Description

Build a self‑hosted AI assistant during hackweek that runs on the home workstation, provides a local chat interface, and understands the user’s GitHub projects well enough to help with organisation, triage, and lightweight code review. The project should start as a “local ChatGPT‑style” helper and be structured so it can later evolve into an automated GitHub bot that reacts to issues and pull requests.

Goals

• Stand up a local LLM (via Ollama or similar) plus a browser‑based chat UI, usable for general Q&A and pasted GitHub content.
• Define and capture a few core workflows for GitHub help: issue triage, label suggestions, PR summaries, and “next action” hints.
• Sketch (or partially implement) an integration plan where GitHub events can be passed to the local model and its responses posted back as comments.

Resources

• Home workstation with AMD GPU and 16 GB RAM.
• Local LLM runtime (e.g. Ollama + a 7B‑class coder/instruct model).
• Existing GitHub repositories to use as testbeds for prompts and workflows.