Chimera-Protocol/csl-core

v0.1.0 Alpha - First Release

We are excited to announce the first public alpha release of CSL-Core.

CSL-Core is a deterministic safety layer for probabilistic AI systems. It enables developers to define rigid, formally verified "laws" for AI agents and enforce them at runtime, independent of prompts or model training.

Our goal is to achieve "Solidity for AI Policies."

Key Features

Core Engine

• Chimera Specification Language (CSL): A domain-specific language designed for defining immutable safety constraints.
• Fail-Closed Runtime: A deterministic Python runtime that blocks actions by default unless explicitly allowed.
• Mathematical Rigor: Full AST parsing and semantic validation ensures policy integrity.

Formal Verification (Z3)

• Compile-Time Verification: Powered by the Microsoft Z3 Theorem Prover to mathematically prove policy consistency before deployment.
• Logic Checks: Automatically detects contradictions, unreachable states, and logical fallacies in rules.

Integrations

• LangChain Support: Native guard_tools wrapper and LCEL gate components to protect agents with minimal code.
• Context Injection: Securely inject runtime variables (e.g., user_role, environment) that the LLM cannot override.

Developer Experience

• CLI Tools: Verify (cslcore verify), simulate (cslcore simulate), and debug policies interactively (cslcore repl).
• Rich Visualization: Detailed terminal outputs for verification traces and errors.

What's Included

• Source code for Compiler, Verifier (Z3), and Runtime.
• Production-ready examples (Banking, Agent Tool Guard, DAO Treasury).
• Comprehensive test suite.
• Documentation and Quickstart guides.

Installation

You can install directly from the source:

pip install csl-core

Disclaimer

This is an Alpha Release. While the core logic is verified, API interfaces may change in future versions. Use in production environments with caution.