All AI systems
Case study · Gwee Per Ming
BersamaAi Community Agents
A deployed Discord community bot and automated content pipeline that filters online sources into topic channels and summarizes videos without captions.
The problem
A community needs a steady flow of relevant material without relying on someone to manually review every source or transcribe every video.
My contribution
Built a source-ingestion workflow using an LLM-as-a-Judge for relevance and topic classification, deployed the Discord bot, and added speech recognition for caption-less videos.
Architecture and stack
Python, GCP Compute Engine, a Discord MCP surface, GLM-5.2, Groq Whisper ASR, yt-dlp, and GitHub Actions.
- Python
- GCP Compute Engine
- Discord MCP
- GLM-5.2
- Groq Whisper ASR
- GitHub Actions
What runs today
The Discord bot runs on a GCP VM. The topic-routing pipeline posts every three hours, with a separate daily job for caption-less video summaries.
Outcome
Routes curated items into nine topic channels every three hours and produces summaries for selected videos that have no captions.