Summary

OpenBiliClaw is a local, open-source AI agent that transcends platform-specific recommendation silos. It builds a deep psychological profile of the user to proactively discover content across Bilibili, Xiaohongshu, Douyin, and YouTube, offering a truly personalized discovery experience by understanding nuanced interests beyond click history.

How It Works

Utilizes a "five-layer soul profile" and proactive "interest probing" to infer user needs and explore novel content domains, contrasting with platform-driven recommenders. Psychological principles guide active exploration beyond current user footprints. All data and LLM interactions are processed locally, ensuring complete privacy and user control.

Quick Start & Requirements

Installation requires a browser plugin (Chrome/Firefox compatible) and a local backend agent, deployable via script, AI assistant, or Docker. Python 3.11+ is mandatory. Users must log into target platforms (Bilibili, Xiaohongshu, Douyin, YouTube) in the same browser. LLM API keys are essential; local embedding via Ollama (bge-m3) is optional. Key links include the project homepage (https://whiteguo233.github.io/OpenBiliClaw/) and installation guides (docs/agent-install.md).

Highlighted Details

• Deep Psychological Profiling: Infers MBTI, cognitive styles, and deep needs.
• Proactive Interest Discovery: Actively hypothesizes and validates new interest areas.
• Cross-Platform Integration: Unifies interests across disparate platforms.
• 100% Local & Private: All data and LLM interactions remain on-device.
• Human-like Explanations: Provides clear, friend-like reasoning for recommendations.
• Local Embedding Option: Supports Ollama for similarity calculations without external API keys.

Maintenance & Community

Recent activity (v0.3.77, May 2026) indicates active development. A discussion link (https://linux.do/t/topic/1978894) is provided, though dedicated community channels are not prominent. The MIT license encourages contributions.

Licensing & Compatibility

Licensed under the permissive MIT license, suitable for commercial use and integration. Its local-first, privacy-centric design enhances compatibility.

Limitations & Caveats

Setup demands technical proficiency. Functionality depends on target platform API stability. Firefox plugin support is currently limited to temporary loading. Core features require user-provided LLM API keys, potentially incurring costs.