autocomplete

• Add the v2 streaming suggestions API with requestSuggestions / streamSuggestions, cancellation, and richer request events.
• Add SuggestionRequestMode.staticThenFinal for fast static results followed by dynamic updates.
• Improve responsiveness by deferring generateSpec during the first static pass.
• Add the stream_suggest_v2.dart example, runtime tests, and benchmark coverage.
• Add built-in specs for claude, codex, gemini, and qwen.

Breaking Changes

• Rename clearCache() to clearDynamicSuggestion().
• Add clearAllCaches() for callers that want the previous full-reset behavior.

• bind AutocompleteEngine to a fixed adapter and simplify engine.getSuggestions(cmd, cwd, shell)
• stabilize dynamic suggestions cache reuse for lazy-loaded specs, postProcess generators, and custom function generators
• fuzzy filter opt

• feature fix, opt

• fuzzy suggestion filter, and tests

• performance improvements
• add object-oriented API

• Fix multi-version spec selection (e.g. heroku, shopify, infracost) so the correct spec is chosen at runtime based on CLI version.
• Add CONTRIBUTING.md with contribution and spec-writing guidelines.
• Update spec definitions (elixir, fin, git, heroku, pnpm, shopify) and generator/runtime logic.

• Update spec definitions.
• Refactor code structure.

• Optimization and bug fixes.

• Prerelease for testing.