Vector Quest

Tracks

1. 01🧭Why Vector Search? Foundations

   0/6 lessons

   Meaning as geometry, not as keywords

   Build the mental model: text becomes vectors, vectors live in a space, and similarity is measured by angle and distance. By the end of this track you can write semantic search by hand with NumPy and explain why your distance metric matters.

   Lesson list (6)

   • 01
     Where Keyword Search Breaks~22 min · embeddings, search, intuition
   • 02
     Embeddings: Text Becomes Numbers~24 min · embeddings, models, intuition
   • 03
     Cosine, Euclidean, Dot Product~22 min · math, metrics
   • 04
     Picking an Embedding Model in 2026~26 min · models, selection
   • 05
     Tokens, Context Limits, and Truncation~20 min · tokens, models, gotchas
   • 06
     Build Semantic Search by Hand~28 min · practice, numpy

   ✓Quiz · 4 questions→

2. 02🧩Chunking and Ingestion

   0/5 lessons

   How you slice the corpus decides what you can find

   Chunking is the most underrated lever in RAG quality. Pick the wrong chunk size or splitter and the best embedding model in the world cannot save you. This track covers token-aware splitters, structure-aware splitters, and the metadata that turns a chunk into a source.

   Lesson list (5)

   • 01
     Why Chunk at All~18 min · chunking, intuition
   • 02
     Tokens vs Characters: Stop Slicing Blind~18 min · chunking, tokens
   • 03
     Structure-Aware: Markdown, HTML, Code~22 min · chunking, markdown, code
   • 04
     Metadata That Earns Its Keep~20 min · chunking, metadata
   • 05
     Semantic Splitting and When to Skip It~22 min · chunking, semantic, advanced

   ✓Quiz · 4 questions→

3. 03🌈ChromaDB in Practice

   0/5 lessons

   The vector store cwkPippa actually runs

   Chroma is the easiest serious vector store to bring up — pip install, persist to disk, done. This track covers the operations you will run a hundred times: create, add, query, filter, update, delete, and back up.

   Lesson list (5)

   • 01
     PersistentClient and Collections~18 min · chroma, setup
   • 02
     Add, Query, and the Three Result Modes~22 min · chroma, queries
   • 03
     Metadata Filtering with Where Clauses~20 min · chroma, filtering
   • 04
     Updating, Deleting, and Re-embedding~20 min · chroma, lifecycle
   • 05
     Backup, Restore, and Disaster Drills~18 min · chroma, operations

   ✓Quiz · 4 questions→

4. 04🐘pgvector and the SQL Hybrid

   0/5 lessons

   When your vectors should live next to your relational data

   pgvector turns Postgres into a credible vector store. The win is locality — vectors next to the rows they describe, transactions, joins, and one operational target instead of two. This track covers installation, indexes, hybrid queries, and the safe migration path.

   Lesson list (5)

   • 01
     When SQL Beats a Pure Vector DB~20 min · pgvector, architecture
   • 02
     Install pgvector and Define a Vector Column~18 min · pgvector, setup, sql
   • 03
     HNSW vs IVFFlat: Picking the Index~22 min · pgvector, indexes, performance
   • 04
     Hybrid: WHERE Filters + Vector Ranking~22 min · pgvector, hybrid, sql
   • 05
     Migrations, Re-embedding, and Zero Downtime~20 min · pgvector, operations, migrations

   ✓Quiz · 4 questions→

5. 05🔀Hybrid Search and Reranking

   0/5 lessons

   Vector + keyword + cross-encoder, in that order

   Pure vector search misses exact-match queries (product codes, error strings, names). Pure keyword search misses paraphrases. Hybrid systems combine both and add a reranker on top. This track shows you how to assemble the stack — and how to measure that it actually helps.

   Lesson list (5)

   • 01
     BM25 Isn't Dead — Pair It With Vectors~22 min · hybrid, bm25, keyword
   • 02
     Reciprocal Rank Fusion~18 min · hybrid, fusion
   • 03
     Cross-Encoder Reranking~22 min · hybrid, reranking
   • 04
     Evaluating and Tuning the Stack~22 min · hybrid, evaluation
   • 05
     Query-Side Tricks: HyDE and Multi-Query~20 min · hybrid, queries, advanced

   ✓Quiz · 4 questions→

6. 06📚RAG System Design

   0/5 lessons

   From retrieval results to a grounded LLM answer

   Retrieval-Augmented Generation is the dominant pattern for building LLM apps over private data. This track covers the contract between retrieval and generation, prompt assembly, citations, and the cases where RAG is the wrong answer.

   Lesson list (5)

   • 01
     The Retrieval Contract~20 min · rag, design
   • 02
     Prompt Assembly: Stuff, Map-Reduce, Refine~22 min · rag, prompts
   • 03
     Citations and Provenance That Hold Up~22 min · rag, citations
   • 04
     When RAG Is the Wrong Answer~18 min · rag, design, tradeoffs
   • 05
     Feedback Loops and Self-Correcting Retrieval~22 min · rag, advanced

   ✓Quiz · 4 questions→

7. 07⚙️Production Operations

   0/5 lessons

   Cost, latency, versioning, and the metrics that catch silent decay

   Once retrieval ships, it stops being a model problem and becomes an ops problem. This track covers cost control, latency engineering, embedding versioning, observability, and tenant isolation — the disciplines that decide whether your retrieval still works in six months.

   Lesson list (5)

   • 01
     Cost: Embedding Budgets and Call Patterns~22 min · ops, cost
   • 02
     Latency: Batching, Async, and Where the Time Goes~22 min · ops, latency, performance
   • 03
     Versioning Embeddings Without Breaking Production~22 min · ops, migrations
   • 04
     Observability: Watching for Silent Decay~22 min · ops, monitoring
   • 05
     Multi-Tenant Isolation~20 min · ops, multi-tenant, security

   ✓Quiz · 4 questions→

8. 08🧠Memory Systems — Pippa's Recall

   0/5 lessons

   From single-turn chat to a stateful AI with long-term memory

   Vector retrieval is the substrate underneath every long-term AI memory system. This track covers the design choices behind episodic vs semantic memory, summary memory, decay, and closes with a case study of how cwkPippa actually does it.

   Lesson list (5)

   • 01
     Episodic vs Semantic Memory~20 min · memory, design
   • 02
     Conversation Summary Memory~22 min · memory, summarization
   • 03
     Embedding-Based Recall and Memory Indexing~22 min · memory, retrieval
   • 04
     Decay, Pruning, and Forgetting on Purpose~18 min · memory, operations
   • 05
     Case Study: cwkPippa's Memory Stack~28 min · memory, case-study, cwkpippa

   ✓Quiz · 4 questions→