Skip to content

Glossary

Definitions of legal, technical, and domain-specific terms used throughout Juca and the sens.legal ecosystem.

TermDefinition
AcordaoAppellate court decision — the primary document type in Juca’s corpus. Each acordao has an ementa (summary), tese (thesis), and full text (integra).
Dispositivo LegalA specific legal provision or article from legislation (e.g., Art. 927 do Codigo Civil).
EmentaSummary or headnote of a court decision. Used as the primary search field in Juca’s hybrid search.
IntegraFull text of a court decision. Accessed via SlideOver panel or exported to PDF.
IRACIssue, Rule, Application, Conclusion — a structured legal analysis framework used by the Ratio tool for extraction.
JurisprudenciaCase law — the body of court decisions that forms binding or persuasive precedent.
MinistroMinister (judge) of the STJ. Juca tracks which ministro authored each decision for divergence analysis.
ParecerFormal legal opinion — one of the 4 delivery modes in the Briefing Progressivo.
PrecedenteA previous court decision used as authority to support a legal argument.
Ratio DecidendiThe core legal reasoning or principle behind a decision — what makes it a precedent.
STJSuperior Tribunal de Justica — Brazil’s highest court for non-constitutional federal matters. Juca’s corpus contains 23,400+ STJ decisions.
SumulaBinding summary of consolidated jurisprudence issued by the STJ. Validated by Juca’s anti-hallucination system.
TeseLegal thesis or argument extracted from a decision. Searchable field in Juca’s search engine.
TurmaDivision or panel of the STJ that decided a case (e.g., 1a Turma, 2a Turma).
TermDefinition
BlockTyped UI content unit — the fundamental building block of Juca’s interface. 11 types: message, progress, diagnosis, action_prompt, precedent, precedent_picker, summary, risk_balance, chart, delivery, exit_card.
Block FactoryPure functions that create typed block data with defaults. Located in src/lib/blocks/types.ts (e.g., createDiagnosisData(), createRiskBalanceData()).
Briefing ProgressivoJuca’s 4-phase progressive disclosure system: Diagnosis → Precedents → Risks → Delivery. Each phase reveals information incrementally based on user interaction.
ComposerInput component where users type queries and submit them. Supports tool hints and pipeline mode selection.
G,C,R PipelineGenerate → Criticize → Revise — Juca’s multi-LLM processing pattern where a generator creates content, critics evaluate it, and a revisor produces the final output.
HubJuca’s architectural role: a lightweight frontend that orchestrates backend agents (Valter, Leci) rather than processing data locally.
Liquid LegalThe design language for Juca’s UI. Palette: bg-paper (#F0F0EE), ink-primary (#1A161F), accent (#FFF06D).
PhaseRailVisual navigation component showing briefing phase progress (4 phases as a vertical rail).
SlideOverSliding panel for viewing full decision text (integra) without leaving the current context.
Tool RegistryPattern that routes user intents to specialized handler tools. 5 tools registered with priority-based selection: analyzer (9), juris (8), ratio (7), compare (5), insights (4).
WorkCanvasMain content area that renders blocks in sequence. The primary viewport of the Unified Home interface.
TermDefinition
sens.legalThe parent platform encompassing Juca, Valter, and Leci — an AI-powered legal analysis ecosystem.
ValterBackend agent for STJ jurisprudence. Python/FastAPI service with 23,400+ decisions, 28 MCP tools, BM25 + semantic search, Neo4j knowledge graph, and multi-LLM pipeline. Production URL: https://valter-api-production.up.railway.app.
LeciBackend agent for federal legislation. TypeScript/Next.js with Drizzle ORM. Currently in early development (v0.1-pre) with database schema ready but no public API yet.
MCPModel Context Protocol — standardized interface for AI tool interaction. Valter exposes 28 MCP tools for search, verification, graph analysis, and ingestion.
Adapter LayerPlanned abstraction (v0.3) that provides a unified interface for Juca to communicate with any backend agent (Valter, Leci, future agents).