A governed memory layer
EDN sits between your application and the language model you choose. Your application writes content into EDN and asks EDN for context. EDN stores that content, classifies it by where it came from, and returns a compact, governed package of context that your application places in front of its model. EDN does not call a model to answer questions and does not route to a provider. The model is yours. EDN’s job is to make the context that reaches your model honest and traceable.What “governed” means here
Every record EDN holds carries a provenance class that records its origin, for example a fact the user stated, a fact confirmed by the user, a fact verified from a document, or an AI inference. Retrieval and context assembly respect those classes: inferences are never presented as established facts, and AI-generated content is kept out of the sections your model treats as ground truth.What EDN gives your application
Structured persistence
Records are typed and classified, not free text blobs.
Deterministic retrieval
Context is selected by a fixed similarity query, not by a model.
Typed corrections
A dedicated channel to correct what memory holds.
Provenance and citations
Every returned item can be traced to its origin.
Next
How EDN works
The path from content in to governed context out.
Core concepts
Provenance, retrieval, abstention, and the SIP.