resonantiq/resonantforge

ResonantForge is a Python command-line tool that creates batches of synthetic two-party conversation transcripts designed for testing AI systems that score, classify, route, or analyze dialogues. Instead of using real customer conversations, you point it at a "profile" describing the conversation domain, supply a seed number, and it produces a structured dataset of dialogues with quality-signal annotations, source-document citations, participant trajectories, and planted human-review corrections already embedded. The central design principle is determinism. Run the tool twice with the same seed and you get the identical simulation structure, making it a reliable fixture in automated test builds. In dry-run mode, the tool skips all AI model calls and generates placeholder prose, so no API key is needed to start. In live mode, it calls an AI model to produce realistic conversational prose while keeping the structural metadata and planted signals reproducible across runs. The included profiles cover customer-support conversations for software-as-a-service companies and professional services businesses, but the architecture is domain-agnostic: developers can create profiles for any two-party conversation domain by subclassing a base class in the codebase. The abstract base class enforces the required interface so gaps surface immediately rather than producing silent empty outputs. ResonantForge is aimed at engineers building conversation evaluation pipelines who need realistic, controlled, and reproducible dialogue data without exposing real user conversations.