Repowise tutorial adds repository-level AI Code Intelligence to itsdangerous

A step-by-step tutorial demonstrates using Repowise to build repository — level AI code intelligence for the itsdangerous Python project, by indexing the repo and producing artifacts that reveal structure, influence and maintenance signals. The guide walks a user from an already cloned repository through Repowise setup, the indexing pipeline, and inspection of generated outputs. This matters because maintainers can use those outputs to prioritize work and reduce technical debt.

The walkthrough begins with basic CLI checks such as repowise --version and repowise init, and shows how to configure LLM credentials via environment variables (ANTHROPIC_API_KEY or OPENAI_API_KEY) before writing a.repowise/config.yaml. The example configuration lists provider/model choices (anthropic → claude — sonnet-4-5, openai → gpt-4o-mini, or mock), sets embedding_model: voyage-3 and reasoning: auto, and enables Git and maintenance settings including git co_change_commit_limit: 200 and blame_enabled: true.

Dead-code detection is enabled with safe_to_delete_threshold: 0.7, and maintenance is constrained with cascade_budget: 10. Concrete analyses produced by the tutorial include repository graph metrics (PageRank and community detection), Git intelligence, dead-code detection, captured architectural decisions, a generated CLAUDE.md file, and MCP-style CLI interactions. The guide emphasizes combining graph analysis with LLM context and embeddings so Repowise can enrich repository signals while working with multiple LLM backends and a dedicated embedding model.

As a result, maintainers can visualize the most important nodes in the repository graph to understand structure, influence, dependencies and maintenance priorities. The tutorial’s configurable thresholds and budgets (safe_to_delete_threshold 0.7, cascade_budget 10, co_change_commit_limit 200) illustrate how the pipeline flags deletion candidates and scopes maintenance effort in a reproducible workflow.