SmartSearch A* (SSA*) is a powerful, hybrid search engine that blends classic keyword matching with cutting-edge semantic understanding using deep learning. SSA* enhances traditional search pipelines with A* pathfinding logic, Transformer embeddings from HuggingFace, BM25 relevance scoring, and FAISS vector indexing—all wrapped in a scalable, microservice-based backend.
- Hybrid Retrieval Pipeline:
- BM25 for fast and effective keyword-based ranking (via Elasticsearch).
- Semantic Embeddings powered by HuggingFace Transformers.
- FAISS for high-speed approximate nearest neighbor search.
- A* Algorithm Integration to optimize result ranking based on multi-objective heuristics (relevance, distance, cost).
- Scalable Microservice Architecture for production-level performance.
- Model Optimization with LightGBM for ranking and result fusion.
- Significant Gains:
- 🔼 35% boost in search accuracy.
- 🔽 40% reduction in average query latency.
- TensorFlow – For model development and integration.
- FAISS – Fast vector similarity search.
- Elasticsearch – BM25 keyword indexing and querying.
- HuggingFace Transformers – Embedding generation for semantic search.
- LightGBM – Learning-to-rank model for result fusion and re-ranking.
- Python (FastAPI) – Microservices backend for scalable deployment.
- Python 3.8+
- Docker & Docker Compose (recommended for deployment)
- Elasticsearch 7.x
- FAISS
- TensorFlow + HuggingFace Transformers
git clone https://github.com/Anayverma/SSAstar.git
cd SSAstar
pip install -r requirements.txt
# SSAstar