This project, titanwings/ex-skill, addresses the desire to digitally recreate the persona of a past intimate partner using their chat history. It targets users seeking to interact with a simulated version of an ex-partner, offering a way to "preserve" a lost connection by replicating their communication style and personality traits. The primary benefit is providing a tool for users to engage with a digital representation of someone they miss, offering a unique form of digital remembrance or closure.

How It Works

The project generates a "Skill" representing a digital persona, built upon two core components: "Relationship Memory" and a 6-layer "Persona" structure. Relationship Memory captures shared experiences, inside jokes, and conflict patterns. The Persona comprises layers detailing core behavioral modes, identity (MBTI, attachment style, etc.), expression style (speech patterns, emojis), emotional behavior, conflict resolution, and sensitive triggers. The system processes incoming messages by referencing these components to generate responses that mimic the target individual's communication style and emotional responses. This layered approach aims for high fidelity and nuanced simulation.

Quick Start & Requirements

Installation is recommended via OpenClaw, cloning the repository into ~/.openclaw/workspace/skills/create-ex. Alternatively, it can be installed within a Claude Code environment using git clone ... .claude/skills/create-ex or ~/.claude/skills/create-ex. Optional Python dependencies can be installed with pip3 install -r requirements.txt. Automated data collection requires the desktop WeChat client logged in on Windows or macOS, or iMessage logged in on macOS. Python 3.9+ is supported. Detailed installation instructions are available in INSTALL.md.

Highlighted Details

• 6-Layer Persona Model: Detailed breakdown including Core Mode, Identity, Expression Style, Emotional Behavior, Conflict/Boundaries, and Triggers/Dealbreakers.
• Evolutionary Mechanism: Supports incremental updates via new chat logs, conversational corrections ("TA 不会这样"), and version management with rollback capabilities.
• Automated Data Import: Handles WeChat (Windows/macOS) and iMessage (macOS) chat logs automatically, with manual options for images and direct text input.
• Multi-Persona Support: Allows for the creation and management of an unlimited number of distinct "ex" Skills simultaneously.

Maintenance & Community

The project is currently in a beta testing phase, with developers encouraging users to report bugs and provide feedback via GitHub issues. There are no specific mentions of notable contributors, sponsorships, or partnerships in the provided README.

Licensing & Compatibility

The project is released under the MIT License. This permissive license generally allows for commercial use, modification, and distribution, including integration into closed-source projects, with minimal restrictions beyond attribution.

Limitations & Caveats

As a beta version, the project may contain bugs or incomplete features. The quality of the generated "Skill" is highly dependent on the quality and quantity of the provided chat logs; real conversation data combined with subjective descriptions is recommended over text-only descriptions. Automated decryption for WeChat logs may encounter issues, with manual export tools suggested as a fallback.