Skip to content

hyperpolymath/patallm-gallery

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

218 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

OpenSSF Best Practices

PatALLM Gallery

MPL-2.0-or-later :toc: macro :toclevels: 3 :icons: font :source-highlighter: rouge

LLM tools gallery - unified verification and antidote systems

Note

Project Status: Infrastructure Ready, Implementation Pending

This monorepo establishes governance, CI/CD, and technology standards for the PatALLM suite. The four satellite projects are defined as git submodules awaiting implementation.

Overview

PatALLM Gallery is a monorepo orchestrating a suite of LLM utility tools focused on:

  • Unification - A single interface to multiple LLM providers

  • Verification - Validating LLM outputs against source material

  • Antidote - Detecting and mitigating hallucinations

Architecture

patallm-gallery/
├── llm-unify-core/    # Core abstractions and traits
├── llm-unify/         # Provider adapters and routing
├── llm-verify/        # Output verification engine
└── llm-antidote/      # Hallucination detection/mitigation

Components

Repository Description Status

llm-unify-core

Core unified LLM interface - shared types, traits, and abstractions

🔲 Planned

llm-unify

LLM unification layer - provider adapters for OpenAI, Anthropic, local models

🔲 Planned

llm-verify

LLM output verification - validate responses against source documents

🔲 Planned

llm-antidote

LLM hallucination antidote - detection, scoring, and mitigation strategies

🔲 Planned

Vision

Problem Statement

Current LLM tooling suffers from:

  1. Fragmentation - Each provider requires different SDKs and patterns

  2. Blind Trust - Outputs accepted without verification against sources

  3. Hallucination Opacity - No systematic approach to detect or quantify confabulation

Solution

PatALLM provides:

  • A unified interface abstracting provider differences

  • Verification pipelines comparing outputs to source material

  • Hallucination detection with confidence scoring

  • Mitigation strategies (re-prompting, source injection, response filtering)

Technology Stack

This project follows the Hyperpolymath Standard language policy.

Allowed Technologies

Technology Use Case Notes

ReScript

Primary application code

Compiles to JS, type-safe

Deno

Runtime & package management

Replaces Node/npm/bun

Rust

Performance-critical, systems, WASM

Preferred for CLI tools and core libraries

Gleam

Backend services

Runs on BEAM or compiles to JS

Guile Scheme

State/meta files

.machine_readable/6a2/STATE.a2ml, .machine_readable/6a2/META.a2ml, .machine_readable/6a2/ECOSYSTEM.a2ml

Explicitly Banned

TypeScript, Node.js, npm, Go, Python (except SaltStack), Kotlin, Swift, React Native, Flutter

See CLAUDE.md for the complete policy.

Quick Start

# Clone with submodules
git clone --recursive https://github.com/hyperpolymath/patallm-gallery.git
cd patallm-gallery

# Initialize submodules if cloned without --recursive
git submodule update --init --recursive

Infrastructure

CI/CD Pipelines

Workflow Purpose

instant-sync.yml

Automatic forge propagation on push/release

mirror.yml

Mirrors to GitLab and Bitbucket

semgrep.yml

SAST security scanning

workflow-linter.yml

RSR security standards enforcement

Security Practices

  • SHA-pinned GitHub Actions

  • Semgrep SAST scanning

  • OpenSSF Scorecard compliance

  • SPDX license headers on all files

  • No hardcoded secrets

Contributing

Development Prerequisites

  • Deno >= 2.0

  • Rust >= 1.75 (for core components)

  • Git with submodule support

License

This project is licensed under the Mozilla Public License, v. 2.0. See the LICENSE file for details.

SPDX-License-Identifier: CC-BY-SA-4.0

Funding

Support this project:

Architecture

See TOPOLOGY.md for a visual architecture map and completion dashboard.

About

Gallery of LLM patterns and implementations

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Releases

Sponsor this project

Packages

Used by

Contributors

Languages