{ "ok": true, "schema": "minike-local-ai-capabilities-v1", "generated_at": "2026-07-12T15:42:52-03:00", "summary": { "total": 11, "by_status": { "available": 1, "available_partial": 2, "planned_optional": 6, "optional_model_required": 1, "assisted_not_guaranteed": 1 }, "ready_now": 3, "needs_install_or_model": 7, "updated_at": "2026-06-18" }, "registry": { "schema_version": 1, "updated_at": "2026-06-18", "title": "Mike Local capability registry", "truth_policy": [ "available significa que o fluxo principal ja existe no produto atual.", "available_partial significa que existe uma parte utilizavel, mas faltam dependencias, formatos ou UX para cobrir todos os casos.", "optional_model_required significa que o recurso depende de modelo, runtime ou API que o usuario precisa instalar, ativar ou conectar.", "planned_optional significa que a arquitetura esta definida, mas a instalacao/execucao ainda nao deve ser prometida como pronta.", "A Mike nao deve afirmar que domina todas as linguagens, impostos, imagens ou audios sem checar fonte, modelo, dependencia e teste." ], "resource_policy": { "normal_local_ai_budget_percent": 50, "agent_mode_budget_percent": 70, "heavy_install_requires_confirmation": true, "reuse_installed_first": true, "auto_downgrade_when_overloaded": true, "avoid_large_models_below_ram_gb": 16, "private_files_default": "local_only_until_user_approves" }, "capabilities": [ { "id": "text_chat", "name_pt": "Conversa local", "name_en": "Local chat", "status": "available", "runtime": [ "ollama", "cloud_byok_optional" ], "summary_pt": "Conversa geral sem exigir projeto, com historico local e respostas leves para tarefas simples.", "summary_en": "General chat without requiring a project, with local history and lightweight answers for simple tasks.", "ready_when": [ "Ollama ativo ou provedor cloud autorizado", "modelo geral compativel instalado" ], "recommended_models": [ "gemma3:1b", "gemma3:4b", "gemma4:e2b", "qwen3.5:2b" ], "user_visible": true }, { "id": "programming", "name_pt": "Programacao assistida", "name_en": "Assisted programming", "status": "available_partial", "runtime": [ "editor_local", "ollama_coder_optional", "codex_optional" ], "summary_pt": "Abre projetos autorizados, edita arquivos, salva automaticamente e pode revisar codigo. LSP completo e Codex embutido ainda sao etapas do roadmap.", "summary_en": "Opens trusted projects, edits files, autosaves and can review code. Full LSP and embedded Codex are still roadmap items.", "ready_when": [ "pasta autorizada", "arquivo textual", "modelo geral ou coder instalado" ], "recommended_models": [ "qwen2.5-coder:1.5b", "qwen2.5-coder:3b", "qwen2.5-coder:7b", "qwen2.5-coder:14b" ], "limitations": [ "Nao garantir todas as linguagens sem analisador/teste instalado", "Acoes destrutivas e deploy exigem aprovacao humana" ], "user_visible": true }, { "id": "file_upload", "name_pt": "Arquivos no chat", "name_en": "Chat file uploads", "status": "available_partial", "runtime": [ "local_upload_store" ], "summary_pt": "Recebe arquivos e guarda localmente. Textos e codigos sao os primeiros formatos seguros; PDF, imagem e audio exigem extratores dedicados.", "summary_en": "Receives files and stores them locally. Text and code are the first safe formats; PDF, images and audio need dedicated extractors.", "ready_when": [ "arquivo escolhido pelo usuario", "tipo permitido", "tamanho dentro do limite" ], "planned_extractors": [ "ocr", "pdf_text", "audio_transcription" ], "user_visible": true }, { "id": "memory_persistence", "name_pt": "Memoria persistente", "name_en": "Persistent memory", "status": "planned_optional", "runtime": [ "local_memory_store_optional", "summary_index_optional" ], "summary_pt": "Guardar contexto util por usuario, projeto e organizacao para continuidade entre sessoes com privacidade e