semantic-search
Run the following command in DeepSeek Harness:
dsh plugin install JohnXu22786/semantic-search
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install JohnXu22786/semantic-search inside DeepSeek Harness to install this plugin; full source is available at https://github.com/JohnXu22786/semantic-search .
About this plugin
As codebases grow, grep only matches literal strings while the real question is often "where does this logic live." semantic-search gives DeepSeek Harness a local, offline, dependency-free semantic code search: it builds a fragment-level index of a workspace and retrieves results through a hybrid vector-cosine + BM25 pipeline with reciprocal-rank fusion, so meaning-based hits are returned even when exact terms differ.
Highlights include a language-aware tokenizer (camel/snake/kebab splitting, CJK bigram n-grams, 16 languages), symbol-aware chunking that keeps functions and classes intact, and a default feature-hashed TF-IDF embedding that works fully offline with no network call or API key. A remote OpenAI-compatible endpoint is also supported, with automatic fallback to the local provider. The index supports incremental refresh, optional file watching, atomic persistence under .sema, and deterministic results.
It is well suited for engineers who search large codebases inside DeepSeek Harness, teams with strict offline or privacy requirements, and projects where Chinese code or comments are prevalent. It plugs in as Harness tools (sema_search / sema_reindex / sema_stats) or runs standalone as a CLI in any Node script.
Use Cases
- Locate relevant code fragments with natural-language queries in a large codebase
- Run semantic code search fully offline without any API key
- Cross-phrasing retrieval over projects with heavy Chinese code or comments
Best For
- Engineers who search code frequently inside DeepSeek Harness
- Teams with strict offline or privacy-isolation requirements
- Developers whose codebases mix Chinese and English comments
Related Plugins
A unified suite combining hot runtime injection, task-aware thinking-mode routing, and a graded session protocol with red-team gates to sustain model diligence across long-horizon inference.
ModLens is a vision plugin for DeepSeek Harness that gives text-only models sight by reading images pasted directly into chat, with zero-config setup and multiple vision engines.
On-demand vision for text-only DeepSeek Harness agents: built-in free keyless vision chain and 14 vision tools, routing image turns as tool calls to vision models with pixel fidelity, no Python needed, one-command install.
Give text-only models in DeepSeek Harness eyes, enabling image Q&A, long-screenshot OCR, UI restoration, and GUI visual tasks.