Glossary
Plain definitions of the terms used across the Apiary suite. For terms specific to one product, see that product's own pages.
Glossary
Plain definitions of the terms used across the Apiary suite. For terms specific to one product, see that product's own pages.
Related:
#The suite
- The Apiary: the whole stack of small programs that give your AI coding agents one shared, lasting memory on hardware you control.
- Honeycomb: the memory. Captures what happens as you work, distills it, and hands it back to any assistant.
- Nectar: the understanding layer. Describes every file in plain language so code can be found by meaning.
- Hive: the one portal you open. A single dashboard for the whole stack.
- Doctor: the watchdog. Keeps the local daemons alive and heals what breaks.
- Queen: the cloud orchestrator for a multi-machine fleet.
#Core ideas
- Daemon: a small always-on background helper. In the Apiary, the daemons are the only things that talk to your memory store.
- The fleet: the set of Apiary daemons, on one machine or across many.
- Capture: quietly recording prompts, tool actions, and results as you work.
- Distill: turning long, noisy transcripts into short, useful notes.
- Recall: asking for the right notes back, automatically at the start of a session and on demand during work.
- Memory: a clean, distilled note: a decision, a fix, a convention, a gotcha.
- The three tiers: the same memory at three levels of detail: a one-line key, a short summary, and the full raw session.
- Priming: the small briefing handed to the assistant at the start of a session.
- Skill: a reusable lesson, written once and shared with a team automatically.
- The pollinating loop: Honeycomb's periodic self-tidying pass: merge duplicates, prune junk, supersede stale facts.
#Codebase understanding
- Nectar (the identity record): a file's small, stable identity tag that survives renames and moves.
- Description: a short plain-language note of what a file does.
- Brooding: Nectar's first read-through of a codebase, a one-time cost.
- Projection: the committed map of Nectar's understanding, which teammates inherit for free.
#Storage and models
- Deep Lake: Activeloop's database for AI, where memory lives. It does both exact lookups and meaning-based search in one engine, with full version history.
- Hivemind: Activeloop's open-source agent-memory project that Honeycomb is built on.
- Embeddings, semantic search: finding things by meaning rather than exact words.
- Hybrid recall: fusing keyword search with meaning-based search into one ranked result.
#Tenancy and health
- Org, workspace, and project: company, team, and repo-or-folder: the three levels that keep data separated.
- Repair ladder: Doctor's escalating sequence for healing a sick daemon.
- Blessed release: a version explicitly approved before Doctor will roll it out.
- Needs-attention report: the plain-language report Doctor writes when it cannot fix something itself.
- Per-agent identity: in a fleet, every agent's own attributable, revocable identity, with no shared forever-key.
#Standards
- MCP: the Model Context Protocol, a standard way for AI tools to call external helpers.
- AGPL: the GNU Affero General Public License, the license the Apiary is released under.