lonelymoon87/dsh-code-intel
dsh-code-intel gives DeepSeek Harness symbol-aware code outlines, persistent workspace indexing, and explicit lexical or embedding-assisted code search. code_search ranks AST symbol chunks and bounded module windows and returns path:line, symbol metadata, snippets, scores and the active retrieval mode; code_outline parses one file immediately or projects a directory from the persistent index. The first workspace search starts a cancellable code-index background job instead of blocking the turn. An index persists under .dsh/code-index/ (SQLite, incremental rebuilds via DSH filesystem versions), dirtying is driven by fs/observed plus a Chokidar watcher for shell/IDE/external changes. TypeScript, TSX, JavaScript, JSX, Python, Go, Rust and Java use install-script-free Tree-sitter WASM grammars. An optional OpenAI-compatible embedding endpoint adds cosine similarity; without it results explicitly say mode: lexical.
Install
dsh plugin --profile web add https://github.com/lonelymoon87/dsh-code-intel/releases/download/v0.1.3/dsh-code-intel-0.1.3.tgznpm dsh-code-intel 0.1.3 verified 2026-09-04 (repository field → github.com/lonelymoon87/dsh-code-intel). The release tarball is prebuilt and needs no build allowance; a pinned source install dsh plugin --profile web add github:lonelymoon87/dsh-code-intel#v0.1.3 runs the package prepare build (pnpm 10 may require an allowBuilds entry printed by the failed command). Supports DSH >=0.1.0-rc.6 <0.2.0 and Node ^22.19 || >=24.
Compatibility
DSH >=0.1.0-rc.6 <0.2.0 plugin APIs; Node ^22.19 || >=24; standard local-filesystem execution world (SQLite + Chokidar need host-accessible processPath).
Details
- Repo: lonelymoon87/dsh-code-intel
- Category: Coding & Development
- Stars: 0
- Version: dsh-code-intel 0.1.3 (GitHub release tarball)
- Last push: 2026-08-14
- First seen: 2026-08-13
Recent updates
v0.1.3 tested with DSH 0.1.0-rc.8 and 0.1.1-rc.1 while keeping the rc.6-compatible peer range.
FAQ
- How does code search work?
- Lexical mode scores exact query matches, tokens, symbol names and paths; hybrid mode combines that with cosine similarity from an optionally configured OpenAI-compatible embedding endpoint. Without an endpoint, results explicitly say mode: lexical.
- Does indexing block the agent?
- No — the first workspace search starts a cancellable code-index background job and returns an indexing result with a job id; file outlines work without a complete workspace index.
- Which languages are supported?
- TypeScript, TSX, JavaScript, JSX, Python, Go, Rust and Java, parsed with install-script-free Tree-sitter WASM grammars.
Alternatives
hellosky983/dsh-skillradar · clouwer/dsh-memsearch