs/)

• [Prover Backend Guides](docs/)

Roadmap

12-Month Implementation Plan

Months 1-2: Foundation - Deploy RSR/CCCP compliance templates - Migrate Python code to Julia - Complete Tier 1 infrastructure

Months 2-4: Tier 1 Provers - Coq/Rocq implementation - Lean implementation - Isabelle implementation - Z3 and CVC5 integration

Months 5-7: Tier 2 Provers - Metamath (start here - easiest!) - HOL Light - Mizar

Months 8-10: Tier 3 Provers - PVS - ACL2

Months 11-12: Tier 4 and Polish - HOL4 implementation - Performance optimization - Documentation completion

Contributing

We welcome contributions! Please see [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines.

Critical Constraints

• ❌ NO PYTHON - Use Julia for all ML code

• ✅ RSR/CCCP Compliance - Follow Rhodium Standard Repository guidelines

• ✅ Justfile PRIMARY - Use Just, not Make

• ✅ GitLab-first - This is a GitLab project

• ✅ Podman not Docker - Always use Podman

• ✅ Dual Licensing - MIT + Palimpsest v0.6

License

This project is dual-licensed under:

• [MIT License](LICENSES/MIT.txt)

• [Palimpsest License v0.6](LICENSES/Palimpsest-0.6.txt)

See [NOTICE](NOTICE) for complete license information.

Citation

If you use ECHIDNA in your research, please cite:

@software{echidna2025,
  title = {ECHIDNA: Extensible Cognitive Hybrid Intelligence for Deductive Neural Assistance},
  author = {ECHIDNA Project Contributors},
  year = {2025},
  url = {https://gitlab.com/non-initiate/rhodinised/quill},
  license = {MIT AND Palimpsest-0.6}
}

Acknowledgments

• Based on the Quill project (Agda neural solver)

• Built on the shoulders of the theorem proving community

• See [AUTHORS.md](AUTHORS.md) for contributor list

Contact

• GitLab Issues: https://gitlab.com/non-initiate/rhodinised/quill/-/issues

• Merge Requests: https://gitlab.com/non-initiate/rhodinised/quill/-/merge_requests

• Security: security@echidna-project.org

---

Version: 0.1.0 Status: Active Development Last Updated: 2025-11-22