VaultAI

Documentation

Getting Started
With VaultAI

Everything you need to know about using VaultAI for private document analysis.

Getting Started

Get up and running with VaultAI in seven steps.

  1. Create an Account — Sign up at vaultai.tech with your email. Creating an account costs nothing; you add funds before your first vault.
  2. Add Funds — Navigate to Billing and add prepaid balance. Each top-up is any amount from €1 to €20,000, or its equivalent in the currency you pick at the day's rate.
  3. Create a Vault — Click “New Vault” from the Dashboard. Choose a use case and model tier.
  4. Wait for Provisioning — Your dedicated GPU instance usually starts and loads the AI model within a few minutes; on a busy host it can take up to 20 minutes.
  5. Upload Documents — Drag and drop files into the sidebar. Supported formats listed below.
  6. Start Analyzing — Ask questions in the chat, generate reports, or edit documents.
  7. Destroy When Done — Click “Destroy” to permanently delete all data and stop billing.

Supported File Types

PDF

Full text extraction with page tracking

DOCX

Paragraph and table extraction

XLSX / XLS

Sheet-by-sheet data extraction

TXT, CSV, MD

Direct text processing

PNG, JPG, JPEG

OCR text extraction (via Tesseract)

Max file size: 50MB per file.

Chat & Analysis

Once documents are uploaded, they're automatically chunked and indexed for retrieval-augmented generation (RAG).

Ask questions in natural language. The AI searches your documents for relevant passages and generates answers with source citations.

Select specific documents using the checkboxes in the sidebar to limit the AI's context to those files.

Citations appear below each response showing the source document and page number.

Document Editing

VaultAI can modify your documents based on natural language instructions.

How to edit a document

  1. Select a document in the sidebar (checkbox).
  2. Type your edit instruction (e.g., “Change the termination period from 6 months to 12 months”).
  3. Click the pencil icon button (between the generate and send buttons).
  4. The AI generates find/replace edits and applies them to a new version.
  5. Download the edited file using the download button in the chat response.

Original files are always preserved. Each edit creates a new version.

File Generation

Generate structured files from your documents using the file generation button (file icon in chat input).

Available output types

  • Summary (DOCX)
  • Report (DOCX)
  • Extraction (XLSX)
  • Comparison (DOCX)
  • Timeline (MD)
  • Action Items (XLSX)

Generated files are available for immediate download.

Model Tiers

FLASH—Qwen 3 4B (FP16)

Fast, good for summaries and extraction

$0.99/hr

TURBO—Qwen 3.5 35B MoE (3B active)

Best quality/speed ratio for professional work

$4.99/hr

ULTRA—GLM-5 744B MoE (40B active)

Coming soon — not available to order yet. Frontier-class reasoning, 8x H200

All models are open-source and served via vLLM.

Billing

  • VaultAI uses a prepaid balance model. Add funds before creating vaults.
  • You are charged for the time a vault is running, metered by the second at its per-minute rate. A running vault is not paused automatically when your balance reaches zero, so pause vaults you are not using.
  • Pause a vault to stop billing. Your chat history is kept. Your documents stay on its GPU: if that GPU is still free when you resume, they come back; if not, the vault starts on a new GPU without them and you upload them again.
  • Destroy vaults to permanently stop billing and delete all data.
  • View your full usage history and transaction log on the Billing page.

Have questions? Read the FAQ or create an account.