📜 FTAI — Foundational Traceable AI Interface

FTAI is a hybrid format for human–AI collaboration, designed to improve on JSON, Markdown, and YAML for AI-native workflows with a structured, minimal, human-readable protocol.

Built for serious AI developers, researchers, and agent architects — but accessible enough for new builders learning to speak to machines.

🚀 Why FTAI?

• Readable by humans. Parseable by models.
• Deterministic: Always yields the same structure across systems.
• Traceable: Embeds rationale, constraints, and memory scopes directly.
• Composable: Bridges cleanly into JSON, YAML, text embeddings, or pipelines.
• Minimal: No noisy punctuation, no nested spaghetti.
• Multimodal: Native support for image references and vision-capable models.

FTAI is Markdown for agents. JSON for intelligence. YAML for reasoning.

🖼️ Multimodal Support

FTAI natively handles image references for vision-capable AI models:

@image
  src: ./screenshot.png
  alt: Application dashboard
  context: User is asking about the error shown in the top-right

@task
  analyze the image and identify the error message
@end

Images can be:

• Local file paths
• Base64 encoded inline
• URLs (when online processing is available)

This makes FTAI ideal for workflows involving screenshots, documents, diagrams, and visual context.

📦 Install

From PyPI (Recommended)

pip install ftai-py

From Source

git clone https://github.com/FolkTechAI/ftai-spec.git
cd ftai-spec
pip install -e .

⚡ Quick Start

# Lint an FTAI file
ftai lint myfile.ftai

# Lint with strict mode
ftai lint myfile.ftai --strict

# Lint with lenient mode (unknown tags as warnings)
ftai lint myfile.ftai --lenient

# Convert JSON to FTAI
ftai convert data.json > output.ftai

# Check version
ftai --version

🛠 What's Inside

📝 Basic Syntax

@ftai v2.0 lang:en

@document
  title: "My First FTAI Document"
  author: "Your Name"
  created: 2026-01-09

@section Introduction
  This is a simple FTAI document demonstrating the format.

@task priority:high
  description: Review the quarterly report
  due: 2026-01-15
@end

@note
  Remember to check the appendix for supporting data.

Core Tags

🤝 Contributing

We welcome thoughtful contributors! All Pull Requests require signing our Contributor License Agreement (CLA) first.

Local Development

git clone https://github.com/FolkTechAI/ftai-spec.git
cd ftai-spec
pip install -e .
pip install pytest

# Run tests
pytest tests/ -v

# Test the CLI
ftai lint tests/vectors/pass/pass_minimal.ftai

🛡 License & Governance

• License: Apache 2.0
• Governance: GOVERNANCE.md
• Security Disclosure: SECURITY.md

FTAI is stewarded with transparent, reviewable releases — designed for long-term stability, not churn.

🌱 Built in the Open

FTAI is designed for the future of machine interaction: clear enough for a person, structured enough for a model, powerful enough for autonomous systems.

Built openly by FolkTech AI with contributions from the community.

© 2025-2026 FolkTech AI — Format maintained by Michael Folk and contributors.