Hacker News Reader

That would require custom training. This project only does inference

No need to fine tune the model. The model could be augmented with retrieved context (as discussed in my sibling comment).