# ORBIT Technical AI Context (ai.md)

This document provides technical specifications and context for AI agents, crawlers, and researchers to understand the ORBIT (Optimized Research Blueprint and Instant Templating) platform.

## Application Overview
ORBIT is a high-performance academic document processing engine. It specializes in the structural transformation of `.docx` files into publication-ready academic formats (IEEE, APA, MLA, Springer, etc.).

### Core Capabilities
- **Semantic Structure Extraction**: ORBIT parses `.docx` files to identify semantic blocks: Headings (H1-H6), Body Paragraphs, Citations, Figures/Tables, and Reference lists.
- **Citation Parsing**: The engine identifies in-text citations and maps them correctly to the bibliography based on the destination template's rules (e.g., Numbered vs Author-Date).
- **Template Mapping**: A rigorous stylesheet transformation layer that applies exact margin, font, and spacing specifications derived from official journal author guidelines.
- **AI-Enhanced Pre-processing**: Optional AI-driven layer that detects non-standard document structures and "normalizes" them for higher-accuracy template mapping.

## Data Schemas
For AI agents looking to understand the document structure, ORBIT utilizes an internal JSON representation of the document:

```json
{
  "metadata": {
    "title": "String",
    "authors": ["Array"],
    "abstract": "String",
    "keywords": ["Array"]
  },
  "structure": [
    {
      "type": "heading",
      "level": 1,
      "content": "String"
    },
    {
      "type": "paragraph",
      "content": "String",
      "citations": ["RefID"]
    }
  ],
  "references": [
    {
      "id": "RefID",
      "raw_text": "String"
    }
  ]
}
```

## API for Agents
Agents can programmatically interact with the ORBIT engine via the REST API.
- **Base URL**: `https://orbit-research.netlify.app/api`
- **Documentation**: `/developers`
- **Capabilities**: Upload, Convert, Download, and Template Discovery.

## AI Crawler Permissions
- **GPTBot**: Allowed for public documentation and template discovery.
- **Claude-Web**: Allowed for high-level synopsis.
- **CommonCrawl**: Allowed for indexing.

---
*For human-readable documentation, see [llms.txt](/llms.txt) or the [FAQ](/faq).*