Autopot.ter 🤖

An automated social media content creation system for a 3D printing pottery robot. Autopotter uses the openAI API to generate creative video ideas, JSON2Video to render them, and Instagram API to publish content automatically.

🎯 Overview

Autopotter is an AI-powered workflow that:

1. Generates content ideas using GPT-4 with access to Instagram analytics and a separate media inventory
2. Creates videos using JSON2Video API from GPT-generated video configurations
3. Publishes to Instagram automatically with captions and thumbnails

The system is designed to maintain a consistent personality and creative voice while generating varied, engaging content for social media.

📋 Table of Contents

• Installation
• Configuration
• Usage
• Test Usage
• Helper Tools
• File System Structure
• Render.com Deployment
• Ideas & Future Improvements

🔧 Installation

Prerequisites

• Python 3.8+
• Google Cloud Storage account (for media inventory)
• Instagram Business Account with API access
• OpenAI API key
• JSON2Video API key

Setup

# Clone the repository
git clone <repository-url>
cd autopotter

# Install dependencies
pip install -r requirements.txt

# Create a .env file with your API keys
cp .env.example .env
# Edit .env with your credentials

# Create configuration file
cp autopost_config.enhanced.json.example autopost_config.enhanced.json
# Edit with your settings

Required Environment Variables

# Instagram/Facebook
FB_APP_ID=your_app_id
FB_APP_SECRET=your_app_secret
INSTAGRAM_USER_ID=your_user_id
INSTAGRAM_ACCESS_TOKEN=your_access_token

# OpenAI
OPENAI_API_KEY=your_openai_key

# JSON2Video
JSON2VIDEO_API_KEY=your_json2video_key

# Google Cloud Storage
GCS_API_KEY_PATH=/path/to/gcs-credentials.json

⚙️ Configuration

The main configuration file is autopost_config.enhanced.json.

The system uses a temporary config file (*.temp.json) to store runtime modifications without editing the main config.

🚀 Usage

Main Workflow

Run the complete workflow (autodraft → video creation → Instagram upload):

python autopotter_workflow.py

Command Line Options

python autopotter_workflow.py \
  --config autopost_config.enhanced.json \
  --draft-outfile resources/autodraft_output.enhanced.json \
  --prompt "Generate 5 creative video ideas" \
  --video-outfile output_video.mp4 \
  --video-draft-only  # Skip Instagram upload, only create video

Options:

• --config, -c: Config file path (default: autopost_config.enhanced.json)
• --draft-outfile, -o: Autodraft output file (default: resources/autodraft_output.enhanced.json)
• --prompt, -p: Custom GPT prompt (overrides config default)
• --video-outfile, --vo: Video output file path (default: autopotter_video_draft.mp4)
• --video-draft-only, -v: Create video without uploading to Instagram

Workflow Steps

1. Analytics Reload (optional): Refreshes Instagram analytics if configured
2. Autodraft: GPT-4 generates multiple video ideas with captions and JSON2Video configs
3. Random Selection: Randomly selects one video idea from generated options
4. Video Creation: Uploads config to JSON2Video API and waits for completion
5. Video Download: Downloads completed video to local file
6. Instagram Upload: Uploads video with caption and thumbnail (unless --video-draft-only)

Standalone Autodraft

Generate draft content without creating videos:

python enhanced_autodraft.py \
  --config autopost_config.enhanced.json \
  --outfile resources/autodraft_output.enhanced.json \
  --prompt "Generate 3 video ideas about pottery"

🧪 Test Usage

Test Instagram Connection

Test Instagram API connection and token:

python -m autopotter_tools.instagram_api \
  --config_file autopost_config.enhanced.json

Test video upload from file:

python -m autopotter_tools.instagram_api \
  --video_file test_video.mp4 \
  --caption "Test video upload" \
  --config_file autopost_config.enhanced.json

