Architecture

ESBMC-AI is a platform that allows for easy development of APR methods. The platform contains components that provide a flexible and extensible framework for automated program repair research.

Core Architecture Components

The ESBMC-AI platform is built around several key architectural components:

Component System

• Base Component - Foundation for all extensible components including commands and verifiers
• Component factory pattern with automatic instantiation validation
• Configuration field integration for component-specific settings

AI Model Integration

• Abstract AIModel base class with concrete implementations for OpenAI, Anthropic, and Ollama
• Variable Substitution System - Dynamic template replacement for AI prompts
• Token counting and model management through AIModels singleton
• Caching system for model lists with configurable refresh intervals

Chat Interface System

• Base class for LLM interactions with message stack management
• Template substitution system for dynamic content
• Cooldown management between API requests
• Message compression and conversation handling

Command Architecture

• Commands extend ChatCommand base class and handle specific user requests
• Support for configuration fields and argument parsing
• Dynamic help generation from loaded components

Verifier System

• Abstract interface for code verification tools
• Standardized output format through VerifierOutput
• ESBMC integration with counterexample parsing and error extraction

Configuration System

• Singleton-based configuration management using BaseConfig
• Supports multiple sources: environment variables, config files, command-line args
• Hierarchical loading with .env file support
• Extensive validation and type checking for all config fields

Extension System

• Addon Loader - Dynamic loading of external components
• Plugin architecture for extending functionality
• Component registration and management