Vault Intelligence

Appearance

Vault Intelligence: Key User Features ​

Core Philosophy: The Active Partner ​

Vault Intelligence transforms Obsidian from a passive storage system into an active collaborator. It doesn't just "search" your notes; it reasons about them, maintains them, and connects them.

1. The Researcher (Agentic Reasoning) ​

Your primary interface for interacting with your knowledge base.

  • Dual-Loop Architecture:
    • Loop 1 (Reflex): Instant, typo-tolerant search that runs locally on your device (<100ms).
    • Loop 2 (Analyst): A deep-reasoning agent that uses "Asymmetric Embeddings" to understand complex questions and trace "hidden threads" across your vault.
  • Deep Recall: The agent follows metadata bridges (tags, active file context, frontmatter topics) to find connections between notes that don't explicitly link to each other.
  • Multilingual Native: Fully supports querying and reasoning in 140+ languages. We combine high-performance local indexing for 30+ major languages with the deep linguistic understanding of frontier AI models.
  • Active Context Awareness: Automatically prioritizes your currently open note and can "see" what you are working on.

2. The Explorer (Discovery & Search) ​

A next-generation search engine built for the "I know I wrote this somewhere" moments.

  • Semantic Galaxy View: A high-performance, interactive 3D-like graph that visualises your vault's relationships in real-time. It centres on your active note, bridging the gap between structural links and semantic similarity.
  • Visual RAG (Retrieval-Augmented Generation): The graph reacts to the Researcher agent. When the AI mentions notes in its reasoning, they are automatically highlighted in the galaxy to provide spatial context.
  • Graph-Augmented Relevance (GARS): A unique scoring algorithm that combines:
    • Vector Similarity: Conceptual matches ("Idea" matches "Thought").
    • Keyword Precision: Essential term matching (BM25).
    • Graph Centrality: Boosting notes that are "hubs" of knowledge (PageRank-like).
  • Zero-Noise Excalidraw: Specialized indexing for visual thinkers. It strips away internal JSON metadata so drawings only appear in search when their text matches your query.
  • Hybrid Search: Merges fuzzy keyword search with vector semantic search for Permissive Recall (finding "cat" in "cats").
  • Slim-Sync Engine: The searchable index is up to 90% smaller on disk, ensuring lightning-fast syncing across devices without hitting storage limits.
  • Model-Specific Sharding: Isolate storage for different embedding models, allowing you to switch between Local and Gemini providers without losing data or risking corruption.
  • Self-Healing Index: Automatically re-indexes when you change models or configurations.

3. The Gardener (Vault Hygiene) ​

The first AI agent dedicated to keeping your vault clean.

  • Proactive Maintenance: Reads your vault's ontology and suggests structure.
  • Topic Classification: Automatically identifies notes that are missing topics and suggests valid ones from your existing hierarchy.
  • Interactive Plans: Generates "Hygiene Plans" (markdown files) that let you review and apply changes with a single click—ensuring you always remain in control.

4. Architecture & Privacy ​

Built for performance, privacy, and sovereignty.

  • Sovereign Intelligence: Option to run Local Embeddings (Nomic, etc.) so your vector data never leaves your device.
  • Privacy-First: Detailed control over what gets sent to the cloud. "Trust but Verify" modals for any agent write operations.
  • Worker-First Design: Heavy lifting (indexing, graph traversal) happens in a background web worker, keeping the Obsidian UI unrelatedly smooth.
  • Computational Solver: The agent can write and execute Python code to analyze data within your notes (eg "Calculate the average rating of books I read this year").