Skip to content

Semantic Code Search

AI-powered semantic code search via RAG. Find code by meaning, not keywords. Local AES-256-encrypted vector index, EU-hosted embedding providers.

🔍

Semantic Search

Find code by meaning, not just keywords. Ask questions about your codebase in natural language.

🔒

Local Index Storage

The vector index lives on your machine, AES-256-GCM encrypted. Code snippets pass through the Lurus Code backend to EU-hosted embedding providers for vector generation — no source code is ever stored on any server.

RAG Pipeline

Retrieval-Augmented Generation feeds relevant context into every AI interaction.

🔄

Auto-Reindexing

Detects file changes and keeps the index fresh without manual steps.

📊

Quota Management

Per-plan quota for embeddings with real-time usage tracking.

⌨️

/indexing Commands

Full control via slash commands in chat.

Keyword Search vs. Semantic Search

Traditional tools like grep only find what you type exactly. Semantic search understands meaning — finding the authentication module even if you ask about "user login" or "token validation".

keyword
grep "auth"Finds all files containing the string "auth"
semantic
Where is login handled?Finds auth.service.ts, jwt.middleware.ts, session.guard.ts
keyword
grep "validate"Finds every occurrence of "validate"
semantic
How is user input validated?Finds the validation pipe, DTOs, and decorator usage

Indexing Commands

/indexing onEnable semantic indexing for this project
/indexing statusShow: chunks, files, model, quota used
/indexing rebuildDelete index and rebuild from scratch
/indexing offDisable indexing (index preserved)
/indexing clearDelete local index and reset quota

When Semantic Search Helps Most

Large Codebases

In repos with 500+ files, grep becomes slow and noisy. Semantic search returns the 3–5 most relevant files immediately.

Onboarding

New team members can ask "How does X work?" and get accurate, context-aware answers without reading the entire codebase.

Finding Duplicates

Ask "Are there similar implementations to this function?" and get a ranked list of candidates for deduplication.

🇪🇺 Privacy & GDPR

Privacy & Security

Your source code stays on your machine. Only vector embeddings (mathematical representations, not readable code) are transmitted to EU-hosted embedding providers.

  • Vector index stored locally, AES-256-GCM encrypted
  • Code snippets sent to EU-hosted infrastructure for embedding generation only
  • No source code stored on any server after embedding
  • Embeddings are mathematically irreversible — cannot be decoded back to source code

Ready to code smarter?

Join the waitlist and let Lurus Code understand your entire codebase.

Get started