WeClone
A one-stop tool for creating a personal AI chat persona from exported chat history, fine-tuning a language model on the logs to match an individual's writing style and connecting the result to a chatbot.
WeClone fine-tunes a language model on a person's own chat history to reproduce their writing style, then connects the tuned model to a chatbot, creating a personal digital avatar. It covers the full pipeline: exporting chat data, preprocessing it, training the model, and deploying the result.
It supports fine-tuning with image data alongside text for more authentic results, includes privacy filtering so sensitive information can be removed before training, and runs training and deployment locally for data control. Chat history can be imported from Telegram, with WhatsApp support planned, and the resulting chatbot can be deployed to Telegram, Discord, Slack or personal WeChat.
By default it fine-tunes the Qwen2.5-VL-7B-Instruct model with LoRA. The project notes it is still in rapid iteration, and that larger base models and more training data tend to produce better results. It is released under the AGPL-3.0 license.