30. Glossary
[!ref] Full definitions in the User Manual, "Appendix: Glossary".
| Term | Definition |
|---|---|
| Workspace | The smallest unit of execution; binds a local directory and a config set; called a "project workspace" in the UI |
| Session | One continuous conversation context, persisted locally |
| Work mode | One of five (auto / plan / prepare / execute / assess) |
| Persona | The prompt set deciding AI style and boundaries |
| Tool | A capability the AI can invoke |
| Skill | A prompt-template pack |
| Subtask | A subordinate job dispatched to an isolated sub-assistant |
| Plan / Todo | Structured task list |
| History compaction | Compresses older messages into a summary; originals are kept |
| Approval | The deny / allow-once / always-allow decision before tool execution |
| Companion messaging | Connecting the AI assistant to an IM channel |
| Plan quota | Model-call quota that comes with your account's plan, metered per period (5-hour / 7-day) |
| Error drawer | Panel aggregating recent tool / service errors |
| Binding code | 6-digit / 10-min code for IM binding |
| SBOM | Software bill of materials — the project's third-party components (CycloneDX / SPDX / DSDX / SWID) |
| VEX | Vulnerability Exploitability eXchange (OpenVEX) — whether a vuln actually affects this project |
| OSV / KEV | OSV: open-source vulnerability DB; KEV: CISA Known Exploited Vulnerabilities catalog |
| CVSS | Common Vulnerability Scoring System; v2 / v3 / v4 vectors supported |
| Sentinel | The interception layer that auto-scans file read / write / execute and blocks malicious hits; decoupled from detection capability |
| VirusTotal AI / GTI | Cloud threat-intel backends: VirusTotal AI (free-hosted) / Google Threat Intelligence (your own key) |
| Self-repair loop | The mechanism that auto-selects a repair strategy on errors (compact / switch / backoff / terminal) — §18.5 |
| AgentParallel | The subtask tool that fans one task out into 2–16 parallel legs — §18.3.1 |
| GRAI+KISS | The auto-reflect frame: GRAI analysis + KISS (Keep/Improve/Stop/Start) lesson taxonomy — §18.6 |
| Whole-repo RAG | code.search / code.ask: non-vector (BM25 + identifier tokenization + CJK bigram tokenization, zero-embedding) symbol-level whole-repo search / Q&A (§11.12.3) |
| Restate intent | The AI restates the goal / plan for confirmation before starting the session's first message (intent_confirm, default on) |
| Trace | The step-level replay of one run: timeline + aggregate metrics + per-tool metrics, splittable by run and drillable into subagents (§19.5) |
| Run | The unit a trace is split by = the whole execution triggered by one user message |
| Local mode | A session-level switch that disables the built-in internet tools (shell and external MCP network access are not blocked) — §19.4 |
| Self-reported pause | The model emitting <todo-status>paused</todo-status> to signal it stopped as you asked, which suppresses the todo check (§16.2.1) |
| Image input | Sending images with a message to a model marked as accepting image input; images go out only with their own turn (§6.8) |
| Model routing | Choosing a model per scenario (subtasks / routines / hooks / sec.decompile / code.ask / memory extraction); local mode always uses the session's model (§6.8.1) |
| Sandbox attachment | Letting a session reach directories / files outside the workspace; the scope is either "this session only" or "project-workspace-wide" |
| Live guidance | Messages added while running, which go into the "Pending" queue; step_end (Prioritize) / interrupt (Handle now) (§19.6) |
| Marked messages | Message bookmarks recorded in the session YAML marked_messages, which can be located / exported (§19.7) |
| Auto-repair | The strategy of switching to another provider with the same-named model when the selected model is unavailable or the upstream fails (§6.8.1) |
| Effort | Landi N2.7's reasoning_effort level: none / low / high / xhigh / max (§19.8) |
| Repetition spin | Repeated body text without tool calls; cut off within a turn, halted across turns (§19.10) |
| Session routine | A timed prompt the model sets inside a session with RoutineCreate, stored in session-routines.yaml (§18.7) |
License: AVL Code Proprietary Software License — Copyright © 2024–2026 Antiy. All Rights Reserved. Document version: 2026-10-15, matching app version v0.10.15-alpha.1442.