Table of Contents

AI-Assisted Development

This section explains how to integrate AI agents into the development workflow — from choosing an LLM provider to running a fully automated issue-to-PR pipeline.

What this section covers

  • Overview of the AI agent pipeline (issue-resolver, pr-reviewer, pr-resolver)
  • LLM inference: cloud API vs. self-hosted (llama.cpp / LiteLLM gateway)
  • Self-hosted inference: llama.cpp Docker setup, quantized models, LiteLLM gateway
  • Claude Code setup: RTK installation, global config
  • Agent workflow end-to-end: YouTrack issue to Gitea PR to merge

The agent pipeline

The platform supports an AI agent pipeline that automates the software development inner loop:

  1. issue-resolver — reads a YouTrack issue, implements the change, opens a PR.
  2. pr-reviewer — reviews the PR and requests changes or approves.
  3. pr-resolver — addresses review comments and re-submits.

Each agent is a Claude Code session orchestrated by Gitea Actions workflows.

Pages in this section