Skip to main content

​
/ingest - Material Ingestion

Ingests external materials into the Mega Brain inbox, extracting metadata and preparing content for pipeline processing.

​
Syntax

/ingest [SOURCE] [FLAGS]
​
SOURCE
string
required
YouTube URL, local file path, or Google Drive link

​
Supported Sources

YouTube

Videos with automatic transcription

Local Files

.txt, .pdf, .docx documents

Google Drive

Docs, PDFs, Sheets

​
Flags

​
--person
string
Manually specify the expert/person name (overrides automatic detection)Example: --person "Cole Gordon"
​
--type
string
Content type: PODCAST, MASTERCLASS, COURSE, BLUEPRINT, VSL, etc.Example: --type MASTERCLASS
​
--process
boolean
Automatically start pipeline processing after ingestionExample: --process

​
Examples

/ingest https://youtube.com/watch?v=abc123

​
How It Works

​
Step 1: Source Type Detection

The command identifies the source type: 
IF URL starts with "http":
  IF contains "youtube.com" or "youtu.be":
    → TYPE = "YOUTUBE"
    → Fetch transcript via youtube-transcript-api
  ELSE IF contains "docs.google.com":
    → TYPE = "GDOC"
    → Download content via Google API
  ELSE:
    → TYPE = "WEB"
    → Fetch page content
ELSE:
  → TYPE = "LOCAL"
  → Read file directly

​
Step 2: Metadata Extraction

Automatic metadata detection: 
Known Patterns:
  "hormozi" OR "acquisition" → Alex Hormozi
  "cole gordon" OR "closers" → Cole Gordon
  "leila" → Leila Hormozi
  "setterlun" OR "sam ovens" → Sam Ovens
  "jordan lee" → Jordan Lee
If --person flag provided, uses that instead.

​
Step 3: Destination Path

Files are saved to structured inbox: 
inbox/
├── {PERSON} ({COMPANY})/
│   ├── PODCASTS/
│   ├── MASTERCLASSES/
│   ├── COURSES/
│   └── BLUEPRINTS/
└── ...
Filename Format:
  • YouTube: {VIDEO_TITLE} [youtube.com_watch_v={ID}].txt
  • Local: {ORIGINAL_NAME}.txt
  • Google Drive: {DOC_TITLE}.txt

​
Step 4: Generate Source ID

Each file gets a unique identifier: 
Format: {INITIALS}{NUMBER}

Examples:
  CG005  # Cole Gordon, 5th file
  AH023  # Alex Hormozi, 23rd file
  JL010  # Jordan Lee, 10th file

​
Step 5: Ingest Report

Display confirmation: 
═══════════════════════════════════════════════════════════════════════════════
                          INGEST REPORT
                     2026-03-06 10:45:23 UTC
═══════════════════════════════════════════════════════════════════════════════

📥 MATERIAL INGERIDO
   Fonte: https://youtube.com/watch?v=abc123
   Tipo: VIDEO (YouTube)
   Título: "How to Scale from $1M to $10M ARR"

📁 DESTINO
   Path: inbox/ALEX HORMOZI (Acquisition.com)/MASTERCLASSES/
   Arquivo: How to Scale from $1M to $10M ARR [youtube.com_watch_v=abc123].txt
   Source ID: AH024

📊 ESTATÍSTICAS
   Palavras: 8,542
   Duração: 45m 32s
   Pessoa detectada: Alex Hormozi
   Empresa: Acquisition.com

⭐ PRÓXIMA ETAPA
   Para processar: /process-jarvis "inbox/ALEX HORMOZI/MASTERCLASSES/How to Scale from $1M to $10M ARR [youtube.com_watch_v=abc123].txt"
   Ou: /inbox para ver todos pendentes

═══════════════════════════════════════════════════════════════════════════════

​
Step 6: Auto-Process (if —process flag)

If --process flag is present, automatically triggers: 
/process-jarvis "inbox/ALEX HORMOZI/MASTERCLASSES/filename.txt"

​
Audit Logging

All ingestions are logged to /logs/AUDIT/audit.jsonl: 
{
  "timestamp": "2026-03-06T10:45:23Z",
  "operation": "INGEST",
  "source": "https://youtube.com/watch?v=abc123",
  "destination": "inbox/ALEX HORMOZI/MASTERCLASSES/How to Scale from $1M to $10M ARR [youtube.com_watch_v=abc123].txt",
  "source_id": "AH024",
  "word_count": 8542,
  "status": "SUCCESS"
}

​
YouTube Transcription

​
Requirements

OPENAI_API_KEY required for YouTube transcription via Whisper
YouTube videos are transcribed using OpenAI Whisper:
  1. Download audio using yt-dlp
  2. Convert to MP3
  3. Send to Whisper API
  4. Save transcript with timestamps

​
Transcript Format

[00:00:00] Welcome to today's masterclass on scaling...
[00:00:15] The first thing you need to understand is...
[00:00:45] Let me give you a specific example...

