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

# Ingesting Materials

> Learn how to ingest videos, PDFs, podcasts, and other expert materials into Mega Brain

Mega Brain accepts any format of expert content and transforms it into structured knowledge. This guide walks you through ingesting different material types into the system.

## Overview

The ingestion process automatically:

* Downloads and transcribes content
* Extracts metadata (author, title, date)
* Saves files with source traceability
* Detects duplicates to prevent reprocessing
* Prepares materials for the processing pipeline

<Note>
  All ingested materials are saved to the `inbox/` directory organized by person and content type.
</Note>

## Supported Material Types

<Tabs>
  <Tab title="YouTube Videos">
    ### Ingesting YouTube Videos

    The most common source. JARVIS automatically downloads transcriptions using the YouTube Transcript API.

    ```bash theme={null}
    /ingest https://www.youtube.com/watch?v=EXAMPLE_ID
    ```

    **What happens:**

    1. Video metadata is fetched (title, author, duration)
    2. Transcript is downloaded automatically
    3. File is saved as `inbox/[PERSON]/[TYPE]/video-title.txt`
    4. A unique SOURCE\_ID is generated (e.g., `CG003` for Cole Gordon video 3)

    **Example output:**

    ```
    JARVIS: Material received.

      Source:    YouTube
      Title:     "How to Create Irresistible Offers"
      Author:    Alex Hormozi
      Duration:  42:15
      Words:     6,230

      Saved to: inbox/alex-hormozi/MASTERCLASS/how-to-create-irresistible-offers.txt

      Next step: execute /process-jarvis to process.
    ```
  </Tab>

  <Tab title="PDFs & Documents">
    ### Ingesting PDFs and Documents

    Upload local files including PDFs, Word documents, and text files.

    ```bash theme={null}
    /ingest /path/to/playbook-vendas.pdf
    ```

    **Supported formats:**

    * PDF documents
    * Microsoft Word (.doc, .docx)
    * Plain text (.txt, .md)
    * Google Docs (via URL)

    <Tip>
      For best results, ensure PDFs have text layers (not just scanned images).
    </Tip>
  </Tab>

  <Tab title="Podcasts & Audio">
    ### Ingesting Podcasts and Audio

    Process audio content through transcription.

    ```bash theme={null}
    /ingest https://podcast-url.com/episode-123
    ```

    **Requirements:**

    * `OPENAI_API_KEY` configured (uses Whisper for transcription)

    **Content detection:** The system automatically detects podcast episodes based on keywords ("podcast", "episode", "ep", "interview"), file metadata, and URL patterns.
  </Tab>

  <Tab title="Courses & Training">
    ### Ingesting Course Materials

    Process complete training courses with multiple modules.

    ```bash theme={null}
    /ingest /path/to/course-module-1.pdf --type COURSE
    ```

    **Scope detection:**
    Course materials are automatically tagged with scope:

    * `scope: "course"` - Complete training programs
    * `corpus: "[course-name]"` - Groups related modules

    **Organization:**

    ```
    inbox/
    └── setterlun/
        └── COURSES/
            ├── module-01-foundation.txt
            ├── module-02-strategy.txt
            └── module-03-execution.txt
    ```
  </Tab>
</Tabs>

## Advanced Ingestion Options

### Manual Metadata Override

Override auto-detected metadata:

```bash theme={null}
# Specify person explicitly
/ingest https://youtube.com/watch?v=abc123 --person "Cole Gordon"

# Set content type
/ingest /path/to/file.txt --type MASTERCLASS

# Ingest and immediately process
/ingest https://youtube.com/watch?v=abc123 --process
```

### Content Type Classification

<Accordion title="Available Content Types">
  | Type        | Auto-Detection Keywords                 | Use Case                 |
  | ----------- | --------------------------------------- | ------------------------ |
  | PODCAST     | "podcast", "episode", "interview"       | Conversational content   |
  | MASTERCLASS | "masterclass", "mastermind", "training" | Expert workshops         |
  | COURSE      | "course", "module", "lesson"            | Structured programs      |
  | BLUEPRINT   | "blueprint", "pdf", "playbook"          | Strategic documents      |
  | VSL         | "vsl", "webinar", "sales letter"        | Sales presentations      |
  | SCRIPT      | "script", "template", "copy"            | Templates and frameworks |
</Accordion>

