Notely

AI & MCP Subsystem Overview ​

Notely features a modular, local-first AI platform designed around private data control. Markdown files remain the absolute source of truth, parsed and indexed into offline-first SQLite databases to fuel semantic search, knowledge discovery, and external agent integrations via Model Context Protocol (MCP).

Capabilities at a Glance ​

1. Embedded Model Context Protocol (MCP) Server ​

  • Runs a dual-transport (Streamable HTTP and SSE) MCP server on http://127.0.0.1:3700.
  • Connects external AI assistants (Google Antigravity, Claude Desktop, Cursor, and IDE agents) directly to workspace notes, tasks, and diagrams.
  • Exposes 7 high-signal enterprise tools (search, read_note, edit_note, manage_tasks, manage_diagrams, workspace_overview, git_control) with atomic write safety and dry-run options.
  • Exposes 9 standard MCP prompts (summarize_note, plan_tasks, explore_knowledge_graph, refactor_note, daily_review, synthesize_document, create_diagram, atomic_split, codebase_sync).

2. SQLite Knowledge Graph Engine ​

  • Extracts conceptual entities and typed relationships from Markdown documents using an offline gliner2-multi-v1-onnx neural model.
  • Outbound relations, tags, and CTE traversals are mapped into {workspace}/.notes-app/ai-graph.db.
  • Explore notes and entity clusters interactively via Workspace → Workspace Graph (Ctrl/Cmd + Shift + G).

3. Local Vector Embedding Indexer ​

  • High-performance vector database ({workspace}/.notes-app/ai-embeddings.db) storing note chunk vectors.
  • Runs entirely offline using a local ONNX runtime for BGE-small-en-v1.5 384-dimensional dense vectors, with optional fallback to cloud embedding APIs.
  • Background worker process handles queue processing and debounced note indexing to prevent UI thread latency.

4. Speech-to-Text (STT) Transcription Pipeline ​

  • Transcribe voice memos and meeting discussions into Markdown and companion JSON transcript files.
  • Runs offline via on-device WebAssembly/ONNX Whisper models (whisper-tiny.en, whisper-base.en, whisper-small), with optional sub-second cloud transcription via Groq or OpenAI Whisper.
  • Synchronized transcript view with key points and action items embedded into notes and the Media Gallery.

5. External IDE & AI Bridge ​

  • Automatically configures external AI coding assistants (VS Code, Cursor, Claude Code, GitHub Copilot, Antigravity, Windsurf).
  • Real-time debounced synchronization of workspace sitemaps, open task registries, knowledge hubs, and extracted binary documents in .notes-app/ai/dynamic-context.md.
  • Transpiles custom workspace prompts into executable VS Code .prompt.md slash commands and Cursor .mdc rules.
  • → External IDE & AI Bridge — full guide

6. In-App Tools Catalog & MCP Diagnostics ​

  • MCP Tools Page (Ctrl/Cmd + Shift + M): Interactive catalog to view tool definitions, test inputs, and verify outputs.
  • AI Health / Diagnostics: Real-time telemetry dashboard monitoring MCP connection status, client sessions, request volume, and error diagnostics.
  • AI Settings (Ctrl/Cmd + Shift + ,): Configure API providers, embedding options, STT engines, and Knowledge Graph extraction confidence thresholds.