Notate Logo

What's a Collection?

A Collection in Notate is a vectorstore that enables powerful LLM-augmented interactions with your content. It transforms your documents and notes into embeddings, allowing AI models to understand and work with your data more effectively.

Key Features

  • • Vector embeddings for semantic search
  • • LLM-powered content understanding
  • • Contextual AI responses
  • • Efficient similarity matching
  • • Automatic content indexing

Use Cases

  • • AI-powered document analysis
  • • Semantic search across documents
  • • Context-aware AI chat
  • • Knowledge retrieval
  • • Content recommendation

Creating a Collection

Creating a new collection is straightforward. Follow these steps to set up your vectorstore and start organizing your content.

Collection Setup Steps

  1. Select Add New Data Store from the dropdown menu

    Selected store contains 75 documents

    Last ingested: 2024-12-29

  2. Name Your Collection

    Choose a unique name for your collection. This will be used to identify and access your vectorstore.

  3. Add a Description

    Optionally add a description to help remember the collection's purpose and contents.

  4. Choose Storage Location

    Select between:

    • Cloud Store: Uses OpenAI's embeddings for high-quality vector representations. Requires an OpenAI API key but enables synchronized access across devices.
    • Local Store: Uses Ollama's local embedding models for complete privacy and offline access. No API key required, but runs computations on your machine.
  5. Configure Embedding Model (Optional default: granite-embedding:278m for local stores)

    For local stores using Ollama, select your preferred embedding model (default: granite-embedding:278m). This model runs entirely on your machine for private and secure vector generation.

  6. Select Collection Type

    Choose between:

    • Notes: For document and text-based content
    • Chats: For conversation histories and chat interactions
    • Add New Type: Create a new type of collection for more specific use cases

Collection Management

Collections automatically handle the complexity of vectorization and embedding generation, making it easy to create AI-enhanced experiences with your content.

Vector Storage

Notate currently uses ChromaDB as its vectorstore backend, providing efficient and reliable vector storage and retrieval. Local management with additional vectorstore options (like Pinecone, Weaviate, and more) are coming soon.

Collection Deletion

When you need to remove a collection, simply click the trashcan icon in the collection management interface. This action only removes the collection's vector embeddings and metadata, while preserving your original files. This means you can:

  • Safely delete collections without losing source files
  • Reingest the same files later into a new collection
  • Access and manage original files through the File Manager
  • Optionally remove source files separately via the File Manager if needed

After Creation

  1. Your collection will be ready for immediate use
  2. Add documents or content to be vectorized
  3. The system will automatically generate embeddings
  4. Start using AI-powered features like semantic search or contextual chat
  5. Manage your collection through the intuitive interface

Using Your Collection

Interact with your collections through Notate's chat interface, where AI responses are enhanced with source-validated Notations from your content.

How to Use Collections

  1. Select Your Collection

    Choose which collection to use from the dropdown menu in the library. This determines which knowledge base the AI will reference.

  2. Ask Questions

    Type your questions or prompts in the chat. The AI will search your collection for relevant context before responding.

  3. Explore Notations

    AI responses include expandable Notations that show:

    • Source documents used for context
    • Relevant excerpts supporting the response
    • Confidence scores for retrieved content
    • Direct links to original documents
  4. Verify Sources

    Click on any Notation to:

    • View the full context around the referenced content
    • Navigate to the original document
    • Understand how the AI used this information
    • Validate the accuracy of responses

Understanding Notations

Notations are interactive citations that provide transparency into the AI's reasoning. Each Notation includes:

  • • Source document title and location
  • • Relevant text excerpts with highlighting
  • • Semantic relevance score
  • • Context window showing surrounding content
  • • Quick actions for document navigation

Understanding Notations

Notations are interactive citations that provide transparency into how the AI uses your collection's content to generate responses.

Notation Components

  • • Source document metadata
  • • Highlighted relevant excerpts
  • • Semantic similarity scores
  • • Expandable context windows
  • • Document navigation links

Interaction Features

  • • Click to expand full context
  • • Direct source navigation
  • • Relevance filtering
  • • Copy citation text
  • • Inline document preview

Benefits of Notations

  • Transparency

    See exactly which parts of your documents the AI used to generate its response.

  • Verification

    Quickly verify the accuracy of AI responses by checking original sources.

  • Context

    Understand the broader context around cited information with expandable windows.

  • Navigation

    Seamlessly move between AI chat and source documents for deeper research.