v0.2.0

Added

• LLM-assisted workstream clustering (shiplog-cluster-llm): optional OpenAI-compatible endpoint for semantic workstream grouping, with automatic fallback to repo-based clustering on failure
• import subcommand: re-render a pre-built ledger directory from an upstream system or previous shiplog run
• Redaction alias cache persistence: redaction.aliases.json provides stable alias mappings across runs
• --zip flag: write a zip archive next to the run folder (available on collect, render, refresh, import, run)
• --run-dir flag: explicit run directory for refresh (overrides auto-detection)
• SourceSystem::Other(String) variant: extension point for third-party source systems (#[non_exhaustive])
• Bundle exclusion of redaction.aliases.json: alias cache is excluded from bundle manifests and zip archives to prevent redaction bypass
• LLM feature gate: shiplog-cluster-llm is now an optional dependency behind the llm feature (default on); build with --no-default-features to omit it
• Module-level documentation (//! doc blocks) for shiplog-schema, shiplog-coverage, shiplog-workstreams, shiplog-engine, and shiplog-ports
• CLI Reference section in README.md with full flag table
• LLM clustering flags (--llm-cluster, --llm-api-endpoint, --llm-model, --llm-api-key) documented across all 4 doc files
• Missing sections in GEMINI.md and copilot-instructions.md to sync with CLAUDE.md

Changed

• Crate-specific descriptions for all 15 publishable crates (replacing generic workspace description)
• Added keywords and categories to all publishable crate Cargo.toml files
• Marked shiplog-testkit as publish = false
• CACHE_FILENAME in shiplog-redact is now pub for cross-crate documentation