> ## Documentation Index
> Fetch the complete documentation index at: https://docs.gosutox.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Multimodal Data Files Vault

> Supported file categories, comprehensive format extensions, size limits, and 10-file multi-context staging.

## Supported File Categories & Exhaustive Formats

AXON supports native multimodal ingestion across **6 core file categories** with strict zero-loss guarantees:

| Category                        | Supported Format Extensions                                                                                                                                 | Single-File Limit    | Policy & Pre-processing                                                                              |
| :------------------------------ | :---------------------------------------------------------------------------------------------------------------------------------------------------------- | :------------------- | :--------------------------------------------------------------------------------------------------- |
| **Documents**                   | `PDF`, `TXT`, `MD`, `MDX`, `HTML`, `HTM`                                                                                                                    | **30MB / 300 pages** | Word (`DOCX`), PowerPoint (`PPTX`), and Hangul (`HWP`) must be converted to **PDF** prior to upload. |
| **Spreadsheets & Tabular Data** | `XLSX`, `XLS`, `CSV`, `TSV`, `PARQUET`                                                                                                                      | **30MB**             | Direct vector & formula parsing.                                                                     |
| **Code & Configuration**        | `PY`, `JS`, `TS`, `TSX`, `JSX`, `VUE`, `CSS`, `SCSS`, `JSON`, `JSONL`, `YAML`, `YML`, `SQL`, `SH`, `BASH`, `XML`, `TOML`, `JAVA`, `GO`, `RS`, `INI`, `CONF` | **30MB**             | High-density semantic AST code indexing.                                                             |
| **Images & Visuals**            | `PNG`, `JPG`, `JPEG`, `WEBP`, `HEIC`, `HEIF`, `BMP`, `TIFF`, `SVG`                                                                                          | **10MB**             | Ultra-high-resolution multimodal vision analysis.                                                    |
| **Audio & Voice**               | `MP3`, `WAV`, `AAC`, `OGG`, `FLAC`, `AIFF`                                                                                                                  | **50MB**             | Whisper-based automatic transcription & semantic indexing.                                           |
| **Video & Recordings**          | `MP4`, `MOV`, `AVI`, `WEBM`, `MPEG`, `MPG`, `WMV`, `3GPP`                                                                                                   | **50MB**             | Multimodal visual frame & audio track extraction.                                                    |

<Note>
  **The PDF Pre-Processing Policy**: To eliminate server-side layout skew, dropped table cells, and AI hallucinations from raw office files, AXON strictly requires Word (`.docx`), PowerPoint (`.pptx`), and Hangul (`.hwp`) documents to be baked to **PDF** prior to uploading. See [Why Office Files Must Be Converted to PDF](/data/why-pdf-conversion) for full architectural details.
</Note>

***

## 10-File Multi-Context Turn Staging

In the **Right Context Tools Panel**, users can select up to **10 files simultaneously** across any combination of categories for a single reasoning turn:

```mermaid theme={null}
flowchart TD
    subgraph Staged_Vault ["📁 10-File Multimodal Staging Vault"]
        F1["📄 Patent_Claims.pdf"]
        F2["📊 Q3_Financial_Model.xlsx"]
        F3["🖼️ System_Architecture.png"]
        F4["💻 Contract_Logic.py"]
    end

    Staged_Vault --> Synthesis["⚡ Unified 1-Turn Cognitive Reasoning"]
    Synthesis --> Output["🎯 Comprehensive Boardroom Deliverable"]
```

### The Power of Multimodal Staging

* **Direct Token Stream**: Files are ingested with zero OCR degradation and structured mathematical precision.
* **Complex Data Fusion**: Cross-examine a corporate legal PDF, an Excel valuation model, and an architectural diagram in a single conversation turn.