Skip to content

Latest commit

 

History

History
65 lines (51 loc) · 3.62 KB

File metadata and controls

65 lines (51 loc) · 3.62 KB

litewiki Roadmap (Ideas & Directions)

This is a living document collecting ambitious ideas for where litewiki could go. These are intentionally broad and speculative to encourage exploration.

1) Agentic Workflows (Devin/Cursor-class UX)

  • Autonomous workflows: run multi-step plans (scan repo, propose fixes, open PRs, write docs) with human approval gates
  • Goal-driven sessions: accept a high-level objective and generate a plan with milestones/tasks
  • Inline PR iteration: read CI output, propose/code fixes, push commits until green
  • Guardrails: dry-run mode, diffs-only mode, and “explain plan” before execution

2) First-class Editor Integrations

  • VS Code/JetBrains/Neovim extensions: trigger “wiki run”, preview reports, jump-to-code from report
  • Inline hover cards: show summaries/graphs for functions/types/dirs
  • One-click “document this file/module” from editor context

3) Rich Knowledge Graphs

  • Build a semantic graph (files ↔ symbols ↔ modules ↔ commits ↔ issues/PRs)
  • Interactive visualizations (Mermaid/D3): dependency graphs, call graphs, ownership overlays
  • Cross-repo views: organization-level knowledge map; detect reused patterns and anti-patterns

4) Advanced Report Types

  • Design docs extraction: infer architecture views (C4), data flow, invariants, boundaries
  • Test docs: summarize test coverage, flaky hotspots, dead tests, suggested new tests
  • Security doc: secrets scanning (defensive only), risky patterns, dependency posture
  • Release notes: summarize since last tag, user notable changes, migration impact

5) i18n and Localization at Scale

  • Automatic language detection from repo metadata and contributors
  • Multi-lingual reports with side-by-side toggle
  • Community translation packs; crowdsource glossary per ecosystem (React, Rust, etc.)

6) Model & Runtime Flexibility

  • Local/offline mode: support small local models (gguf) for private runs; hybrid fallback to cloud
  • Provider mesh: choose best provider per subtask (summarization vs. code reasoning vs. graph)
  • Cost-aware routing: budgets per run; offline cache + deduplication of unchanged chunks

7) Extensible Plugins (Safe by Default)

  • WASM sandboxed plugins for analysis (no network by default; capability-scoped)
  • Plugin categories: framework analyzers (Next.js, Django), language analyzers (Go, Rust), infra docs (Terraform)
  • Plugin marketplace with signed manifests and reputation

8) Continuous Knowledge

  • Watch mode: incremental re-analysis on changes; PR comment bot posts deltas
  • Drift detection: code vs. docs divergence alerts; suggest updates as patch files
  • Temporal views: how the architecture evolved over time, hotspots trending up/down

9) Collaboration & Review

  • Shareable previews with private links; per-section comments that map back to code
  • Suggested reviewers based on ownership graph and past contributions
  • “Explain this diff” and “Summarize risk” for PRs

10) Education & Onboarding

  • First-week onboarding pack: guided repo tour, glossary, key decisions, coding standards
  • Skill maps for new contributors: starter issues, safe subsystems, mentored tasks

11) Performance & Scale

  • Parallel crawl + chunk schedulers; memoized analysis with structural hashing
  • Large monorepo strategies: partial runs by scope; repo-aware rate limits and caches

12) UX Experiments

  • TUI dashboard: fuzzy search across report sections, symbol lookup, live graphs
  • Web UI: progressive render of large reports; searchable across all past runs
  • “Ask the Repo”: chat over report + code (retrieval grounded on indexed artifacts)

If any of these resonate, open an issue or a discussion. Contributions welcome! 🧭