Concepts
How the platform actually works
These pages explain rather than instruct. Each one takes a single part of the platform, starts from something familiar, and then says plainly where the comparison stops being true. Read them when you want to understand a guarantee, not when you want a command.
Turning documents into answers
What happens between uploading a file and getting a reply you can trust.
How your documents become searchableParsing, chunking, and what the node stores once a file arrives.How your node searches by meaningWhy a question finds the right passage even when it shares no words with it.How hybrid search finds the right passageMeaning and exact terms run together, and the reranker reads the candidates.How every answer shows its sourcesProvenance from the chunk to the claim, so an answer can be checked.How your node keeps answers accurateWhat grounding actually constrains, and where it stops.Why your search has no meter runningRetrieval runs on your node, so asking more questions costs nothing extra.
Where the work happens
Which machine holds your data, and which one does the thinking.
Your data lives on your node, and so does the reading of itThe containment boundary, stated as what the control plane can and cannot reach.The model that answers you lives in your nodeGeneration on the node itself, and what that removes from the path.Bring your own model, and keep it localPointing a node at a model runtime you already run.How your node remembersWhat persists between conversations, and where it is kept.
Nodes working together
How separate nodes answer one question without pooling what they hold.
Evidence a reviewer can check
The parts you point at when someone asks you to prove it.
Next: Platform Status lists which of these capabilities each tier includes today, what is still in beta, and what is on the roadmap.