Valstorm VFS Hybrid Search & AI Workspace Assistant

User Guide & Best Practices

1. Introduction & Overview

Valstorm's Virtual File System (VFS) Hybrid Search is a next-generation workspace intelligence engine. It combines:

1. Sub-millisecond Metadata Search: Instantly matches filenames, folder paths, modules, and file extensions as you type.
2. Semantic Vector Search: Understands the meaning and context of your documents, allowing you to find files even if you don't remember the exact name.
3. Workspace AI Assistant (RAG): Ask natural-language questions across your cloud files and receive immediate, factual answers grounded strictly in your team's knowledge base, complete with clickable source citations.

2. Getting Started: The Spotlight Search Bar

2.1 Opening Search

• Keyboard Shortcut (Mac): Press ⌘K anywhere in Valstorm Desktop.
• Keyboard Shortcut (Windows/Linux): Press Ctrl + K.
• Top Header Bar: Click the search bar at the top of the application window.

┌────────────────────────────────────────────────────────────────────────┐
│ 🔍 Search cloud files, contracts, invoices, or ask AI...          ⌘K   │
└────────────────────────────────────────────────────────────────────────┘

3. The 3 Ways to Search

Mode 1: Exact Filename & Path Navigation (Exact Lookup)

Type any part of a file title, vault folder, or extension to immediately pull up matches with zero latency.

• Examples:
  • invoice 2026 $\to$ Returns all 2026 client & vendor invoices.
  • Lamm Migration $\to$ Jumps straight to Lamm Data Migration.md.
  • .pdf or .json $\to$ Filters files matching that filetype.

Mode 2: Semantic Concept Search (Semantic Search)

When you don't remember a filename but remember what the document is about, describe the concept in plain English. The AI engine searches the underlying text embeddings across all indexed cloud files.

• Examples:
  • confidential obligations and non disclosure $\to$ Automatically finds and highlights snippets from Valstorm_Mutual_NDA_Signable.pdf.
  • cryogenic temperatures for quantum systems $\to$ Retrieves research notes containing physics excerpts.
  • steps for database cutover during migration $\to$ Finds the operational migration runbooks.

Mode 3: Natural Language Questions & Workspace AI Answers (RAG Query)

Need a direct answer without opening and skimming through multiple 20-page PDFs? Just ask a full question!

• Examples:
  • What are the payment terms in the mutual NDA?
  • How do we configure call routing in ValPhone?
  • What were the key takeaways from the Bob & Jared meeting on August 1st?

How AI Workspace Answers Work:

1. Instant Document Retrieval: The engine pulls the most relevant document chunks across your organization.
2. Live Token Streaming: A dedicated AI model (gemini-flash-lite-latest) streams a concise, direct answer into an AI Workspace Answer card.
3. Verified Sources: Every answer displays clickable Source Chips (e.g. [ 📖 Mutual_NDA.pdf ]). Clicking any source opens the exact document directly in your Valstorm Editor or Preview pane.

┌────────────────────────────────────────────────────────────────────────┐
│ ✨ AI WORKSPACE ANSWER                                                 │
│ The payment terms outlined in the agreement specify net-30 upon receipt│
│ of invoice, with late penalties accruing after 45 days.               │
│                                                                        │
│ Sources: [ 📖 Mutual_NDA.pdf ]  [ 📖 Sales_Quote_Template.pdf ]         │
└────────────────────────────────────────────────────────────────────────┘

4. Keyboard Shortcuts & Power Navigation

5. Supported Document Formats

All text, code, document, and data files are automatically indexed for semantic search and AI answers upon upload:

Note: Media formats like .png, .jpg, .mp4, .mp3, and .heic are indexed by filename and metadata, while their binary bodies are preserved without consuming vector index capacity.

6. Real-Time Indexing & Data Lifecycle

• Instant Upload Indexing: Whenever you upload a document, save changes in the Editor, or create a new file version, background Celery workers automatically extract text and update vector embeddings in real time (~100–500ms).
• Automatic Vector Cleanup: When you delete a file or vault, all associated vector chunks and embeddings are instantly purged from the search index.
• Tenant Isolation: Every search query and vector lookup is strictly partitioned by your organization's ID (org_id), ensuring data never leaks across organization boundaries.