## Known Source Detection

The system recognizes expert sources automatically:

<Steps>
  <Step title="Path Analysis">
    Analyzes the file path or URL to detect known experts:

    ```
    inbox/COLE GORDON/MASTERMINDS/video.txt
           ↓
    SOURCE_PERSON: Cole Gordon
    SOURCE_COMPANY: Cole Gordon
    CORPUS: closers_io
    ```
  </Step>

  <Step title="Known Sources Library">
    | Detected Pattern         | Person        | Company              | Corpus           |
    | ------------------------ | ------------- | -------------------- | ---------------- |
    | "hormozi", "acquisition" | Alex Hormozi  | Alex Hormozi         | acquisition\_com |
    | "cole gordon", "closers" | Cole Gordon   | Cole Gordon          | closers\_io      |
    | "leila"                  | Leila Hormozi | Alex Hormozi         | acquisition\_com |
    | "setterlun", "sam ovens" | Sam Ovens     | Setterlun University | setterlun        |
  </Step>

  <Step title="Automatic Corpus Assignment">
    Related materials from the same expert are grouped into a corpus for cross-referencing.
  </Step>
</Steps>

## Duplicate Detection

<Warning>
  The pipeline includes 6 levels of duplicate detection to prevent reprocessing:
</Warning>

### Detection Levels

1. **Exact MD5 Match** - Same file already processed
2. **Content Hash** - Different file, same content
3. **Partial Match** - Content is substring of processed material
4. **YouTube ID** - Same video from different source
5. **File Registry** - Cross-reference with processing history
6. **Chunk Fingerprint** - First 1000 characters comparison

**When duplicates are found:**

```
┌─────────────────────────────────────────────────────────┐
│  ⛔ DUPLICATE DETECTED - PROCESSING INTERRUPTED        │
├─────────────────────────────────────────────────────────┤
│                                                         │
│  Current file:  inbox/new-video.txt                     │
│  MD5:           a1b2c3d4e5f6...                         │
│                                                         │
│  Duplicate of:  inbox/old-video.txt                     │
│  Registered:    2026-02-15                              │
│  SOURCE_ID:     CG005                                   │
│                                                         │
│  ACTION: File will NOT be processed.                    │
│                                                         │
└─────────────────────────────────────────────────────────┘
```

## Batch Ingestion

Process multiple materials at once:

<CodeGroup>
  ```bash Ingest Multiple URLs theme={null}
  # List of YouTube videos
  /ingest https://youtube.com/watch?v=video1
  /ingest https://youtube.com/watch?v=video2
  /ingest https://youtube.com/watch?v=video3
  ```

  ```bash Ingest Directory theme={null}
  # Process all files in a directory
  /scan-inbox /path/to/materials/
  ```
</CodeGroup>

## Verification

### Check Inbox Status

View pending materials:

```bash theme={null}
/inbox
```

**Output:**

```
═══════════════════════════════════════════════
                INBOX STATUS
═══════════════════════════════════════════════

📥 PENDING: 3 files

  Cole Gordon (2 files)
  ├─ masterclass-closing.txt (6,234 words)
  └─ podcast-interview.txt (8,901 words)

  Alex Hormozi (1 file)
  └─ offer-creation-framework.txt (4,567 words)

⭐ NEXT: /process-jarvis to process pending files
```

### Check System Status

```bash theme={null}
/jarvis-briefing
```

Provides:

* Health score (0-100)
* Processed materials count
* Pending inbox files
* Active agents status

## Next Steps

<Card title="Process Ingested Materials" icon="gears" href="/guides/processing-pipeline">
  Learn how to run the 5-phase processing pipeline on ingested materials
</Card>

<Card title="View Processing Results" icon="chart-line" href="/guides/session-management">
  Check processing logs and generated artifacts
</Card>