THE RECALL PLAYGROUND
Search. Rank.
See what changes.
Explore retrieval through small, focused experiments. Change the inputs and inspect indexes, rankings, metrics, and every stage of a RAG pipeline.
One question, four ways of searching
How Search Systems Evolved
Interactive experimentAhead of time or during the search?
How Search Systems Evolved
Interactive experimentShould these count as the same word?
Boolean Search and Inverted Indexes
Interactive experimentLook up words and combine them
Boolean Search and Inverted Indexes
Interactive experimentStrong clue or weak clue?
Ranking with TF-IDF and BM25
Interactive experimentDoes repeating a word help?
Ranking with TF-IDF and BM25
Interactive experimentMark the results, watch the scores
Measuring Retrieval Quality
Interactive experimentWhich score fits the job?
Measuring Retrieval Quality
Interactive experimentMissed page or wrong page?
Where Lexical Search Falls Short
Interactive experimentClose the word gap by hand
Where Lexical Search Falls Short
Interactive experimentThe same question, four ways
Why LLMs Need External Knowledge
Interactive experimentFind the unsupported claims
Why LLMs Need External Knowledge
Interactive experimentPick the starting approach
RAG, Prompting, Fine-tuning or Long Context
Interactive experimentFollow a question through RAG, then break it
Anatomy of a RAG System
Interactive experimentWhich stage would you check first?
Anatomy of a RAG System
Interactive experimentFind the nearest meaning
What Embeddings Are
Interactive experimentMatch the idea to what it does
How Embedding Models Learn Meaning
Interactive experimentAngle, length and distance
Similarity Metrics
Interactive experimentSame query, three models
Modern Embedding Models
Interactive experimentWhat will embeddings cost?
Choosing an Embedding Model
Interactive experimentPick a starting model
Choosing an Embedding Model
Interactive experimentOne Telugu question, three approaches
Multilingual and Indic Retrieval
Interactive experimentWhat kind of query is this?
Multilingual and Indic Retrieval
Interactive experimentWhich method wins this query?
Sparse vs Dense Retrieval
Interactive experimentHow much work is one brute-force search?
Building Semantic Search from Scratch
Interactive experimentWhich format causes this problem?
Enterprise Documents and Their Formats
Interactive experimentThree ways to read the same PDF page
Parsing Documents into Markdown
Interactive experimentThe same scanned circular, three ways
OCR for Scanned Documents
Interactive experimentPick the fix
OCR for Scanned Documents
Interactive experimentThree versions of the same fee table
Extracting Tables, Forms and Figures
Interactive experimentHow would you store this?
Extracting Tables, Forms and Figures
Interactive experimentClean the page without losing the facts
Cleaning and Normalizing Text
Interactive experimentClean it or keep it?
Cleaning and Normalizing Text
Interactive experimentMask the identifiers, keep the request
Handling Sensitive Data at Ingestion
Interactive experimentDrop, mask or restrict?
Handling Sensitive Data at Ingestion
Interactive experimentOne night of incremental sync
Document Connectors
Interactive experimentHow should we sync this source?
Document Connectors
Interactive experimentFour files through the same pipeline
Building an Ingestion Pipeline
Interactive experimentHow long will a full run take?
Building an Ingestion Pipeline
Interactive experimentWhere does this document belong?
Designing a Knowledge Base
Interactive experimentFind a chunk size that keeps the answer together
Chunk Size and Overlap
Interactive experimentFixed size, sentences or headings?
Recursive and Structure-aware Chunking
Interactive experimentWhere should the cuts go?
Semantic Chunking
Interactive experimentThe same messy notice, chunked three ways
Semantic Chunking
Interactive experimentWhat does the model get to read?
Parent-Child Chunking
Interactive experimentWhich fix fits?
Late Chunking
Interactive experimentSame search, with and without metadata
Metadata Engineering
Interactive experimentWhat does this change require?
