# Glossary | Term | Definition | | -------------------- | --------------------------------------------------------------------------------------------------------------------------------------- | | **Auto-approve** | Routing outcome where the policy ships the AI suggestion without human review. | | **Auto-threshold** | Confidence value at or above which (and no risk flags) an item is auto-approved. | | **Calibration** | The property that a model's confidence scores match its actual accuracy. A 0.9-confidence prediction should be right \~90% of the time. | | **Decision** | Terminal state of an item: `approved`, `rejected`, or `escalated_final`. | | **Edit-and-approve** | Reviewer action: modify the AI suggestion, then approve. Captured as an override. | | **Escalate** | Routing outcome (or reviewer action) that sends an item to a senior reviewer. | | **Guidelines** | Project-level rules shown to reviewers when deciding items. | | **Hard-block flag** | A risk flag that triggers immediate rejection with no human involvement. | | **HITL** | Human-in-the-loop. In LoopDesk specifically: humans on the exception path, not the hot path. | | **Hot path** | The default execution path of an AI pipeline. "Human on the hot path" means every action waits for a human. | | **Exception path** | The path taken only when policy can't auto-decide. "Human on the exception path" means humans only see items that need them. | | **Input** | The original request that produced the AI suggestion. Stored as `jsonb`. | | **Item** | A single AI decision flowing through LoopDesk. | | **Learning** | A captured override, with full context, used as training signal. | | **Override** | A reviewer's decision that differs from the AI suggestion. | | **Policy** | The per-project config (thresholds, flags, reviewer pool, etc.) that drives routing. | | **Project** | A logical grouping of items sharing the same policy. | | **Queue** | Items awaiting human review. | | **Reviewer** | A user with the `reviewer` role who acts on queued items. | | **Reviewer pool** | The set of reviewers allowed to pick up a given project's items. | | **Risk flag** | A short namespaced string attached to an item that influences routing. | | **Senior reviewer** | A reviewer (often admin) who handles escalated items. | | **SIWS** | Sign-In-With-Solana. Wallet-based authentication using signed messages. | | **Suggestion** | The AI's proposed answer for an item. | | **Webhook** | HTTP callback. Outbound (LoopDesk → you on decision) or inbound (external → LoopDesk). | | **$HITL** | The token whose balance gates access to `/payments` routes. | --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://hitl-01.gitbook.io/hitl-docs/reference/glossary.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.