Updated
trovex glossary
The terms behind one idea: give your agents the single current, authoritative doc for a query instead of a pile of files to sift. Each entry is a one-sentence definition an engine can lift, then a short, concrete expansion. The only metric here is the real ~60% fewer tokens per lookup.
A canonical doc is the single current, authoritative document for a topic, the one a coding agent should treat as the source of truth when several files cover the same ground.
A freshness marker is a label attached to a document (canonical, stale, or duplicate) that tells a coding agent which copy is current, independent of how relevant the text looks.
An SSOT for agents is one shared store that every agent and teammate reads from and writes to, so the whole fleet works from the same current knowledge instead of scattered, drifting copies.
MCP is the protocol a coding agent uses to call an external tool for context: the agent sends a request, the tool returns content the agent folds into its reasoning.
A section-level read returns only the part of a document that answers a query, not the whole file.
A record is a note an agent writes back into the shared store: an incident, a decision, a note on what actually worked, so other agents and teammates can read it later instead of re-deriving it.
Re-derivation is when a coding agent works out an answer that another agent, or an earlier session, already figured out, because that knowledge was not saved anywhere it could read.
Give your agents one source of truth.
One current answer per query, with a freshness marker, in about a minute.
Open source. No cloud, no API keys. Your docs never leave your machine.