Keeping the Index Fresh
Interactive experimentFull rebuild or incremental?
Keeping the Index Fresh
Interactive experimentHow much work is one brute-force search?
Why Vector Databases Exist
Interactive experimentKeyword SQL versus vector search on the same table
Why Vector Databases Exist
Interactive experimentMatrix, library or database?
Why Vector Databases Exist
Interactive experimentSearch fewer buckets, find most neighbours
Exact vs Approximate Nearest Neighbour Search
Interactive experimentClusters and probes
ANN Index Types
Interactive experimentHop towards the query
ANN Index Types
Interactive experimentWhich index would you start with?
ANN Index Types
Interactive experimentFlat, IVF and HNSW on one million vectors
Vector Search with FAISS
Interactive experimentWhich operator?
PostgreSQL + pgvector
Interactive experimentIndex or no index?
PostgreSQL + pgvector
Interactive experimentWhere should this field live?
Schema Design for Vector Storage
Interactive experimentWhat ends up in the store?
Vector Database Operations
Interactive experimentWhich operation does this need?
Vector Database Operations
Interactive experimentHow many passages reach the model?
Tuning Retrieval Parameters
Interactive experimentRelevance or variety?
Tuning Retrieval Parameters
Interactive experimentFuse two rankings with RRF
Hybrid Search and Rank Fusion
Interactive experimentWhich method rescues this query?
Hybrid Search and Rank Fusion
Interactive experimentSame question, different filters
Filtered Vector Search
Interactive experimentFilter before or after the search?
Filtered Vector Search
Interactive experimentDoes compression change the ranking?
Vector Compression
Interactive experimentHow much memory do the vectors need?
Vector Compression
Interactive experimentPick a vector store
Choosing a Vector Database
Interactive experimentRetrieve, then build the prompt
Building a Naïve RAG Pipeline
Interactive experimentWhich step failed?
Building a Naïve RAG Pipeline
Interactive experimentWhich component do you change?
RAG with LangChain
Interactive experimentPack the prompt
Constructing the Context and Prompt
Interactive experimentCheck the citations
Citations and Grounded Answers
Interactive experimentThree ways to improve one query
Query Rewriting, Expansion and HyDE
Interactive experimentWhich technique would help most?
Query Rewriting, Expansion and HyDE
Interactive experimentPlain search, multi-query or self-query?
Multi-query and Self-query Retrieval
Interactive experimentMulti-query, self-query or neither?
Multi-query and Self-query Retrieval
Interactive experimentDoes reranking change what the model sees?
Reranking Retrieved Results
Interactive experimentSearching a follow-up question
Conversational RAG
Interactive experimentDoes this message need rewriting?
Conversational RAG
Interactive experimentRoute the question
Query Routing
Interactive experimentAn agent answers a multi-part question
Agentic RAG
Interactive experimentFixed pipeline or agent?
Agentic RAG
Interactive experimentFive hard questions for a naïve RAG system
Where Naïve RAG Breaks
Interactive experimentName the failure pattern
Where Naïve RAG Breaks
Interactive experimentWhat does adding context cost?
Contextual Retrieval
Interactive experimentWould contextual retrieval help here?
Contextual Retrieval
Interactive experimentWalk the graph
Knowledge Graphs for Retrieval
Interactive experimentVector search or graph walk?
Knowledge Graphs for Retrieval
Interactive experimentA global question with GraphRAG
GraphRAG
Interactive experimentLocal search, global search or plain vector RAG?
GraphRAG
Interactive experimentHow much bigger is a multi-vector index?
Multi-vector Retrieval
Interactive experimentWould multi-vector retrieval help?
Multi-vector Retrieval
Interactive experimentText search or SQL?
Retrieval over Structured Data
Interactive experimentWhere should this question go?
Retrieval over Structured Data
Interactive experimentLong context or RAG: tokens and cost
Long Context vs RAG
Interactive experimentLong context, RAG or both?
