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

# Document Sources

> Ingest static files into the Organization Brain.

If your team relies on external SaaS platforms or legacy systems that can export data but don't support direct database connections, you can ingest static files (Sources) into the Organization Brain.

## Uploading a document

Navigate to **Brain → Sources** and use the **Upload Data** interface (or simply drop a file into the Corgtex Slack bot, if configured).

Supported formats include:

* PDF
* DOCX
* TXT
* MD / MDX
* CSV

## The ingestion and chunking process

When you upload a file, Corgtex doesn't just store the blob. The worker processes the file through our background pipeline:

1. **Extraction:** Specialized parsers (like `pdf-parse` or `mammoth`) extract the raw text content.
2. **Chunking & Storage:** The text is sliced into manageable semantic chunks (`KnowledgeChunk`), embedded, and saved to the vector database.

## PII Classification

Before any text is embedded, it runs through the Zero-LLM PII Regex Classifier. This is a crucial safety step that happens entirely locally (without sending your data to external APIs):

* The system scans the chunks using deterministic regex patterns.
* If it detects Personal Identifiable Information (PII) or confidential markers, it tags the chunk accordingly.
* Tagged chunks are strictly isolated and are only returned to workflows explicitly authorized to handle them. They will *never* be exposed in standard conversational agent responses.