Explorar.dev

Explorar is a Next.js web application with a VS Code-like interface for browsing large source repositories through curated guides, indexed search, cross-references, diagrams, and knowledge checks.

Build your own local copy of the Explorar application, which will download all the git tags to your filesystem (get a coffee, this may take some time), and subsequently build the "shell" web application that works entirely offline and instantly.

npm install
npm run dev      # starts at localhost:3000

npm run lint     # tsc + eslint + prettier + depcheck
npm run build    # static export to out/ for a static host
npm run deploy   # production deploy: R2 sync of repo corpus
npm test         # Playwright tests (requires built output)

Online you can visit https://explorar.dev for free. Repository source files are loaded on demand from a public bucket-backed origin, with R2 as the default source.

Roadmap

Explorar should stay a guide-first source browser. The goal is to improve guides, search, graph views, and editor navigation with structured indexing and language-aware retrieval, not to turn the product into a chat-first tool.

The long-term architecture is:

Repository
  ↓
Config + Ignore Rules
  ↓
Language Detection
  ↓
Per-Language Indexers
  ↓
Unified Knowledge Base
  ↓
Retrieval Layer
  ↓
Monaco + Guides + Graph UI

For public web deployment, expensive analysis should happen in offline index builders. The browser client and read-only query APIs should consume prebuilt source snapshots and semantic artifacts rather than spawning per-user language-server sessions.

Milestone 0: Index Foundation

• Generate a SQLite code index for curated repository snapshots.
• Store files, symbols, references, file edges, guide links, concept links, and search tables in the index schema.
• Load and cache code-index.sqlite in the browser for static deployments.
• Keep heuristic navigation available as a fallback when an index is missing or incomplete.
• Add explorar.toml for repo-specific indexing config.
• Add .explorarignore for excluded paths.
• Add a local .explorar/ cache layout for generated index artifacts.

Milestone 1: Editor Backend Abstraction

• Add a live editor language backend interface keyed by languageId.
• Wire Monaco definition, reference, and hover flows through the indexed backend before falling back to heuristics.
• Provide indexed symbol, file, reference, graph-neighbor, guide-link, and concept query helpers.
• Share one backend contract between offline indexing and live editor queries.
• Implement backend-provided diagnostics and document symbols.

Milestone 2: C / C++ Semantic Navigation

• Integrate clangd for C/C++ indexing and editor queries.
• Replace heuristic-only C/C++ navigation where clangd data is available.
• Support parser-backed hover, definitions, references, diagnostics, type/member traversal, and include chains.
• Validate on Linux, XNU, seL4, and CPython native runtime code.

Milestone 3: Python Semantic Navigation

• Integrate pyright or basedpyright for Python indexing and editor queries.
• Resolve Python symbols across stdlib, tests, tools, and scripts.
• Support Python hover, definitions, references, and diagnostics through the backend path.

Milestone 4: Graph Retrieval And Concepts

• Persist basic file edges, guide links, and concept links in the index.
• Add callers/callees, include/import expansion, and related-files panels.
• Resolve guide mentions to multiple symbol candidates instead of only file paths.
• Add concept views that group docs, tests, headers, and implementations.
• Add scoped search tabs for files, symbols, and references.

Milestone 5: CPython Mixed Traversal

• Connect Include/*.h, Objects/*.c, Python/*.c, Lib/*.py, Lib/test/*.py, Tools/*.py, and docs/python_cpython.md in one retrieval model.
• Make PyObject, PyTypeObject, and PyDictObject resolve correctly.
• Make a concept like dict jump across docs, tests, headers, implementations, constructors, and guide chapters.
• Prove the same retrieval model on Linux and XNU kernel code.

Milestone 6: Optional Semantic Retrieval

• Add embeddings only after exact symbol and graph retrieval are strong.
• Keep semantic search additive; exact structural lookup remains the primary retrieval path.

UX Backlog

• Deep-linkable file, line, and symbol URLs.
• File outline panel.
• Per-file metadata strip with language, size, source, and related guide sections.
• Persist open tabs and workspace state locally.
• Side-by-side file compare for two refs or branches.
• Guide-aware navigation polish such as opening all files in a chapter.
• Subsystem map / architecture view.
• Blame and history overlays.
• Dependency graph explorer.
• API surface explorer.
• Version-aware guides.
• Ownership or maintainer overlays.
• Build-target awareness.
• Automatic interesting-files ranking.