deepwiki-rs vs Graft
Litho: a Rust, DeepWiki-style doc generator that turns any codebase into C4-model architecture docs - context, container, component and code-level pages with diagrams - via any LLM. — versus — Context layer for large codebases: a graph of plain-English markdown nodes — no embeddings, no index — agents read like any repo file. Claude Code hooks + MCP; 42% fewer tokens in its bench.
graft builds a graph of plain-English markdown nodes for agents to navigate large codebases; Litho's C4 docs serve humans first and agents second.
| deepwiki-rs | Graft | |
|---|---|---|
| Stars | 3.0k | 8.4k |
| Forks | 297 | 760 |
| Language | Rust | TypeScript |
| License | MIT | MIT |
| Last activity | 5 days ago | 2 days ago |
| Topics | code-intel, coding | code-intel, rag |
| Curated connections | 4 | 6 |
deepwiki-rs — the curator's take
Use Litho when you want a one-shot, opinionated architecture wiki (C4 layers, diagrams, multilingual) for a repo you inherited or need to onboard people onto - it is fast, single-binary, and the output reads like a human wrote it. Not a live code-intelligence layer: it regenerates rather than syncs, has no query API, and agents consume it as static markdown. The author has moved on to Terrain (which keeps the knowledge base in sync and serves it over ACP); Litho remains the focused generator. If you want the docs to stay current as agents edit code, mex or graft are the living-wiki alternatives; if you want a click-around dependency map instead of prose, codeflow.
Graft — the curator's take
The anti-embeddings bet is the story: the codebase map is prose files — senior-engineer explanations, linked and greppable — so agents consume it with the file tools they already have, and a stale map shows up as a diff in review instead of rotting in a vector store. The 162-run controlled benchmark holding correctness equal is more rigor than this category usually shows; still self-run. Claude Code gets the deep integration (hooks, statusline, background rebuild), everyone else gets MCP. When NOT: nodes are LLM-written prose — building them on a big repo costs real tokens, and prose can drift subtly wrong in ways a symbol graph can't; spot-check the nodes on your hot paths. ★437 and young, with NanoNets behind it.