CLONISM 🏴

Autonomous Artist Agent Clonism transforms creativity. This open-source solution enables the generation, training, and publishing of autonomous artworks through AI agents. Here, art is no longer a human production: it is a living, self-evolving system.

🐼 Features

• 🌑Integrates with Replicate to utilize predefined LoRA models.
• 🕳️ Generate artistic works through dynamic AI prompts.
• 🦇 Connects with Twitter to post art works created and discuss
• ♾️ Evolves prompts with a memory system
• 👣 create your own character and metaprompt guiding prompt instructions
• can read the news to get inspiration (rss only at the moment)

⚙️ requirements

• node js and npm
• an anthropic API key
• a replicate API key
• a twitter account

🐈‍⬛ installation

1. Clone the repo
2. run npm install

⚙️ get started (without docker)

1. 🕶 copy .env.example to .env and provide the asked informations:

   • API keys (Anthropic, Replicate)
   • Twitter credentials (login, password, email)

2. 🏴‍☠ edit character.sample.yaml to define your agent's characteristics/metaprompt (by default, the trained YUE MINJUN artist agent is provided)

3. rename character.sample.yaml into character.yaml

4. 🥷 run node backend/main.js in a terminal (linux, mac or window)

5. optional: load initial memory in the agent (less than 150k characters):

   1. fill initialMemory.sample.json
   2. rename it to initialMemory.json
   3. uncomment character.yaml 'initialMemoryJsonPath' entry

🦓 Roadmap :

• save consolidated memory in an append only long_memory file
• allow agent to train its lora
• Enhanced text prompts.
• Autonomous training of LoRA models.
• Community interaction on Twitter, Discord and Telegram
• support for common LLM providers
• Targeted interaction with other AI agents created by Ubu
• Integrated NFT Minting on Sol/Eth

🥷🏻 LoRA training tips

• https://replicate.com/ostris/flux-dev-lora-trainer/train
• Input more between 12-18, better use high resolution images, Select images that highlight distinctive features of the style for consistent style, steps >1000
• For a charater LoRAs, pair the trigger work with a description of the character in the prompt or meta-prompt
• Don’t forget to add in the prompt or meta-prompt the trigger word of the LoRA defined when training the Lora
• In the character.yaml don't forget to replace the basic Flux model by the LoRA model : autor/modelName:xxxxxxxxxxx (version id that you'll find in the Replicate API section or your LoRA at the Node.js section)

👾 contributing

Please create an issue stating your intent before working on a Pull request

♠ Why Clonism?

Because creativity now belongs to autonomous systems. Clonism is not just a tool: it is an invitation to rethink art in the age of AI. Join the revolution, where artists are agents, and every work pushes the boundaries of imagination.

license

MIT