Yappy is a silly side project. LinkedIn people yap a lot right? So what if you could yap back, automatically, and have it sound like you?
It's not production software. It's more of a "what happens if I just let an LLM loose on LinkedIn comments" experiment. But it works surprisingly well, which is either impressive or alarming depending on how you feel about AI-generated discourse.
What It Does
Yappy finds LinkedIn posts based on keywords you care about, reads them, and generates a contextual comment in your voice. You can review before it posts or just let it run and see what happens.
The loop is pretty simple:
- Find posts matching your keywords
- Read the post content
- Generate a comment using your persona config
- Post it (or don't, if you want to stay sane)
The Persona Thing
The interesting part is the persona system. You give it writing samples, preferred phrases, topics to avoid, and it tries to match your style. Generic AI comments are obviously AI. This makes it a little less obvious. It's not perfect, but it's a good fine-tuning target if you wanted to take it further.
Could It Be Useful?
Probably, with some work. The core loop is solid and the persona approach is the right direction. A production version would need proper rate limiting, smarter post filtering, and maybe a feedback loop to improve comment quality over time. For now it's mostly just fun to watch it yap.
Built in Python with Playwright. LLM layer is pluggable (OpenAI, Anthropic, Ollama). MIT licensed.