Tool-Store
The Tool-Store is the central registry for discovering, evaluating, and managing reusable computational assets in the Federated Learning Net network.
It provides access to global tools, user-private tools, trained models, and workflows, enabling users to compose and execute data-driven analyses in a consistent and reproducible manner.
This page serves as a reference overview of the Tool-Store and its core concepts.
Conceptual Overview
The Tool-Store aggregates different types of executable artifacts under a unified interface:
-
Tools
Executable, containerized applications that perform a specific computation (e.g. data transformation, model training, inference). -
Models
Pre-trained tools. Conceptually, models are tools whose primary artifact is a trained model that can be reused without retraining. -
Workflows
Pipelines composed of multiple tools, defining a structured sequence of execution steps.
All tools available in the Tool-Store are either:
- Global tools, accessible to all users, or
- Private tools, visible only to the owning user.
Browsing the Tool-Store
The Tool-Store presents all available entries as cards in a searchable and filterable list.
Each tool card displays:
- Name
- Short description
- Rating
- Tags
- Tool type (Tool, Model, or Workflow)
- Version information
Search and Filtering
Users can efficiently explore the Tool-Store using:
- A search field to query by name, description, or tag
- Filter panels to refine results by:
- Tool type (Tool, Model, Workflow)
- Rating
- Tags
- Visibility (global or private)
Sorting options allow ordering by relevance, name, or rating.
Ratings and Tags
Ratings
- Any user can rate a tool.
- Ratings are visible on tool cards and can be used as a filter criterion.
- Ratings provide a lightweight signal for tool quality and community adoption.
Tags
- Tags are assigned based on tool metadata and export-related classifications.
- They provide contextual information such as:
- Intended usage
- Data compatibility
- Execution characteristics
Tags can be combined with search and rating filters to narrow down relevant tools.
Managing Tools
The Tool-Store allows users to:
- Browse all available tools and workflows
- Distinguish between global and private tools
- Compare tools based on metadata, ratings, and tags
Actions such as installing tools into projects or starting executions are accessible from the Tool-Store interface, but are documented in the respective project and experiment sections.
Scope and Limitations
This page focuses exclusively on the Tool-Store itself:
- It does not describe individual tool detail pages.
- It does not cover execution, monitoring, or configuration workflows in detail.
For related topics, see:
docs/frontend/experiments.mddocs/frontend/connectors.md
Summary
The Tool-Store acts as the discovery and entry point for reusable computational assets in the Federated Learning Net network.
By unifying tools, trained models, and workflows under a single interface with rich metadata, ratings, and filtering, it enables users to efficiently identify suitable building blocks for their analytical tasks.