Skip to content

Instantly share code, notes, and snippets.

@jmanhype

jmanhype/delightful-tickling-crab.md

Created January 16, 2026 20:14
Show Gist options

  • Select an option

    Learn more about clone URLs
  • Save jmanhype/01ab1457449098672c611fa4b16e64b5 to your computer and use it in GitHub Desktop.

Select an option

Learn more about clone URLs
Save jmanhype/01ab1457449098672c611fa4b16e64b5 to your computer and use it in GitHub Desktop.
Download ZIP
Content Creation Pipeline - Letta MAS to Social Media (Skool + FB Groups)
Raw
delightful-tickling-crab.md

Plan: Content Creation Pipeline - Letta MAS to Social Media

Goal

Monetize your AI video knowledge by building a content distribution system that:

  1. Transforms Letta MAS videos into platform-ready formats
  2. Auto-generates captions with Ollama
  3. Schedules posts via Postiz (self-hosted)
  4. Distributes across Twitter/X, LinkedIn, TikTok, YouTube

Current State

What you have:

  • 100+ AI-generated videos in /Users/speed/proxmox/videos/
  • Letta MAS running (6 agents, port 8283)
  • ComfyUI LTX-2 (port 8188)
  • Frame server (port 8189)
  • PostgreSQL + Redis already running
  • Ollama for local inference

What's missing:

  • Postiz (not deployed)
  • Video format conversion (vertical, square)
  • Social media OAuth setup
  • Content templates
  • Automated caption generation

Architecture

LETTA MAS → VIDEO PROCESSOR → POSTIZ → SOCIAL PLATFORMS
(generate)    (transform)     (schedule)   (distribute)

┌──────────┐   ┌──────────┐   ┌──────────┐   ┌─────────┐
│ ComfyUI  │──>│ FFmpeg   │──>│ Postiz   │──>│ X/LI/TT │
│ 768x512  │   │ 9:16,1:1 │   │ Drafts   │   │ YT/IG   │
└──────────┘   └──────────┘   └──────────┘   └─────────┘

Phase 1: Deploy Postiz

Location: /home/straughter/content-pipeline/

1.1 Create Databases

# On Proxmox - create postiz + temporal databases
docker exec -it postgres-vector psql -U vectorgraph -c "
CREATE DATABASE postiz;
CREATE USER postiz WITH PASSWORD 'postiz_secret';
GRANT ALL PRIVILEGES ON DATABASE postiz TO postiz;
CREATE DATABASE temporal;
CREATE USER temporal WITH PASSWORD 'temporal_secret';
GRANT ALL PRIVILEGES ON DATABASE temporal TO temporal;
"

1.2 Docker Compose

File: /home/straughter/content-pipeline/docker-compose.postiz.yml

Key services:

  • postiz - Main app on port 4007
  • postiz-redis - Dedicated Redis
  • temporal - Workflow engine on port 7233
  • temporal-ui - Dashboard on port 8088

1.3 Environment Variables

File: /home/straughter/content-pipeline/.env

Required:

  • POSTIZ_JWT_SECRET - Random 32+ char string
  • DATABASE_URL - PostgreSQL connection
  • REDIS_URL - Redis connection
  • Social OAuth keys (Phase 4)

Phase 2: Video Processing Pipeline

Purpose: Convert 768x512 landscape videos to platform formats

2.1 Platform Specs

Platform Dimensions Aspect Method
TikTok/Reels 1080x1920 9:16 Center crop
Instagram Square 1080x1080 1:1 Center crop
Twitter/LinkedIn 1920x1080 16:9 Letterbox
YouTube Shorts 1080x1920 9:16 Center crop

2.2 Thumbnail Extraction

Size Dimensions Use
og 1200x630 Open Graph/Link previews
twitter 1200x675 Twitter Cards
youtube 1280x720 YouTube thumbnails
square 1080x1080 Instagram

2.3 Files to Create

/home/straughter/content-pipeline/video_processor.py

  • Watch for new videos in ComfyUI output
  • Extract thumbnails at multiple sizes
  • Convert to all platform formats
  • Output to /content-pipeline/processed/{video_id}/

/home/straughter/content-pipeline/text_overlay.py

  • Add title text (first 2 seconds)
  • Add CTA text (last 2 seconds)
  • Branded watermark (optional)

