Ryan Zucchi
Community@Ryanzucchi
Local-first AI platform for Portuguese-language fiction writing, combining LLM inference, knowledge graphs, and real-time collaborative rich-text editing.
Agent Skills by Ryan Zucchi
Showing 45 vetted skills indexed across 1 GitHub repositories.
Ryanzucchiotimizacao-de-modelos-generativos-locais
Configures llama.cpp inference with Q4_K_M quantization, KV caching, and asynchronous generation for local language models.
Ryanzucchiautosave-e-historico-de-versoes
Implements debounced autosave, snapshot versioning, and diff-based version restoration for editor documents.
Ryanzucchidesign-de-interacao-nao-intrusivo-para-ia
Defines UX policies for non-intrusive AI writing assistants in text editors.
Ryanzucchidesign-de-quadro-kanban-para-escrita-criativa
Designs Kanban board interfaces for novel writing planning with low-intrusion status signaling.
Ryanzucchiconcorrencia-editor-texto-rico-crdt
Implements real-time collaborative rich text editing using Yjs CRDTs and Peritext anchoring.
Ryanzucchiexportacao-multiplos-formatos
Convert ProseMirror manuscripts to DOCX, PDF, ePub, Markdown, and TXT formats.
Ryanzucchianalise-de-co-ocorrencia-e-cliches-literarios
Analyzes trope co-occurrence density against genre benchmarks and suggests optional narrative subversions.
Ryanzucchiauditoria-global-consistencia-entidades
Detects plot holes by auditing entity continuity across chapters using temporal entity graphs.
Ryanzucchivisualizacao-interativa-time-slicing
Renders character and location relationship graphs with temporal filtering and semantic clustering.
Ryanzucchipipeline-de-notificacoes-em-tempo-real
Implements real-time push notifications via WebSockets and Server-Sent Events for collaboration.
Ryanzucchiautorizacao-rbac-por-recurso
Implements role-based access control with per-resource permission checks and hierarchical inheritance.
Ryanzucchiresolucao-correferencia-e-aliases-literarios
Resolves pronoun references and character name aliases into canonical entity IDs in Portuguese fiction.
Ryanzucchidesfazer-refazer-colaborativo-seletivo
Implements distributed per-client undo/redo for selective rollback of AI-generated text in collaborative editors.
Ryanzucchidestilacao-espinha-dorsal-narrativa
Extracts dramatic story spine elements to generate loglines and commercial synopses.
Ryanzucchigeracao-perfis-psicologicos-e-roleplay
Extracts psychological character profiles from manuscripts and configures local LLMs for in-character interviews.
Ryanzucchimodelagem-de-grafo-de-metas-narrativas-gmn
Models plot causal dependencies as a DAG with cycle detection and inconsistency propagation.
Ryanzucchigrafo-de-lore-visualizacao-e-filtros
Renders and filters fictional universe entity graphs with Cytoscape.js force-directed layouts.
Ryanzucchibusca-hibrida-local-first
Implements local hybrid search combining SQLite-VSS vector retrieval and BM25 with Reciprocal Rank Fusion.
Ryanzucchianalise-consistencia-narrativa-tempo-e-eventos
Validates causal and temporal consistency of branching story timelines using event graphs.
Ryanzucchiverificacao-global-de-tropos-com-llm
Validates narrative trope candidates using a local LLM with global plot context.
Ryanzucchiparticionamento-de-modelos-de-ia-com-lora-dinamico
Implements per-tenant LLM personalization using dynamically loaded LoRA adapters on shared GPU infrastructure.
Ryanzucchimodelagem-ontologica-worldbuilding
Structures fictional universe entities into a temporal knowledge graph using quadruples.
Ryanzucchivetorizacao-e-contexto-de-escrita
Compiles writing session state into a unified context vector for local recommendation engines.
Ryanzucchisincronizacao-e-armazenamento-offline-first
Synchronizes local SQLite databases with remote servers using Delta-State CRDTs and LWW conflict resolution.
Frequently Asked Questions About Ryan Zucchi
FAQPage SchemaWhat tasks can I accomplish with Ryan Zucchi's skills?▼
You can build a local-first fiction writing platform: extract characters and relationships from Portuguese manuscripts with NER, classify narrative tropes with BERTimbau, run GraphRAG lore queries, generate hierarchical summaries, track emotional arcs, and enable real-time collaborative editing with CRDTs and selective undo/redo.
Who are these skills designed for?▼
Developers and engineers building AI-assisted creative writing software for Portuguese-language authors. The skills target full-stack engineers implementing rich-text editors, NLP pipelines, knowledge graphs, offline-first synchronization, and multi-tenant SaaS infrastructure for novelists and fiction writers.
How do the local AI inference skills work in practice?▼
Skills optimize llama.cpp inference with Q4_K_M quantization, KV cache, and async interfaces for minimal latency. They configure local LLMs for prose critique, in-character roleplay interviews, trope validation, and recursive summarization that fits long novels within limited context windows.
What are the key dependencies and prerequisites?▼
Core dependencies include llama.cpp for local inference, Yjs/Peritext CRDTs with Tiptap/ProseMirror for collaborative editing, SQLite-VSS with BM25 for hybrid search, PostgreSQL with Row-Level Security, IndexedDB for offline storage, and BERTimbau models for Portuguese text classification.
How do these skills handle data privacy and compliance?▼
Dedicated skills implement LGPD/GDPR compliance by design: permanent physical deletion, structured data portability, and immutable consent logs. Multi-tenant isolation uses PostgreSQL Row-Level Security and dynamic LoRA adapters, while JWT with rotating refresh tokens secures authentication.