Architecture Diagrams

This page shows how Squirro Neo, the generative AI service, and the Squirro platform fit together. For how projects, items, and labels are organized, and for the path your data takes through a project, see the General Concepts page.

Interfaces and Tools

• Squirro Neo

  The main Squirro user interface, at the address of your Squirro instance followed by /neo/. With Agentic AI, you chat with agents that plan, use tools such as document retrieval, and answer from your content, citing their sources. Enterprise Search finds documents and filters them by facets, and personalized dashboards are built for your employees and added to the projects that need them. For more information, see the Squirro Neo page.

• Command line tools

  Neo also comes with command line tools for headless management. Squirro CLI, the squirro command, manages the configuration of an instance from files that you keep in version control, and is available to selected customers and partners. Administrators use the genai command line tool to configure agents and schedule tasks. Developers use the neo-ui command line tool to build and deploy dashboards. For more information, see the Squirro CLI, genai CLI Reference, and neo-ui CLI Reference pages.

• Classic Squirro Interface

  The former Squirro user interface, with Squirro Chat, Search, and dashboards. It remains supported and available next to Neo, and both interfaces work with the same projects and content. It also holds the administration screens, where administrators set up projects, data sources, and pipelines, and manage system settings. For more information, see the Chat, Search, Dashboards, and Squirro Spaces pages.

Every interface and tool reaches the platform through the Squirro APIs.

Generative AI Service

The generative AI service runs the agents that answer in Chat. In Neo, the agents follow the configuration that your administrators deploy, which covers the following:

• Agents

  The instructions each agent follows, the settings users cannot change, and which agents users can pick. For more information, see the Authoring Agents page.

• Models

  The language models that power the agents. A model can be set for the project, for an agent, by a user, or for a single request, within the limits that the configuration allows. For more information, see the Resource Reference page.

• Tools

  The actions an agent can take besides writing an answer, such as retrieving documents, querying a knowledge graph, or calling an external system. For more information, see the Agent Configuration Overview page.

• Tasks and approvals

  The agents that run on a schedule, and the actions that need a person to approve them before they run. For more information, see the Scheduled Tasks and Approvals page.

To answer, the agents call the LLM providers of the models you configure, the knowledge graphs in Squirro Graphite, and external systems through MCP servers. For more information, see the Agent Configuration page.

Squirro Platform

Neo and the Classic Squirro Interface share the same platform, so the projects, content, and settings you already have carry over to Neo.

• Data loading

  Connectors and the data loader bring content in from your data sources. For more information, see the Data Loading page.

• Processing pipeline

  Enriches every item before it is indexed, for example with known entity extraction and custom pipelets. For more information, see the Data Processing Pipeline page.

• Search

  Processes queries, ranks the results, and provides the facets that Search in Neo uses as filters. For more information, see the Search page.

• AI Studio

  Trains the machine learning models that the pipeline applies to your items. For more information, see the AI Studio page.

• Squirro Classifier

  Detects entities and labels in your documents. For more information, see the Squirro Classifier page.

• Permissions

  Controls which users and groups can reach each project. For more information, see the Squirro Roles & Permissions page.

Squirro Graphite is a separate Squirro product that holds your taxonomies, ontologies, and knowledge graphs. The pipeline uses the taxonomies to tag items with concepts, and the agents query the knowledge graphs to answer. For more information, see the Squirro Graphite page.

Data Stores

• Search index

  Holds your indexed items. Elasticsearch is the default, and OpenSearch is available as a technical preview. For more information, see the Elasticsearch Management and System Requirements pages.

• Settings and state

  A SQL database holds the settings and state of the platform services. MariaDB is the default, and PostgreSQL is supported as an alternative. For more information, see the System Requirements and Configure PostgreSQL as the Database Backend pages.

• Cache and queues

  Redis caches data and queues work between the platform services. For more information, see the Securing Redis Instance over SSL page.

For the versions installed with each release, see the Third-Party Component Versions page.