How to Use
Browsing Techniques
Use the search box to find specific techniques. Filter by category or pipeline stage to narrow down results. Click on any technique card to view detailed information.
Understanding Relationships
When viewing a technique, scroll down to see related techniques grouped by relationship type. Click on any related technique to explore it. Use these relationships to design complete RAG pipelines.
Contributing
Found a technique we're missing? Click "Suggest Technique" to submit it. Notice incorrect information? Click "Report incorrect data" on any technique panel to let us know.
Categories
Query Transformation & Expansion
Query reformulation, expansion, rewriting, and augmentation techniques
Routing & Query Classification
Query routing, logical/semantic classification, conditional pipeline selection
Agentic & Multi-Agent Systems
Autonomous agents, multi-agent collaboration, iterative workflows
Hybrid & Ensemble Retrieval
Combining dense and sparse retrieval, ensemble methods, fusion strategies
Chunking & Document Segmentation
Semantic chunking, adaptive chunking, proposition chunking strategies
Hierarchical & Tree-Based Indexing
Tree structures, RAPTOR, hierarchical indices, multi-level navigation
Multi-Vector & Advanced Embeddings
Multi-vector representations, late interaction models, attention mechanisms
Multimodal & Visual Retrieval
Visual document retrieval, image-based search, multimodal RAG
Reranking & Post-Retrieval Filtering
Cross-encoder reranking, LLM-based scoring, adaptive K selection
Context Compression & Optimization
Prompt compression, context window optimization, information density
Verification & Self-Correction
Self-RAG, CRAG, verification loops, reflection-based quality control
Knowledge Graph & Structured Grounding
GraphRAG, knowledge graph integration, neurosymbolic retrieval
Training & Alignment Methods
Fine-tuning embeddings, retriever-generator alignment, contrastive learning
Memory & Cross-Session State
Persistent memory, caching, conversational continuity, state management
Relationship Types
Alternative Approaches
Techniques in the same category and pipeline stage that solve similar problems in different ways. These are interchangeable options — you might choose one over another based on your specific constraints or preferences.
Complementary Techniques
Techniques from different categories but the same pipeline stage. These work well together in combination — using multiple complementary techniques can enhance overall performance by addressing different aspects of the same stage.
Pipeline Chain
Techniques from different pipeline stages that naturally connect in sequence. These represent potential next steps in your RAG pipeline — the output of the current technique can feed into these downstream techniques.
Pipeline Stages
Pre-Retrieval
Techniques that prepare or transform the query before searching the knowledge base. Includes query expansion, rewriting, and routing decisions.
Retrieval Core
The main search and retrieval operations. Includes embedding-based search, keyword search, hybrid approaches, and multimodal retrieval.
Post-Retrieval
Processing and refinement of retrieved results before generation. Includes reranking, filtering, compression, and context optimization.
Generation Layer
Techniques applied during or after the LLM generates a response. Includes verification, self-correction, citation, and reasoning enhancement.
Cross-Session
Techniques that persist or leverage information across multiple interactions. Includes memory systems and feedback loops.