Semantic Search With Hybrid Retrieval Strategies

via “semantic search and retrieval with query-time reranking”

<p align="center"> <img height="100" width="100" alt="LlamaIndex logo" src="https://ts.llamaindex.ai/square.svg" /> </p> <h1 align="center">LlamaIndex.TS</h1> <h3 align="center"> Data framework for your LLM application. </h3>

Unique: Abstracts retrieval strategies behind a pluggable Retriever interface, allowing developers to compose vector search, BM25, and LLM-reranking without changing application code, and supporting query-time metadata filtering across heterogeneous vector stores

vs others: More composable than LangChain's retriever chain because it separates retrieval strategy from reranking logic, enabling A/B testing of different reranking models without modifying the retrieval pipeline

via “hybrid search combining vector and full-text retrieval”

Serverless embedded vector DB — Lance format, multimodal, versioning, no server needed.

Unique: Integrates full-text and vector search at the storage layer using Lance's columnar format, avoiding separate indices and enabling single-pass retrieval; combines both modalities without requiring external search engines like Elasticsearch

vs others: Simpler than Elasticsearch + vector plugin because both search modes share the same columnar storage, but less mature than Pinecone's hybrid search in terms of tuning options and performance optimization

via “hybrid search combining dense and sparse retrieval”

LangChain reference RAG implementation from scratch.

Unique: Implements hybrid search by running parallel dense (vector similarity) and sparse (BM25) retrieval and merging results using configurable weighting (e.g., 0.7 * dense_score + 0.3 * sparse_score), enabling developers to tune the balance between semantic and lexical relevance.

vs others: More effective than pure semantic search for specialized vocabularies because BM25 captures exact term matches; more practical than pure keyword search because dense retrieval captures semantic relationships and synonyms that keyword search misses.

via “hybrid retrieval combining vector and keyword search”

LlamaIndex starter pack for common RAG use cases.

Unique: LlamaIndex's retriever composition pattern enables pluggable fusion strategies and easy swapping of retrieval methods, whereas most RAG systems hard-code a single retrieval approach

vs others: More flexible than Elasticsearch's hybrid search because LlamaIndex's retriever abstraction decouples fusion logic from storage backend, enabling experimentation with different ranking strategies without re-indexing

via “hybrid retrieval with bm25 keyword search and semantic reranking”

LlamaIndex is the leading document agent and OCR platform

Unique: Combines vector search, BM25 keyword matching, and optional semantic reranking with configurable fusion algorithms and support for multiple reranker backends. Unlike LangChain's retriever composition (which chains retrievers sequentially), LlamaIndex's hybrid retrieval merges results with configurable fusion.

vs others: Provides integrated hybrid retrieval with automatic result fusion and optional reranking, whereas LangChain requires manual retriever composition and result merging.

via “vector semantic search with hybrid ranking”

Lightning-fast search engine with vector search.

Unique: Implements hybrid search through configurable weighted fusion of keyword and vector scores at query time, allowing dynamic adjustment of semantic vs lexical emphasis without reindexing. Uses arroy library for vector storage, which is optimized for LMDB-backed persistence rather than in-memory indexes.

vs others: Simpler to integrate than Pinecone or Weaviate because it's a single self-hosted binary; more flexible than Elasticsearch vector search because it supports external embedding providers without requiring Elasticsearch's inference API.

via “sparse-embedding-generation-for-hybrid-search”

Framework for sentence embeddings and semantic search.

Unique: Provides sparse encoder models for hybrid search, enabling combination of dense semantic embeddings with sparse keyword-aware embeddings in unified framework; differentiates by supporting both embedding types without requiring separate libraries or complex integration

vs others: More flexible than dense-only search because it combines semantic understanding with keyword matching, and simpler than building custom hybrid systems with separate dense and sparse components

via “hybrid search combining full-text and vector results”

🌌 A complete search engine and RAG pipeline in your browser, server or edge network with support for full-text, vector, and hybrid search in less than 2kb.

Unique: Implements score normalization and weighted combination of BM25 and cosine similarity in a single unified query interface, allowing developers to tune the balance without maintaining separate search endpoints. Most vector databases treat hybrid search as an afterthought; Orama makes it a first-class citizen with configurable weighting.

vs others: Simpler API than Elasticsearch's hybrid search which requires separate queries and manual score combination; more flexible than Pinecone's hybrid search which uses fixed weighting algorithms.

via “hybrid vector + full-text search with combined ranking”

Low-cost vector database — pay-per-query, S3-backed, up to 10x cheaper at scale.

Unique: Provides native hybrid search combining vector and full-text signals in a single query without requiring application-level result merging or separate API calls, with unified ranking across both modalities within the same namespace isolation model

vs others: More efficient than querying vector and full-text search separately and merging results in application code because ranking is unified server-side, reducing latency and eliminating deduplication logic

via “semantic and hybrid retrieval with query expansion”

Unified framework for building enterprise RAG pipelines with small, specialized models

Unique: Implements query expansion at retrieval time using small specialized models (SLIM models) to inject synonyms and related concepts, improving recall without expensive reranking. Hybrid retrieval combines vector similarity with keyword matching through configurable alpha weighting, enabling both semantic and exact-match queries in a single call.

vs others: Built-in query expansion via SLIM models improves recall vs static vector-only retrieval; hybrid approach handles both semantic and keyword queries vs pure vector solutions like Pinecone; integrated with llmware's small model ecosystem for on-device expansion.

via “fusion-retrieval-with-multi-strategy-ranking”

This repository showcases various advanced techniques for Retrieval-Augmented Generation (RAG) systems. Each technique has a detailed notebook tutorial.

