AI-Powered Semantic Search for Your Documents
Search your documents by meaning, not just keywords. Our local semantic search engine uses AI embeddings to understand context and find relevant information even when exact words don't match. Perfect for researchers, writers, developers, and anyone managing large document collections.
Unlike traditional keyword search, semantic search understands the meaning and context of your query, returning results that are conceptually related even if they don't contain the exact search terms.
Find documents by meaning using vector embeddings and AI, not just keyword matching. Search for "happy" and find results containing "joyful" or "delighted".
All data stays in your browser using IndexedDB. No server, no uploads, complete privacy. Your documents never leave your device.
Uses Transformers.js for in-browser embeddings with all-MiniLM-L12-v2 model (384 dimensions). No external API calls required.
Powered by PGlite with HNSW indexing for lightning-fast similarity search across thousands of document chunks.
Upload .txt, .md, or .pdf files, or paste text directly. Click the "Add Document" button to get started.
Documents are automatically chunked into smaller pieces and converted to vector embeddings using AI. This happens locally in your browser.
Type natural language queries to find relevant chunks. Ask questions or describe what you're looking for - no need for exact keywords.
Our system combines semantic similarity with keyword matching for the best results, giving you both conceptual relevance and exact matches.
View all your documents in the document list and delete ones you no longer need. Your data is always under your control.
Find related papers, notes, and concepts across your research library. Perfect for literature reviews and connecting ideas.
Search technical docs, API references, and code documentation to find solutions faster.
Build a personal knowledge base with notes, articles, and bookmarks. Your second brain for information retrieval.
PGlite - PostgreSQL running entirely in your browser with the pgvector extension for vector similarity search. Stores data in IndexedDB for persistence.
Xenova/all-MiniLM-L12-v2 via Transformers.js - Generates 384-dimensional vectors representing the semantic meaning of text. Runs completely in-browser using WebAssembly.
HNSW (Hierarchical Navigable Small World) indexing for approximate nearest neighbor search. Provides fast and accurate similarity search.
Combines cosine similarity (semantic vector search) with PostgreSQL full-text search (keyword matching) for comprehensive results.
Smart chunking with overlap for better context preservation. Documents are split into ~300 character chunks with 50 character overlap to maintain semantic continuity.
PDF.js for parsing and extracting text from PDF documents directly in the browser.
This is a fully client-side application. All document processing, embeddings generation, and search happen locally in your browser. Your documents never leave your device.
No data is sent to any server. Your documents are stored in IndexedDB and persist only in your browser. You have complete control over your data - you can export, delete, or clear everything at any time.
We don't track you, collect analytics, or store any information about your searches or documents. This tool is built with privacy as a core principle.
Traditional keyword search (Ctrl+F) only finds exact matches. Semantic search understands meaning - if you search for "automobile accidents", it will also find content about "car crashes" or "vehicle collisions" even without those exact words.
Yes, 100%. Everything runs in your browser. Your documents are stored in IndexedDB (browser storage) and never sent to any server. You can verify this by checking your browser's network tab - no document data is transmitted.
Browser storage limits vary by browser (usually 50-100GB for modern browsers). IndexedDB will use as much as your browser allows. The app will notify you if you're running low on storage.
Currently supports .txt (plain text), .md (Markdown), and .pdf (PDF) files. You can also paste text directly into the app.
After the initial load and model download, the app works offline. Your documents and search functionality will be available without an internet connection.
The all-MiniLM-L12-v2 model is trained on millions of text pairs and provides excellent semantic understanding for most use cases. Results combine both semantic similarity and keyword matching for optimal accuracy.
This tool implements Retrieval Augmented Generation (RAG) techniques using vector embeddings and similarity search. Vector databases like PGlite with pgvector enable efficient storage and retrieval of high-dimensional embeddings.
By converting text into numerical vectors, we can measure semantic similarity using cosine distance. This enables powerful features like finding similar documents, clustering related content, and building intelligent search systems.
Whether you're building a knowledge base, document management system, or research tool, semantic search provides a more natural and effective way to find information compared to traditional keyword-based approaches.
Start searching your documents with AI-powered semantic search. Add your first document and experience the power of vector embeddings.
Start Searching Now