Installation & Prerequisites

Prerequisites

yt-issue-reviewer is read-only against YouTrack and does all AI work on a self-hosted Ollama instance — no third-party hosted AI is ever used.

  1. uv — Python 3.11+ toolchain and runner.

  2. ``youtrack-cli`` installed & authenticated — this tool reads YouTrack only through the yt CLI and reuses its existing auth. It is a separate install (intentionally not a bundled dependency, so yt-issue-reviewer stays free of youtrack-cli’s heavy native transitive deps — see issue #22). Set it up once: .. code:: bash

    uv tool install youtrack-cli # or: pipx install youtrack-cli yt auth login yt issues list –project-id PROJ –state Open # sanity check

    Config lives at ~/.config/youtrack-cli/.env (or YOUTRACK_BASE_URL / YOUTRACK_TOKEN env vars).

  3. A reachable self-hosted Ollama with the models pulled: .. code:: bash

    ollama pull nomic-embed-text # embeddings (default, 768-dim) ollama pull qwen2.5 # only needed for –label (generated group labels)

    Ollama may be on localhost or reached over Tailscale — set the address with --ollama-host or the OLLAMA_HOST env var (see configuration).

Install

From source (until a released version is on PyPI — see releasing):

git clone https://github.com/ryancheley/yt-issue-reviewer
cd yt-issue-reviewer
uv sync
uv run yt-issue-reviewer --help

Once published, it will be installable directly:

uv tool install yt-issue-reviewer     # or: pip install yt-issue-reviewer

Verify connectivity

uv run yt-issue-reviewer doctor --ollama-host http://<host>:11434

doctor checks that yt is present + authenticated and that Ollama is reachable with the required model pulled. Next: the quickstart.