General Concepts

This page explains how the main parts of the Squirro platform relate to each other, and where each part is documented in full. For the definition of an individual term, see the Squirro Glossary page.

How Squirro Is Organized

The Squirro platform is structured as a hierarchy, from a whole deployment down to the metadata on a single piece of content. Each level holds the levels below it.

• Instance

  A single deployment of the Squirro platform. One instance can hold several projects. Smaller deployments run on a single server, and larger ones spread the Squirro services across several nodes for capacity and resilience. For more information, see the System Requirements page.

• Workspace

  An optional boundary that groups projects and isolates them from each other. Workspaces control which users reach which projects. The feature is turned off by default, and must be turned on directly after installation, before any users are added. For more information, see the Workspaces page.

• Project

  The level you work in day to day. Each project has its own data sources, data processing pipeline, dashboards, and members. Most configuration applies to a single project. For more information, see the Project Browser page.

• Item

  The fundamental unit of content in Squirro, such as a news article, a PDF document, a service ticket, or an email. Every item belongs to one project. For more information, see the Item Format page.

• Label

  A key/value pair attached to an item, for example, a city label with the value London. Labels are what you filter, group, and navigate by. The Squirro API and codebase refer to the same concept as a keyword, a tag, or a facet. For more information, see the Labels page.

Permissions are granted for each project, so a user can hold access to a single project or to several projects within the same instance. For more information, see the Squirro Roles & Permissions page.

How Data Moves Through Squirro

Data follows the same path through every project.

1. Data loading

   Data enters a project from its sources. The primary method is the data loader in the user interface, and a command line data loader is available as well. Built-in connectors cover common source systems. For more information, see the Data Loading page.

2. Data processing

   Every item that arrives flows through the data processing pipeline, which enriches, relates, classifies, and indexes it. Pipeline steps add labels and entities to each item, and you can add steps of your own as pipelets. An item becomes searchable once the pipeline has indexed it. For more information, see the Data Processing Pipeline page.

3. Search and applications

   Indexed items reach your users through search, dashboards, and chat. For more information, see the Search, Dashboards, and Chat pages.

Which AI Capabilities Act on Your Data

The following capabilities act on items as they pass through the pipeline, or on indexed items in applications such as search and chat.

• AI Studio

  Trains and deploys AI models through a no-code interface. For more information, see the AI Studio page.

• Machine Learning Service

  Builds machine learning models that classify items, score them for relevance, group them into topics, and extract people, organizations, and places. For more information, see the Machine Learning page.

• Known Entity Extraction

  Links unstructured text to structured records, such as company or product names. For more information, see the Known Entity Extraction page.

• Retrieval Augmented Generation

  Combines search across your own items with a large language model, so that generated answers cite the items they came from. For more information, see the Retrieval Augmented Generation page.

• Agents

  Combine a large language model with dedicated tools to complete a task. For more information, see the Agent Framework page.

• Squirro Graphite

  Holds your taxonomies and ontologies. Dedicated pipelets use a large language model to tag items with concepts from those taxonomies, and to propose new concepts and relationships found in your content. For more information, see the Squirro Graphite page.