| MAESTRO | The whole system — generate, run, and evolve reasoning chains. You use it via the MAESTRO CARE TUI or the care CLI. |
| MAESTRO CARE | The Textual TUI — the terminal interface to MAESTRO (the care command launches it). |
| CARL | Collaborative Agent Reasoning Library — the format reasoning chains are written in (and the library that runs them). See the CARL docs. |
| MAGE | The generator that turns a task into a CARL chain. |
| GigaEvo Memory | The store for saved entities (chains, agents, skills, run records) — the Library. |
| GigaEvo Platform | The evolution service (genetic search over chains). |
| Chain | A CARL reasoning chain — the unit MAGE generates and CARL executes. |
chain_id | The stable id a chain is saved under in Memory. |
| Agent | A saved, named chain in the Library you return to and re-run. |
| AgentSkill | A portable skill folder (SKILL.md + scripts) a chain can run. See Capabilities. |
memory_card | A saved note / run digest entity in Memory. |
| Channel | A version channel for a saved chain; latest is the default care run reads. |
| Lineage | A chain’s ancestry DAG (parents it was derived/evolved from) — see care lineage. |
| Fitness | An evolved chain’s score on the dataset; higher wins. |
| Ad-Hoc mode | Generate + run inline; nothing is saved. |
| Interactive mode | The default chat surface — generate a chain, run it on the spot, answer inline; save / evolve are your call (button-driven). The legacy ad_hoc name normalises to this. See Chat surfaces. |
| Production mode | Generate a reproducible chain, save it to Memory, seed a dataset, optionally evolve. |
| Per-stage policy | How each pipeline stage behaves: ask (confirm gate first), auto (do it silently), or skip (never, and skip dependents). Set per surface via CARE_CHAT__MODE__<MODE>__<STAGE>. See Stage policy. |
| Soft config parsing | A malformed or legacy default_mode never crashes the config load: known aliases are normalised, anything unrecognised falls back to the default (interactive) with a logged warning. |
| Version | An immutable point in a saved entity’s history. Channels (latest, stable) are named pointers into that history — see care versions. |
rollback | Repoint a channel at a specific version — a non-destructive pin, not a revert (care rollback, TUI /rollback). |
promote | Copy one channel pointer to another (default latest → stable) — care promote. The TUI /promote layers a baseline/eval gate on top. |
| Library bundle | A portable tarball packing saved chains (+ AgentSkills) for transfer to another machine — care export / care import. |
| Dataset (eval case) | A chain’s set of task → expected test cases used to baseline and score it — care dataset list/add/run/export. |
| Long-term memory (LTM) | A per-session notes store that persists facts and preferences across runs (care remember / care notes), separate from the entity Library. |
| Agent hub | A lightweight local process that serves a deployed chain as an HTTP agent with its own Swagger UI — care deploy / deployments / metrics. See Deploy. |
| DAG modal | The full-screen chain view (F) that puts the box-and-arrow graph, step list, and details side by side; under /revise it overlays the new version’s diff. See Reading the DAG. |
| Latency heat-map | A DAG overlay that shades each step box by a run metric (latency / tokens / cost) on a low→high scale. |
| Pareto front | The table of non-dominated individuals from an evolution run; with ≥ 2 objectives a 2D scatter sits beneath it. |
| Evolution dashboard | The TUI screen (/evolution) that lists active + recent runs and streams a live Fitness chart, Pareto front, Programs, and Versions. See Evolution. |
| Preflight | Static validation of a chain before execution (care validate / care run). |
| Catalog | The list of installed capabilities (skills / MCP / tools / cards) — Ctrl+K or care catalog. |