Long Context vs RAG
Interactive experimentMatch the idea to the problem
Emerging Retrieval Paradigms
Interactive experimentRecommend an architecture
Choosing a Retrieval Architecture
Interactive experimentText-only or multimodal?
Multimodal Enterprise Knowledge
Interactive experimentOne annual report page, three parsers
Vision-Language Document Parsing
Interactive experimentOCR or vision-language model?
Vision-Language Document Parsing
Interactive experimentImages and text in one space
Multimodal Embeddings
Interactive experimentGood fit for a CLIP-style model?
Multimodal Embeddings
Interactive experimentFinding an image three ways
Cross-modal Retrieval
Interactive experimentWhich direction is this search?
Cross-modal Retrieval
Interactive experimentAnswering from a page image
Page-image Retrieval
Interactive experimentHow big is a page-image index?
Page-image Retrieval
Interactive experimentText extraction or page images?
Page-image Retrieval
Interactive experimentHow should this be stored?
Multimodal Knowledge Base Design
Interactive experimentOne question, three answer pipelines
Multimodal RAG
Interactive experimentWhat do images add to a prompt?
Multimodal RAG
Interactive experimentHow long should a transcript chunk be?
Audio and Video Retrieval
Interactive experimentHow would you make this searchable?
Audio and Video Retrieval
Interactive experimentFind the first broken stage
How RAG Systems Fail
Interactive experimentFour deliberately broken pipelines
How RAG Systems Fail
Interactive experimentGood test item or weak one?
Building an Evaluation Dataset
Interactive experimentScore one configuration's results
Benchmarking Retrieval Configurations
Interactive experimentWhat does this benchmark result tell you?
Benchmarking Retrieval Configurations
Interactive experimentJudge faithfulness claim by claim
Evaluating Generated Answers
Interactive experimentWhere should the judge's passing mark be?
Evaluation Frameworks
Interactive experimentWhich tool fits?
Evaluation Frameworks
Interactive experimentSame chunks, three embedding set-ups
Debugging Chunking and Embeddings
Interactive experimentChunking or embedding?
Debugging Chunking and Embeddings
Interactive experimentIs it a ranking problem?
Debugging Retrieval
Interactive experimentWhich fix matches the log?
Debugging Retrieval
Interactive experimentTrace each sentence to its evidence
Debugging Generation
Interactive experimentName the generation failure
Debugging Generation
Interactive experimentLabel the failure
Error Analysis
Interactive experimentWhere should we spend the next fortnight?
Error Analysis
Interactive experimentThree changes, three gate reports
Regression Testing for RAG
Interactive experimentExplore a trace
Tracing and Observability
Interactive experimentWhat does this signal suggest?
Monitoring in Production
Interactive experimentFit the evidence into the budget
Context Compression and Token Budgeting
Interactive experimentDoes the index fit in memory?
Performance Tuning for Vector Search
Interactive experimentWhich setting would you change?
Performance Tuning for Vector Search
Interactive experimentHow long does the first full index take?
Scaling to Millions of Documents
Interactive experimentShards, replicas or workers?
Scaling to Millions of Documents
Interactive experimentSet the cache threshold
Caching for RAG
Interactive experimentWhat does a response cache save?
Caching for RAG
Interactive experimentMonthly cost estimate
Cost Engineering
Interactive experimentWhich lever fits?
Cost Engineering
Interactive experimentChoose a deployment
Deployment Architectures
Interactive experimentWho sees what?
Permission-aware Retrieval
Interactive experimentSafe or unsafe?
Permission-aware Retrieval
Interactive experimentSpot the injected instructions
Securing RAG Applications
Interactive experimentMatch the risk to the defence
Securing RAG Applications
Interactive experimentLaunch, or not yet?
Production Readiness
Interactive experimentNo experiments found.
Try a shorter search or choose another unit.