---

name: integrations-and-extensions description: "Handle code-review-graph custom languages, embeddings, wiki, registry, daemon, GitHub Action, and eval workflows." disable-model-invocation: true metadata: disco-role: operating license: MIT

Integrations and Extensions

Use this sub-skill for advanced or optional workflows: custom parser languages, semantic embeddings, wiki generation, multi-repo registry and daemon support, GitHub Action integration, and evaluation/benchmark guidance.

Start here

1. Confirm the requested feature is truly optional or integration-focused.
2. Read the narrow reference for the exact surface.
3. Install only the extra needed for that surface, not all by default.

Read references/custom-languages.md for .code-review-graph/languages.toml workflows. Read references/embeddings.md for semantic-search and provider behavior. Read references/registry-and-daemon.md for multi-repo registry, watch daemon, and repo listing. Read references/github-action.md for PR review automation. Read references/eval.md for benchmark/reproduction guidance. Read references/troubleshooting.md when an optional dependency, provider, grammar, daemon state, or eval run fails.

Route by task

User task	Do this
"Add a custom language"	Edit .code-review-graph/languages.toml and rebuild the graph.
"Enable semantic search"	Install the embeddings extra, then run the embedding workflow.
"Generate wiki pages"	Use the wiki generation workflow after communities exist.
"Register multiple repos"	Use the registry/daemon commands.
"Run a PR review in CI"	Use the GitHub Action pattern and the bundled PR renderer.
"Reproduce benchmarks"	Follow the eval reference and be prepared for expensive, repo-clone-heavy work.

Optional dependency policy

Only install an optional extra when the task requires it:

• code-review-graph[embeddings] for local sentence-transformers embeddings.
• code-review-graph[google-embeddings] for Google Gemini embeddings.
• code-review-graph[communities] for igraph-based community detection.
• code-review-graph[wiki] for Ollama-backed wiki summaries.
• code-review-graph[eval] for benchmark and report helpers.
• code-review-graph[enrichment] for deeper Python call-resolution enrichment.

Do not install all unless the user explicitly wants every optional feature in one environment.

Safe starting points

• Custom language users should start with a small .code-review-graph/languages.toml and a rebuild.
• Embedding users should confirm provider/environment variables before enabling cloud providers.
• Registry/daemon users should ensure each registered repository has a valid graph or watchable checkout.
• GitHub Action users should prefer the split analysis/comment workflow for fork safety.

Boundaries

• For installation and server setup, use install-and-setup.
• For diff review and PR comments, use review-changes.
• For structural exploration, search, and refactor previews, use graph-exploration.

Verification anchors

Native tests that ground this route include custom-language tests, embeddings tests, registry and daemon tests, wiki tests, eval tests, and PR workflow security tests. Some of these are optional or service-backed; record unverified optional capabilities explicitly instead of implying full support.