Unique: Implements Reciprocal Rank Fusion and weighted scoring to combine dense semantic retrieval with sparse keyword retrieval, allowing developers to balance semantic understanding with exact-match precision without choosing one strategy — a hybrid approach that's more robust than single-strategy retrieval

vs others: More comprehensive than pure semantic search because it captures both meaning and keywords, and more practical than pure BM25 because it includes semantic understanding; fusion is more maintainable than building a custom unified ranking function

via “hybrid retrieval with semantic and keyword search fusion”

Open-source LLM knowledge platform: turn raw documents into a queryable RAG, an autonomous reasoning agent, and a self-maintaining Wiki.

Unique: Decouples semantic and keyword retrieval into independent pipelines with pluggable reranking, allowing fine-grained control over fusion strategy per knowledge base. Supports multiple reranking backends (BM25, cross-encoder models) without requiring model retraining.

vs others: More flexible than pure semantic search (handles domain jargon better) and more intelligent than keyword-only search (understands intent), with configurable reranking that adapts to domain-specific precision/recall tradeoffs.

via “multi-strategy document search with tree, metadata, semantic, and description-based retrieval”

📑 PageIndex: Document Index for Vectorless, Reasoning-based RAG

Unique: Implements four orthogonal search strategies (tree-based, metadata, semantic, description) all operating on the same hierarchical index, allowing composition and fallback mechanisms. Unlike vector-only systems, it provides explicit control over retrieval strategy and can combine multiple approaches for improved recall.

vs others: More flexible than single-strategy vector RAG because it supports metadata and description-based search without requiring separate indices, and allows explicit strategy composition rather than relying solely on embedding similarity.

via “hybrid search combining graph traversal and vector semantic similarity”

The memory for your AI Agents in 6 lines of code

Unique: Implements a search router (cognee/modules/search/methods/get_retriever_output.py) that dynamically selects between graph traversal, vector similarity, and hybrid fusion based on query characteristics, rather than forcing a single search strategy. Uses configurable scoring functions that allow developers to weight structural vs. semantic relevance per use case, enabling fine-tuned retrieval behavior.

vs others: More sophisticated than pure vector RAG (like Pinecone) because it preserves and leverages explicit relationships for multi-hop reasoning; more flexible than pure graph databases (Neo4j alone) because it combines structural queries with semantic similarity to handle ambiguous or paraphrased queries that wouldn't match exact relationship patterns.

via “hybrid-search-semantic-and-keyword-fallback”

MCP server and Claude plugin for Postgres skills and documentation. Helps AI coding tools generate better PostgreSQL code.

Unique: Implements both semantic (pgvector cosine similarity) and keyword (BM25) search on the same documentation corpus, allowing AI models to choose the most appropriate method per query. Both methods are in-database, avoiding external search service dependencies. Results are returned in the same format, enabling easy comparison and combination.

vs others: More flexible than semantic-only or keyword-only search because it supports both approaches and allows AI models to choose. More cost-effective than external search services because both methods use in-database indexing. More effective than single-method search because it enables fallback strategies and hybrid result combination.

via “semantic search and retrieval with ranking”

A data framework for building LLM applications over external data.

Unique: Implements a pluggable Retriever abstraction supporting multiple retrieval strategies (similarity, MMR, fusion, custom) that can be composed and chained. Built-in support for re-ranking via LLM or cross-encoder, and hybrid search combining dense and sparse retrieval without custom integration code.

vs others: More flexible retrieval composition than LangChain's retrievers; built-in re-ranking and fusion strategies reduce boilerplate for advanced retrieval pipelines.

via “semantic-context-retrieval-with-hybrid-search”

MineContext is your proactive context-aware AI partner（Context-Engineering+ChatGPT Pulse）

Unique: Implements hybrid search combining vector similarity with structured SQL filters, enabling queries that blend semantic relevance with temporal and categorical constraints. Supports both programmatic API and UI-based search with configurable ranking and filtering.

vs others: More powerful than vector-only search because it enables structured filtering (date range, type) combined with semantic similarity, whereas vector-only databases lack efficient categorical filtering. More intelligent than SQL-only search because it understands semantic meaning rather than just keyword matching.

via “two-stage retrieval with dense-sparse hybrid search”

A modular Agentic RAG built with LangGraph — learn Retrieval-Augmented Generation Agents in minutes.

Unique: Implements parallel dense+sparse search with reciprocal rank fusion (RRF) merging in a single Qdrant query, rather than maintaining separate indices or sequentially executing searches. The VectorDatabaseManager class abstracts the hybrid search logic, enabling transparent switching between retrieval strategies without changing the agent code.

vs others: Outperforms pure dense retrieval on keyword-heavy queries and pure BM25 on semantic queries; the hybrid approach captures both signal types in a single retrieval pass, reducing latency vs sequential search strategies.

via “hybrid semantic and keyword search with adaptive strategy selection”

Memento MCP: A Knowledge Graph Memory System for LLMs

Unique: Implements adaptive strategy selection that automatically routes queries to semantic or keyword search based on query characteristics, rather than requiring explicit user configuration. Combines Neo4j's vector index and full-text index capabilities in a single unified search interface.

vs others: More intelligent than single-strategy search systems; avoids the latency overhead of always running both semantic and keyword searches by adaptively selecting the optimal path.

via “hybrid keyword-semantic search with weighted fusion”

A lightning-fast search engine API bringing AI-powered hybrid search to your sites and applications.

Unique: Uses weighted fusion of separate inverted indexes (for keyword) and arroy vector stores (for semantic) with configurable semanticRatio parameter, enabling per-index tuning of keyword vs. semantic weight without requiring external ranking services or re-indexing

vs others: Faster than Elasticsearch's hybrid search because Meilisearch's Rust-based milli engine pre-computes both index types at ingest time rather than computing similarity scores at query time, achieving sub-50ms latency on large datasets