Focus Area: LLM-oriented AI search workflows, including query planning, retrieval orchestration, result ranking, provenance signaling, and grounded answer assembly across agentic systems.
This ontology provides citation-quality definitions for 15 foundational terms, backed by authoritative sources from standards bodies (NIST, W3C, IETF, OASIS, ISO) and peer-reviewed research.
15
Technical Terms
75%+
Tier-1 Sources
V1.72
Pipeline Version
Technical Glossary
AGT001Query Planning Envelope
captures a search objective as a structured instruction that separates the user’s ask from the sub-queries, tools, and constraints needed to resolve it. It matters because agentic search only stays coherent when the initial intent is decomposed into machine-tractable pieces before retrieval begins.
defines the ordered path through indexes, APIs, tools, and decision points that an agent follows while assembling evidence for an answer. The graph matters because retrieval quality is shaped not just by sources, but by how calls are sequenced, gated, and recombined.
aligns a query to the vocabulary, entity labels, and conceptual neighborhoods that make hidden but relevant material retrievable. It matters because LLM search fails when lexical matching outruns the semantic relationships that actually connect the evidence.
allocates limited prompt and retrieval capacity across evidence, instructions, and memory so the system does not drown useful material in low-value context. It matters because grounded search quality is constrained as much by what is excluded as by what is included.
chooses the specific passages or objects that will represent a source inside the answer-generation workflow. It matters because citation-grade output depends on selecting chunks that preserve meaning, scope, and traceability rather than arbitrary slices of text.
attaches origin, transformation, and trust metadata to each evidence item as it moves through the search pipeline. It matters because answer grounding weakens the moment supporting material can no longer be traced back to a source and handling path.
generates an answer in a way that preserves the attachment between claims and the sources that justify them. It matters because search systems become auditably useful only when synthesis is coupled to citation instead of replacing it.
reorders retrieved candidates after initial recall using stronger signals about usefulness, authority, and task fit. It matters because first-pass retrieval often surfaces plausibly related material that still needs a second judgment before answer assembly.
tests whether the draft answer is actually supported by retrieved evidence rather than merely plausible to the model. It matters because grounded search requires a final verification step that explicitly checks claims against the retrieved record.
revises an initial search instruction when the first evidence set is sparse, ambiguous, or misaligned with the task. It matters because effective search agents improve retrieval by iterating on the question, not just reusing the first wording.
weights sources according to identity, integrity, policy fit, and prior reliability before they influence the final answer. It matters because agentic search should not treat every retrievable object as equally admissible evidence.
calls retrieval tools, APIs, or indexes as explicit search actions rather than hiding them inside one opaque model pass. It matters because search becomes governable when the retrieval step is observable, permissioned, and separable from generation.
checks whether a retrieved result is timely enough for the task before it is surfaced or cited. It matters because old but authoritative material can still be misleading when freshness is itself part of relevance.
steps through competing entity meanings or query interpretations until the system can commit to the intended referent with evidence. It matters because many search failures begin as unresolved ambiguity rather than missing data.
retains the evolving state of a multi-step search interaction so later actions can build on earlier retrievals and decisions. It matters because agentic search becomes brittle when each turn forgets prior evidence, rejected paths, and established constraints.