exclusao controlada.", "summary_en": "Store useful context per user, project and organization for continuity across sessions with privacy and controlled deletion.", "ready_when": [ "politica de privacidade definida", "armazenamento local habilitado", "escopo aprovado pelo usuario" ], "limitations": [ "Nao misturar memorias entre clientes ou projetos", "Memorias antigas devem ser revalidadas antes de acao critica" ], "official_sources": [ "https://github.com/thedotmack/claude-mem" ], "user_visible": true }, { "id": "image_understanding", "name_pt": "Entender imagens", "name_en": "Image understanding", "status": "optional_model_required", "runtime": [ "ollama_multimodal", "cloud_vision_byok" ], "summary_pt": "Pode ser ativado com modelo multimodal compativel ou provedor cloud autorizado. Nao instalar automaticamente em PC fraco.", "summary_en": "Can be enabled with a compatible multimodal model or an authorized cloud vision provider. Do not auto-install on weak PCs.", "ready_when": [ "modelo multimodal instalado ou cloud autorizado", "arquivo de imagem aprovado pelo usuario" ], "recommended_models": [ "gemma3:4b", "qwen3.5:4b", "gemma4:e2b" ], "minimum_recommended_ram_gb": 16, "user_visible": true }, { "id": "ocr", "name_pt": "OCR local", "name_en": "Local OCR", "status": "planned_optional", "runtime": [ "tesseract_optional" ], "summary_pt": "Extrair texto de imagens e scans usando OCR local quando instalado.", "summary_en": "Extract text from images and scans using local OCR when installed.", "ready_when": [ "tesseract instalado", "idioma treinado disponivel" ], "official_sources": [ "https://github.com/tesseract-ocr/tesseract", "https://tesseract-ocr.github.io/tessdoc/Installation.html" ], "user_visible": true }, { "id": "audio_transcription", "name_pt": "Transcrever audio", "name_en": "Audio transcription", "status": "planned_optional", "runtime": [ "whisper_cpp_optional", "ffmpeg_optional" ], "summary_pt": "Converter audio em texto localmente, com fila e progresso visual. Requer runtime de transcricao e conversor de audio.", "summary_en": "Convert audio to text locally, with queue and visual progress. Requires a transcription runtime and audio converter.", "ready_when": [ "whisper.cpp ou equivalente instalado", "ffmpeg disponivel para converter formatos" ], "official_sources": [ "https://github.com/ggml-org/whisper.cpp" ], "user_visible": true }, { "id": "text_to_speech", "name_pt": "Falar respostas", "name_en": "Text to speech", "status": "planned_optional", "runtime": [ "piper_optional", "system_tts_optional" ], "summary_pt": "Ler respostas em voz alta usando voz local quando instalada ou voz do sistema quando disponivel.", "summary_en": "Read answers aloud using local voices when installed or system voices when available.", "ready_when": [ "voz local instalada ou TTS do sistema disponivel", "usuario ativou audio" ], "official_sources": [ "https://github.com/rhasspy/piper", "https://rhasspy.github.io/piper-samples/" ], "user_visible": true }, { "id": "image_generation", "name_pt": "Gerar imagens", "name_en": "Image generation", "status": "planned_optional", "runtime": [ "comfyui_optional", "cloud_image_byok" ], "summary_pt": "Geracao local de imagens deve ser opcional, pesada e com confirmacao. GPU e espaco em disco sao recomendados.", "summary_en": "Local image generation must be optional, heavy and confirmed by the user. GPU and disk space are recommended.", "ready_when": [ "ComfyUI ou backend equivalente instalado", "modelo de imagem instalado/licenciado", "politica de conteudo ativa" ], "official_sources": [ "https://github.com/Comfy-Org/ComfyUI" ], "minimum_recommended_ram_gb": 16, "gpu_recommended": true, "user_visible": true }, { "id": "video_understanding", "name_pt": "Entender video", "name_en": "Video understanding", "status": "planned_optional", "runtime": [ "frame_extraction", "vision_model_optional" ], "summary_pt": "Primeira versao deve extrair quadros e audio, analisar por etapas e mostrar custo/tempo antes de rodar.", "summary_en": "First version should extract frames and audio, analyze step by step and show cost/time before running.", "ready_when": [ "ffmpeg instalado", "modelo de visao ou cloud autorizado" ], "user_visible": true }, { "id": "all_languages_programming", "name_pt": "Linguagens de programacao", "name_en": "Programming languages", "status": "assisted_not_guaranteed", "runtime": [ "programming_skill", "official_docs_registry", "tests" ], "summary_pt": "A Mike pode aprender e ajudar em muitas linguagens usando documentacao oficial, mas deve validar com ferramentas reais do projeto.", "summary_en": "Mike can learn and help across many languages using official docs, but must validate with real project tools.", "ready_when": [ "linguagem detectada", "fontes oficiais selecionadas", "analisador/teste disponivel quando possivel" ], "limitations": [ "Nao prometer dominio absoluto", "Sempre citar fonte e teste quando o risco for alto" ], "user_visible": true } ], "install_profiles": [ { "id": "lite", "name": "PC leve", "max_ram_gb": 8, "enable": [ "text_chat", "file_upload", "programming" ], "manual_only": [ "memory_persistence" ], "avoid": [ "image_generation", "video_understanding", "large_multimodal_models" ] }, { "id": "balanced", "name": "PC intermediario", "min_ram_gb": 8, "max_ram_gb": 16, "enable": [ "text_chat", "file_upload", "programming", "ocr", "audio_transcription" ], "manual_only": [ "memory_persistence", "image_understanding", "text_to_speech" ] }, { "id": "power", "name": "PC forte ou com GPU", "min_ram_gb": 16, "enable": [ "text_chat", "file_upload", "programming", "image_understanding", "ocr", "audio_transcription", "text_to_speech" ], "manual_only": [ "memory_persistence", "image_generation", "video_understanding" ] } ], "optional_modules": [ { "id": "ocr_tesseract", "capability_id": "ocr", "name": "Tesseract OCR", "install_mode": "optional_background_when_winget_available", "weight": "light", "official_source": "https://tesseract-ocr.github.io/tessdoc/Installation.html", "notes": "Usado para extrair texto de imagens e PDFs escaneados. Revisao humana continua obrigatoria em documentos sensiveis." }, { "id": "audio_ffmpeg", "capability_id": "audio_transcription", "name": "FFmpeg", "install_mode": "optional_background_when_winget_available", "weight": "medium", "official_source": "https://ffmpeg.org/", "notes": "Conversor de midia usado antes de transcricao e em ferramentas de audio/video." }, { "id": "audio_whisper_cpp", "capability_id": "audio_transcription", "name": "whisper.cpp", "install_mode": "guided", "weight": "medium", "official_source": "https://github.com/ggml-org/whisper.cpp", "notes": "Transcricao local; instalacao automatica fica bloqueada ate escolher pacote/runtime por CPU/GPU." }, { "id": "tts_system", "capability_id": "text_to_speech", "name": "Voz do Windows", "install_mode": "built_in_when_available", "weight": "light", "official_source": "https://learn.microsoft.com/dotnet/api/system.speech.synthesis.speechsynthesizer", "notes": "Permite testar fala local sem baixar modelo extra." }, { "id": "tts_piper", "capability_id": "text_to_speech", "name": "Piper TTS", "install_mode": "guided", "weight": "medium", "official_source": "https://github.com/rhasspy/piper", "notes": "Voz neural local; tratar como opcional guiado ate definir canal mantido." }, { "id": "image_comfyui", "capability_id": "image_generation", "name": "ComfyUI", "install_mode": "guided_heavy", "weight": "heavy", "official_source": "https://github.com/Comfy-Org/ComfyUI", "notes": "Geracao de imagem local exige Python, modelos grandes, disco e preferencialmente GPU." } ] } }