Installation Guide

Install the tool locally to analyze your own WeChat chat history. Runs on macOS and Linux.

Quick Start

1. Prerequisites

Python 3.12 or later

python3 --version

WeChat desktop app (for decryption)

2. Install the chatlog extractor

Download and set up the chatlog tool (MIT licensed)

cd original
bash scripts/setup_chatlog.sh

3. Install dependencies

Install the distiller with optional dependencies

cd original
pip install -e ".[full,dev]"

4. Try the demo (no WeChat required)

Generate synthetic data and run the full pipeline

cd original
make demo

Output lands in output/samples/

ls output/samples/
# profile.md  persona_prompt.txt  stats.json

Using Your Real Chats

1. Extract your chats

With WeChat desktop running and logged in, run:

chatlog

Follow the TUI to decrypt your database

2. Run the pipeline

Analyze your messages and generate profile + prompt

python -m wechat_style_distiller.cli run \
  --from-api \
  --out output \
  --name "Your Name"

Check the outputs

ls output/
# profile.md  persona_prompt.txt  dataset.jsonl  stats.json

3. Chat in your voice (optional)

Requires ANTHROPIC_API_KEY in environment

export ANTHROPIC_API_KEY="sk-ant-..."

Start the chatbot

python -m wechat_style_distiller.cli chat \
  --persona output/persona_prompt.txt

Privacy Reminder

All analysis runs locally. Your chat data is read from your machine, written to git-ignored folders, and never uploaded. Network calls only happen if you explicitly enable the chatbot feature with your own API key.

See the Privacy page for full details.

Full documentation and source code available on GitHub:

rNLKJA/wechat-style-distiller