Phase 3: Letta-Postiz Bridge

Purpose: Auto-create Postiz drafts when videos complete

3.1 Bridge Service

File: /home/straughter/content-pipeline/letta_postiz_bridge.py

Flow:

  1. Detect new video in ComfyUI output
  2. Query Letta archival memory for prompt/metadata
  3. Run video processor (thumbnails + formats)
  4. Generate captions via Ollama (platform-specific)
  5. Create draft posts in Postiz

3.2 Caption Generation

Use Ollama qwen2.5:7b to generate:

  • Hook - Attention grabber
  • Description - Brief, engaging
  • CTA - Call to action
  • Hashtags - Platform-appropriate

3.3 Systemd Service

# Auto-start on boot
sudo systemctl enable content-pipeline

Phase 4: OAuth Setup

Required Developer Accounts

Platform Portal Callback URL
Twitter/X developer.twitter.com http://192.168.1.143:4007/api/callback/twitter
LinkedIn linkedin.com/developers http://192.168.1.143:4007/api/callback/linkedin
TikTok developers.tiktok.com http://192.168.1.143:4007/api/callback/tiktok
YouTube console.cloud.google.com http://192.168.1.143:4007/api/callback/youtube

Notes:

  • LinkedIn requires "Marketing Developer Platform" access
  • TikTok requires app review for production
  • YouTube needs "YouTube Data API v3" enabled

Phase 5: Content Templates

Twitter/X (280 chars)

{hook}

{short_description}

{cta}

#AIVideo #AIArt #GenerativeAI #LTX2

LinkedIn (3000 chars)

{professional_hook}

{context_paragraph}

{learning_point}

{cta}

#AIVideo #GenerativeAI #FutureOfWork

TikTok (150 chars)

{punchy_hook} #aiart #aivideo #fyp #viral #darkfantasy

Directory Structure

/home/straughter/content-pipeline/
├── docker-compose.postiz.yml
├── .env
├── video_processor.py
├── letta_postiz_bridge.py
├── text_overlay.py
├── health_check.py
├── config/
│   └── templates.json
├── processed/
│   └── LTX-2_XXXXX_/
│       ├── manifest.json
│       ├── thumb_og.jpg
│       ├── tiktok.mp4
│       ├── twitter.mp4
│       └── ...
└── logs/

Implementation Order

1. Phase 1: Deploy Postiz          (~30 min)
   └── Database setup, Docker compose, verify UI

2. Phase 2: Video Processor        (~45 min)
   └── FFmpeg scripts, thumbnail extraction

3. Phase 3: Bridge Service         (~30 min)
   └── Letta integration, caption generation

4. Phase 4: OAuth Setup            (~60 min)
   └── Developer accounts, API keys, test auth

5. Phase 5: Content Templates      (~15 min)
   └── Configure templates, test generation

Verification Checklist

  • Postiz UI accessible at http://192.168.1.143:4007
  • Temporal UI at http://192.168.1.143:8088
  • Video processor creates all format variants
  • Thumbnails extracted correctly
  • Ollama generates captions
  • Bridge service creates Postiz drafts
  • OAuth works for at least 1 platform
  • Can schedule and publish test post

Critical Files

File Purpose
docker-compose.postiz.yml Postiz deployment
video_processor.py Format conversion
letta_postiz_bridge.py Integration service
.env OAuth credentials

Monetization Path: Skool + Facebook Groups

FB Groups → Skool Community → Course Upsells

Week 1-2: Join AI/automation FB groups, post value (videos + insights)
Week 3-4: Launch Skool community ($29-49/mo)
Month 2:  Add course module to Skool ($97-297)
Month 3+: Premium tier ($497), 1:1 consulting

Skool Strategy

  • Community: $29-49/mo - Access to tutorials, prompt library, support
  • Course: Built into Skool - "Build Your AI Video Factory"
  • Upsell: Done-for-you setup, consulting

FB Groups to Target

  • AI Art & Video Generation groups
  • Automation/No-Code groups
  • Content Creator communities
  • Side hustle/passive income groups

Content Flow

Letta MAS videos → Postiz → FB Groups (free value)
                         ↓
              "Want the full system?"
                         ↓
                Skool community link

The videos prove you can do it. The FB posts give free value. Skool captures the paying audience.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment