Converge: Semantic Governance & Alignment

Converge is a vision for semantic governance. We move from fragmented intent to unified, converged states through a deterministic alignment engine. Our mission is to provide a stable foundation for complex decision-making where human authority and AI agency coexist in a transparent, explainable ecosystem.

Converge Tool 🛠️

Standalone developer tools for the Converge ecosystem. High-performance, LLM-powered validation, and generation of Gherkin specifications.

🚀 Overview

converge-tool is designed to bridge the gap between business requirements and technical implementation. It focuses on the "Truth" behind your specifications—ensuring they are not just syntax-valid, but provide real business value and are realistically automatable.

✨ Key Features

• Multi-Dimensional Validation: Evaluates specs for convention compliance, business sense, and technical compilability.
• AI Spec Generation: Transform free-text requirements into valid Truth (Gherkin) files instantly.
• Truth Syntax Support: Native support for Converge's Truth: keyword, maintaining full Gherkin compatibility.
• LLM Integration: Seamlessly connects with Anthropic, OpenAI, and other providers via converge-core traits.
• Scoring & Feedback: Receive detailed validation reports with confidence scores and actionable suggestions.

🛠️ Components

📐 Gherkin Validation

Validate your feature files against three rigorous dimensions:

1. Convention Compliance: Enforces Given/When/Then structure and naming standards.
2. Compilability: Analyzes if steps can be automated or if they rely on hidden human inputs/externalities.
3. Business Sense: Ensures scenarios are testable, deterministic, and describe meaningful business outcomes.

✍️ Spec Generation (New!)

Convert natural language descriptions into professional specifications.

# Try it out with the included example
ANTHROPIC_API_KEY=your_key cargo run --example generate_spec

💻 Installation

Add this to your Cargo.toml:

[dependencies]
converge-tool = "0.2"

⚡ Quick Start

Validation Example

use converge_tool::gherkin::{GherkinValidator, ValidationConfig};
use converge_provider::AnthropicProvider;
use std::sync::Arc;

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    // 1. Setup LLM Provider
    let provider = Arc::new(AnthropicProvider::from_env("claude-3-5-haiku-20241022")?);

    // 2. Initialize Validator
    let validator = GherkinValidator::new(provider, ValidationConfig::default());

    // 3. Validate Content
    let result = validator.validate(feature_text, "my_feature.truth")?;

    println!("Validation Score: {:.2}", result.confidence);
    Ok(())
}

🗺️ Roadmap & Architecture

See converge-business/knowledgebase/tool-ARCHITECTURE.md for a detailed look at the internal components and workflows.

• Knowledgebase: See converge-business/knowledgebase/

• Collaboration: See COLLABORATION.md

• For LLMs: See AGENTS.md

• Standalone Tool Extraction

• AI Spec Generation

• Validation Scoring

• VS Code Extension Integration

• Custom Domain Lexicons

📄 License & Conduct

This project is licensed under the MIT License - see the LICENSE file for details.

We expect all contributors to adhere to our Code of Conduct.

© 2024-2026 Aprio One AB, Sweden.

Development Tools

This project is built with the help of modern AI-assisted development tools:

AI Assistants

• Claude Code - Primary programming tool for development, refactoring, and architecture
• Cursor - AI-powered IDE for detailed code work
• Antigravity - AI pair programming for complex implementations
• Claude Opus 4.5 & Gemini 3 - Frontier models for staying on top of complex decisions

Version Control & Issue Tracking

• Jujutsu (jj) - Git-compatible VCS with first-class conflicts and undo
• Beads (bd) - Lightweight issue tracker with dependency support

# Quick workflow
jj status                 # See changes
jj commit -m "message"    # Commit
bd ready                  # Find unblocked work
bd create "Task" --type task  # Create issue

Key Rust Crates