DevShelf: Download Distributed Search Engine (Java)

Windows

• 64-bit installer for Windows 10 and 11
• Latest stable release: v1.0.0

⚡ System Overview

DevShelf is a vertical search engine engineered to index and rank Computer Science literature without relying on external search frameworks.

Unlike typical academic or student projects, DevShelf follows a split architecture that separates indexing from query-time execution to guarantee predictable performance.

Architectural Separation

1. Offline Layer (IndexerMain)
   Responsible for corpus traversal, text normalization, and index construction.

2. Online Layer (BookSearchEngine)
   Serves sub-millisecond queries using memory-resident, precomputed data structures.

System Architecture Diagram

The following diagram illustrates the separation between offline indexing and online query execution:

🏗️ Core Architecture

1. Offline Search Indexing

Index construction is treated as a batch operation to eliminate runtime overhead during user queries.

• Index Builder
  IndexBuilder processes the document corpus using a custom NLP pipeline (TextProcessor).

• Primary Data Structure
  A Positional Inverted Index is generated and serialized, enabling constant-time term lookups.

• Design Goal
  Shift all expensive computation out of the query path.

2. Intelligent Query Processing

The online engine (QueryProcessor) applies a multi-stage retrieval pipeline:

• Lexical Matching
  Candidate documents are retrieved directly from the inverted index.

• Fuzzy Matching
  Typographical errors are handled using Levenshtein Distance–based correction.

• Predictive Search
  Autocomplete suggestions are generated via a Trie (Prefix Tree) with O(L) lookup complexity.

🧠 Hybrid Ranking Strategy

Document relevance is computed using a weighted scoring model that combines lexical relevance with behavioral signals.

Scoring Formula (Conceptual):

This approach balances traditional information retrieval techniques with real-world usage patterns.

👥 Engineering Team

DevShelf was built by a focused team of three engineers, each owning a distinct subsystem.

📘 Documentation

For detailed design notes, algorithms, and implementation details:

View the full documentation