​
Supported Languages

Whisper auto-detects language. Supports 50+ languages including:
  • English
  • Spanish
  • Portuguese
  • French
  • German
  • Chinese
  • Japanese
  • And more…

​
Google Drive Import

​
Setup Required

GOOGLE_CLIENT_ID and GOOGLE_CLIENT_SECRET required for Drive import
Configure during /setup or manually in .env: 
GOOGLE_CLIENT_ID=your-client-id.apps.googleusercontent.com
GOOGLE_CLIENT_SECRET=your-client-secret

​
OAuth Flow

First time accessing Drive: 
🔑 Google Drive Authorization Required

Opening browser for authentication...

1. Sign in to your Google account
2. Grant access to Mega Brain
3. Return to this window

Waiting for authorization...

​
Supported Document Types

  • Google Docs - Converted to plain text
  • Google Sheets - Converted to CSV then text
  • PDFs - Extracted with OCR if needed
  • Word Docs (.docx) - Converted to text

​
Error Handling

​
YouTube Video Not Available

✗ INGEST FAILED

YouTube video not available:
  - Video is private
  - Video was deleted
  - Invalid video ID

Please check the URL and try again.

​
Transcription Failed

✗ TRANSCRIPTION FAILED

OpenAI Whisper error: Quota exceeded

Options:
  1. Wait and try again (quota resets hourly)
  2. Upload transcript manually to inbox/
  3. Use pre-existing transcript file

​
File Already Exists

⚠️  FILE EXISTS

A file with the same name already exists:
  inbox/ALEX HORMOZI/MASTERCLASSES/filename.txt

Options:
  [o] Overwrite existing file
  [r] Rename new file (append timestamp)
  [s] Skip ingestion

Choice:

​
Unsupported File Type

✗ INGEST FAILED

Unsupported file type: .pptx

Supported types:
  - Text: .txt, .md
  - Documents: .pdf, .docx
  - Web: YouTube URLs, Google Drive links

Convert the file to a supported format and try again.

​
Performance

​
Processing Times

Source TypeSizeTime
YouTube (1hr)~1 video2-5 min
PDF (50 pages)~5 MB10-30 sec
Local .txt~10k words1-2 sec
Google Doc~10k words5-10 sec
YouTube transcription time depends on video length and Whisper API latency

​
Rate Limits

  • OpenAI Whisper: 50 requests/min (Tier 1)
  • YouTube API: 10,000 quota/day
  • Google Drive API: 1,000 requests/100s/user

​
Best Practices

​
1. Organize by Expert

Use --person flag for consistency: 
/ingest video1.mp4 --person "Alex Hormozi"
/ingest video2.mp4 --person "Alex Hormozi"
/ingest video3.mp4 --person "Alex Hormozi"

​
2. Specify Content Type

Helps with later analysis: 
/ingest course-module-1.pdf --type COURSE
/ingest webinar-replay.mp4 --type MASTERCLASS
/ingest sales-script.txt --type SCRIPT

​
3. Batch Processing

For multiple files, ingest first, then batch process: 
# Ingest without processing
/ingest video1.mp4
/ingest video2.mp4
/ingest video3.mp4

# Then batch process all
/process-jarvis --batch inbox/PERSON/TYPE/

​
4. Use Descriptive Filenames

Before ingesting local files: 
# Good
/ingest "Alex-Hormozi-Pricing-Masterclass-2024.txt"

# Bad
/ingest "transcript_final_v2.txt"

​
Advanced Usage

​
Playlist Ingestion

​
Webhook Integration

​
Custom Metadata

​
Troubleshooting

​
”yt-dlp not found”

Issue: YouTube downloader not installed Solution: 
# macOS
brew install yt-dlp

# Linux
sudo apt install yt-dlp

# Windows
winget install yt-dlp

​
“OpenAI API key invalid”

Issue: API key not configured or incorrect Solution: 
# Re-run setup
/setup

# Or manually edit .env
OPENAI_API_KEY=sk-your-key-here

​
“Permission denied writing to inbox/”

Issue: Insufficient file permissions Solution: 
# Fix inbox permissions
chmod -R 755 inbox/

​
Next Steps

Process Command

Transform raw content into structured knowledge

JARVIS Briefing

Monitor your knowledge base growth

Pipeline Guide

Understand the 8-phase processing pipeline

Inbox Management

Best practices for organizing materials

Build docs developers (and LLMs) love

Get started for freeTalk to us