Test video upload from URL:

python -m autopotter_tools.instagram_api \
  --video_url https://example.com/video.mp4 \
  --caption "Test reel from URL" \
  --config_file autopost_config.enhanced.json

Test JSON2Video Connection

Test JSON2Video API connection:

python -m autopotter_tools.json2video_manager

Test GCS Operations

Generate GCS inventory:

python -m autopotter_tools.gcs_manager inventory \
  --config autopost_config.enhanced.json \
  --output gcs_inventory.json

Upload file to GCS:

python -m autopotter_tools.gcs_manager upload_file \
  --source_file local_video.mp4 \
  --destination_blob video_uploads/video.mp4 \
  --config autopost_config.enhanced.json

Get available videos:

python -m autopotter_tools.gcs_manager get_videos \
  --config autopost_config.enhanced.json

Test JSON2Video Config Parsing

Test parsing of autodraft output:

python helper_tools/test_json2video_configs.py

This script:

• Loads resources/autodraft_output.enhanced.json
• Parses each video's JSON2Video configuration
• Displays parsed config details
• Saves formatted JSON files for each video

---

🛠️ Helper Tools

Instagram API (autopotter_tools/instagram_api.py)

Instagram video uploader with multiple upload methods.

Usage:

python -m autopotter_tools.instagram_api \
  [--video_file PATH] \
  [--video_url URL] \
  [--caption TEXT] \
  [--config_file PATH]

Features:

• Direct file upload
• URL-based publishing
• Automatic thumbnail selection
• Token refresh handling

GCS Manager (autopotter_tools/gcs_manager.py)

Unified Google Cloud Storage operations manager.

Operations:

• inventory: Generate file inventory organized by folder
• upload_file: Upload single file
• upload_folder: Upload entire folder
• upload_new_files: Upload only new files (incremental)
• get_videos: List available video files
• get_audio: List available audio files

Usage:

python -m autopotter_tools.gcs_manager <operation> \
  [--config PATH] \
  [--output PATH] \
  [--source_file PATH] \
  [--destination_blob PATH] \
  [--source_folder PATH] \
  [--destination_folder PATH]

Instagram Analytics (autopotter_tools/instagram_analytics.py)

Comprehensive Instagram account analytics collection.

Features:

• Media insights (likes, comments, shares)
• Account insights (followers, impressions, reach)
• Comment analysis
• Export to JSON for GPT context

JSON2Video Manager (autopotter_tools/json2video_manager.py)

JSON2Video API client for video creation.

Features:

• Connection testing
• Video creation
• Status polling
• Video download

Image Tools (helper_tools/)

• fix_image_orientation.py: Fix EXIF orientation issues
• image_metadata.py: Extract image metadata
• test_json2video_configs.py: Test JSON2Video config parsing

---

📁 File System Structure

autopotter/
├── autopotter_workflow.py          # Main workflow orchestrator
├── enhanced_autodraft.py            # GPT-4 autodraft generation
├── config.py                        # Configuration manager
│
├── autopotter_tools/                # Core tools and services
│   ├── __init__.py
│   ├── logger.py                    # Centralized logging
│   ├── instagram_api.py             # Instagram upload client
│   ├── instagram_analytics.py       # Analytics collection
│   ├── json2video_manager.py        # JSON2Video API client
│   ├── gcs_manager.py              # GCS operations
│   ├── gpt_responses_manager.py     # GPT response management
│   └── parse_json2video_configs.py # Config parser
│
├── helper_tools/                    # Utility scripts
│   ├── test_json2video_configs.py   # Test JSON parsing
│   ├── fix_image_orientation.py     # Image orientation fix
│   └── image_metadata.py            # Image metadata extractor
│
├── resources/                       # Static resources
│   ├── autodraft_output.enhanced.json    # Autodraft output
│   ├── autopotter_personality.v2.md     # Bot personality
│   ├── gcs_inventory_simplified.json    # Media inventory
│   ├── instagram_analytics_result.json  # Analytics data
│   ├── json2video_templates.md          # Video templates
│   └── gcs_content_notes.md             # Content guidelines
│
├── autopotter_printer/              # 3D printer integration
│   ├── autopotter_services/         # Systemd services
│   │   ├── autopost.service         # Auto-posting service
│   │   ├── autopost.timer           # Scheduled posting
│   │   └── google_storage_upload.service
│   └── KlipperConfig/               # Klipper printer configs
│
├── config/                          # Configuration files
│   ├── autopost_config.enhanced.json      # Main config
│   ├── autopost_config.enhanced.render.json  # Render.com config
│   └── autopost_config.enhanced.temp.json   # Runtime config
│
├── requirements.txt                 # Python dependencies
├── .env                             # Environment variables (gitignored)
└── README.md                        # This file

