Skip to main content

AI Pipeline Cycle

Watch the full RAG lifecycle: embed → retrieve → rank → context → generate → stream. generate typically dominates the waterfall — that's the teaching moment.

Pipeline

order & status
CLIENTSERVERBrowser — client-side form submission or fetch callBrowserNetwork — HTTP round-trip between browser and serverNetworkServer — SvelteKit hooks, validation, authServerEmbed — embed query into vector representationEmbedRetrieve — vector + graph search across tiersRetrieveRank — RRF fusion of multi-tier resultsRankContext — assemble final prompt context blockContextGenerate — LLM streaming — first token → final tokenGenerateResponse — serialization back to the clientResponse

Waterfall

relative timing
Browser
Network
Server
Embed
Retrieve
Rank
Context
Generate
Response