KNOWLEDGE LAB / LEARN RAG THROUGH PRACTICE
What do you want
your knowledge to do?
Choose a use case. Get a step-by-step reading plan built from practical lessons, source files, parameter experiments, and troubleshooting experience.
Start with your application.
The goal shapes your learning path. Your document format comes next.
Real practice behind every step.
Start with the fictional Harbor Studio FAQ: find its 60-minute booking rule, trace a printer citation, handle an unanswered parking question, compare Top K, and verify the updated 90-minute rule. Keep the examples beside you as you study.
Start with a one-file example Explore runnable concept labsTHE PRACTICE LIBRARY
Practice one skill at a time.
Read worked examples with source files, expected outcomes, and troubleshooting checks. Browse independently or follow your learning plan.
Six practical lessons
Build a tiny RAG with one file
Prepare a practice file, run local retrieval, and connect the returned evidence to your chosen answer model.
Show the source behind an answer
Carry source IDs through search and generation, then validate each citation against the passage.
Handle questions your documents cannot answer
Add a no-evidence path and test it beside a covered question.
Find a fact your RAG keeps missing
Trace one question from source text to index, search result, model input, and final answer.
Change just one setting: Top K
Compare one returned passage with three while keeping documents and questions fixed.
Update a document without keeping the old answer
Replace a 60-minute rule with 90 minutes and remove the old version from searchable results.
GO DEEPER WHERE YOU NEED TO
Document guides, model choices & agent settings.
Explore document preparation, model and database choices, agent behavior, and fair comparisons.
Choose embedding and answer models separately
Use the printer and parking exercises to separate retrieval failures from answer-model failures, then record a reproducible model choice.
Choose a RAG database and understand its settings
Compare a local baseline, PostgreSQL with pgvector, and a dedicated vector service using source records, version updates, and access-filter exercises.
Choose a RAG workflow or agent, then set its boundaries
Separate the answer model from its workflow, compare direct orchestration with LangGraph, and study a bounded multi-search example.
Build your first RAG knowledge base
Go from a small document set to a cited answer, with a check at every stage.
Make product FAQs answer the right question
Keep product identifiers, exceptions, and versions attached to the answer.
Keep context intact in long manuals
Compare section boundaries and parent-child retrieval without losing conditions or exceptions.
Fix scanned PDFs before tuning RAG
Verify OCR, reading order, and table structure before spending time on retrieval settings.
Know whether a RAG change actually helped
Use paired questions, source checks, and measured usage to compare two configurations.
References & optional worksheets
Use these when a lesson calls for a closer look at configuration, a failure, or a measured comparison.