# MarkItDown Web Studio — Full System & API Architecture (llms-full.txt)

> Comprehensive technical documentation for AI Agents, LLM tool runners, and developers.

## Architecture
- Backend: FastAPI 0.115+ running on Uvicorn.
- Engine: Microsoft MarkItDown 0.1.8+ with native extensions.
- OCR Pipeline: PyPDFium2 for PDF page rasterization + native Apple Vision VNRecognizeTextRequest on macOS.
- Web Scraper: Requests + BeautifulSoup4 noise removal (scripts, styles, ads, cookies) + Markdownify.
- Privacy Guarantee: Zero persistent storage. In-memory temporary files are unlinked in finally blocks.

## REST API Specification

### 1. Document File Upload
- Endpoint: `POST https://markitdown.site/api/convert/file`
- Content-Type: `multipart/form-data`
- Body Parameter: `file` (Binary file content)
- Supported File Extensions: `.pdf`, `.docx`, `.doc`, `.xlsx`, `.xls`, `.csv`, `.pptx`, `.png`, `.jpg`, `.jpeg`, `.webp`, `.tiff`
- Response:
  ```json
  {
    "status": "success",
    "filename": "document.pdf",
    "markdown": "# Extracted Heading\n\nExtracted body text...",
    "elapsed_seconds": 0.42
  }
  ```

### 2. URL Web Clipper
- Endpoint: `POST https://markitdown.site/api/convert/url`
- Content-Type: `application/json`
- Request Body: `{"url": "https://example.com/article"}`
- Response: `{"status": "success", "url": "https://example.com/article", "markdown": "# Article Title..."}`

### 3. Raw HTML to Markdown
- Endpoint: `POST https://markitdown.site/api/convert/html`
- Content-Type: `application/json`
- Request Body: `{"html": "<h1>Title</h1><p>Body</p>"}`
- Response: `{"status": "success", "markdown": "# Title\n\nBody"}`

### 4. Table Formatter
- Endpoint: `POST https://markitdown.site/api/process/table`
- Content-Type: `application/json`
- Request Body: `{"markdown": "|col1|col2|\n|a|b|"}`
- Response: `{"status": "success", "markdown": "| col1 | col2 |\n|------|------|\n| a    | b    |"}`

### 5. AI Markdown Cleaner
- Endpoint: `POST https://markitdown.site/api/process/clean`
- Content-Type: `application/json`
- Request Body: `{"markdown": "<think>Thinking...</think>Sure! Here is the code..."}`
- Response: `{"status": "success", "markdown": "Here is the code..."}`

### 6. TOC Generator
- Endpoint: `POST https://markitdown.site/api/process/toc`
- Content-Type: `application/json`
- Request Body: `{"markdown": "# Section 1\n## Subsection A"}`
- Response: `{"status": "success", "markdown": "- [Section 1](#section-1)\n  - [Subsection A](#subsection-a)\n\n# Section 1..."}`