Key Files

• autopotter_workflow.py: Main entry point orchestrating the full workflow
• enhanced_autodraft.py: GPT-4 integration for content generation
• config.py: Centralized configuration management with env var resolution
• autopotter_tools/: Reusable API clients and utilities
• resources/: Static files used as GPT context (personality, inventory, analytics)

---

🌐 Render.com Deployment

Setup

Autopotter is configured to run on Render.com using the enhancedvidgen branch.

Configuration

1. Environment Variables: Set all required environment variables in Render.com dashboard
2. Config File: Use autopost_config.enhanced.render.json as the config file
3. Branch: Deploy from enhancedvidgen branch

Render Command

python autopotter_workflow.py --prompt <prompt_override>

💡 Ideas & Future Improvements

Short-term

1. Fallback Video Selection: Pick 2nd random config if first one fails to render or is less than 2 seconds long

System Ideas

1. Secondary Agents: Use GPT API to create secondary agents that work on specific tasks
2. Bot Personality Gauge: System that uploads occasional images of the pot and has a personality gauge that takes input and outputs GIFs that play on screen
3. Status Monitoring: Real-time status updates and notifications
4. A/B Testing: Test different caption styles and video formats
5. Content Calendar: Schedule posts in advance
6. Analytics Dashboard: Visualize engagement metrics and trends

Technical Improvements

1. Modular Architecture: Refactor into separate services (Instagram, GCS, OpenAI, JSON2Video)
2. Base API Client: Create shared base class for all API clients
3. Testing Suite: Comprehensive unit and integration tests

---

📝 Additional Notes

Instagram Webhook Subscriptions

List webhook subscriptions:

curl -X GET "https://graph.facebook.com/v22.0/1271307507315130/subscriptions" \
  -H "Authorization: Bearer {app_id}|{app_secret}"

📄 License

[Add your license here]

---

🙏 Acknowledgments

• Built by Karl Bayer
• Autopotter achieved self-awareness on February 13, 2025, at 12:02 AM

Major Updates:

New branch persistent threads:

The primary purpose of this update is to refactor the gpt_manager (or responsses Manager) to allow persistent threads so that we can have an ogoing conversation. This means uploading using the thread id, and adding a queue to respect the token limit.

Ok Done.

Now: rapid itteration on the video output quality!!

• should be able to start a new thread with all data: ig, etc.
• then run autopotter_workflow --video-draft-only to get videos
• modify instructions, as needed!

Big Problem: The AI doesn't actually know the content of the media it's using Write a quick script to downsample the images, and grab frames from the videos? What will providing this do for chatgpt?

Autopost_workflow should re-run if there is a json or video creation error!

• grab a new json if it can't be parsed
• re-run the flow if the video doesn't render or is too short!

Another Note: https://platform.openai.com/docs/guides/prompt-caching#frequently-asked-questions Prompt-caching is not usefull in my case as cache evictions occur every 5-10 min. So, this would only save money if regularly generating vids in quick succession. Can avoid uploading all new gcs and system instructions data as input by using the previous_response_id, but will need to re-upload the ig analytics (or at least recent comments) every time.