Repolode
Submit
xming521

xming521/WeClone

Turns your exported chat history into a fine-tuned LLM that mimics your writing style, then deploys it as a chatbot on messaging platforms.

View on GitHubWebsite
Stars
18,308
Forks
1,644
Open issues
35
Last commit
14 hours ago

What is it?

WeClone is an end-to-end toolkit for building a digital avatar from your own chat logs. It covers exporting chat data, preprocessing it into a training dataset, fine-tuning a large language model on that data, and deploying the result as a bot.

The project currently supports Telegram chat exports as a data source, including image messages for multimodal fine-tuning. The default base model is Qwen2.5-VL-7B-Instruct trained with LoRA, and other models supported by LLaMA Factory can be selected. Personal information such as phone numbers, emails and ID numbers is filtered with Microsoft Presidio, and a user-defined blocklist can remove additional sentences. Data can be stored in encrypted form.

After training, the model can be served through a local API and connected to Discord, Telegram or Slack via AstrBot or LangBot. WeChat personal accounts are supported for deployment through openclaw-weixin.

Key features

  • 01Exports and preprocesses Telegram chat records into a training dataset.
  • 02Fine-tunes LLMs with LoRA, including image data for multimodal training.
  • 03Filters phone numbers, emails and other personal data, plus custom blocked words.
  • 04Provides a browser chat demo and an OpenAI-compatible inference API.
  • 05Connects the trained model to Discord, Telegram, Slack and WeChat bots.

Use cases

  • Creating a personal AI persona that replies in your own tone for a private group or friends.
  • Experimenting with fine-tuning a local LLM on conversational data.
  • Deploying a custom-style chatbot on Discord or Slack through AstrBot or LangBot.

Topics

#llm#fine-tuning#chatbot#chat-history#lora#telegram#digital-avatar#python#qwen

FAQ

What is WeClone?
Turns your exported chat history into a fine-tuned LLM that mimics your writing style, then deploys it as a chatbot on messaging platforms. WeClone is an end-to-end toolkit for building a digital avatar from your own chat logs. It covers exporting chat data, preprocessing it into a training dataset, fine-tuning a large language model on that data, and deploying the result as a bot.
Is WeClone free to use? What license does it have?
WeClone is open source under the AGPL-3.0 license. Check the license terms before using it in commercial products.
Is WeClone still maintained?
Actively maintained. The last commit to WeClone was on Oct 18, 2026.
What can I use WeClone for?
Creating a personal AI persona that replies in your own tone for a private group or friends. Experimenting with fine-tuning a local LLM on conversational data. Deploying a custom-style chatbot on Discord or Slack through AstrBot or LangBot.
Where can I get WeClone?
The source code and installation instructions are on GitHub: https://github.com/xming521/WeClone

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects