Home

Introduction

The project was initiated in 2023 with the 'eSearch' working title. The goal was to create a full-text Desktop search application tailored for the European market. Our focus was on building a user interface that is easily 'user' translatable and modern in design. Due to licensing constraints, we selected the open-source and well-established Lucene .NET search engine.

To provide a contemporary user experience, we adopted the open-source Avalonia framework, using an MVVM architecture and targeting the latest version of .NET Core. The default UI supports dark mode for a modern look and feel.

Traditional Boolean AND, OR, NOT operators can easily be entered using buttons. For stemming, users can choose between the traditional Porter English stemmer or our multilingual stemming rules, originally developed for dtSearch. For phonic search, we implement standard Soundex rules. For synonyms, we utilize the same rules that we created for the User Thesaurus Plus add-on for dtSearch. Beyond these features we added a scrollable word list with alphabetic sorting, and intelligent handling of punctuation characters (.,) the word list keeps numbers and letters separated by punctuation together, for example complete IP4 addresses and sums of money appear in the word list unbroken, as do possessive apostrophes (e.g. Charles's flight) and contractions (e.g. d'un).

Milestones

Apr 2023: Project started
Sep 2024: First public beta (no AI mode)
Nov 2024: First portable beta
Feb 2025: AI Mode with Perplexity search models
Apr 2025: AWS S3 plugin released
May 2025: UI updates with new results grid control and multiline AI input
Jul 2025: Project open-sourced
Aug 2025: Scheduled indexing
Oct 2025: 'Full service local LLM support': Hugging Face downloads, Local Server, Local Model connections with Advance Settings to fine-tune sampler parameters (Temp, Top-K, Top-P, Min-P, etc). Users can migrate GGUF folders/files from LM Studio.
Nov 2025: TARILIO branding. Plugin code open-sourced. AI Mode supports attachments.

For details of all public releases see: https://searchcloudone.com/release-notes-esearch/

Plugins

We introduced a plugin architecture to support indexing from additional data sources such as Amazon S3, Dropbox, SharePoint, and more.

These may be distributed as community plugins, similar to the model used by the Obsidian application, which distinguishes between core and community plugins.

In April 2025 we released an AWS S3 plugin, this also connects to an Amazon Web Services (AWS) S3 bucket as a demonstration.
In November 2025 the plugin code was open sourced. AWS S3 plugin

Create your own plugins

Complete instructions: https://github.com/electronart/S3-Sample-Plugin/wiki/TARILIO-Plugins

AI Mode

As large language models (LLMs) rapidly evolved and gained commercial relevance, we introduced an AI Mode in late 2024. This mode supports connections to the Perplexity API, OpenAI API, OpenRouter.ai and a customizable option for any other OpenAI-compatible API, whether online or offline.

A particular strength is that Custom mode can run multiple offline LLMs on your own PC or local area network. The August 2025 release added Ollama and LM Studio local server connections to the Service dropdown selector. The October 2025 release added downloading of LLMs from Hugging Face and a Local Server so that users can serve LLMs over a network; users can install TARILIO on other machines on the network to connect as clients. LLMs previously downloaded with LM Studio can be imported or migrated. LLMs can be set up with different system prompts to define an assistants role and some context, so that users can choose which assistant to work with for specialised tasks e.g. translator, research assistant, creative writer. System prompts can be exported into a public folder on a network and imported by others on the network.

Users can save conversations in plain text or export them as JSON, CSV or ECONV format, all of which can be imported by others on the network or indexed and searched later. IT Admins can periodically index and search conversations to monitor usage. Users can add notes to any part of the conversation - system, user, assistant - to provide feedback.

We aim to expand these AI capabilities as the field continues to mature.

MCP Client

In June 2025 the MCP Client was introduced to allow Tool access to the LLMs.

Agents

https://github.com/openclaw/openclaw

https://github.com/jiulingyun/openclaw-cn

MoltBook

https://github.com/vnnkl/openclaw-nostr

https://github.com/google/adk-docs

https://github.com/a2aproject/A2A

Portable versions

In addition to the main desktop application, we developed a portable version for use on USB drives. The is ideal for eDiscovery, allowing searches on machines without requiring software installation.

The TARILIO Pro Publish variant is for publishing search-only data (Operating Manuals, eBibles/Production Bibles, etc.). It has additional code and license to facilitate distribution of search-only (no indexing) versions within an organization or more widely. Contact support@electronart.co.uk with your distribution needs for further details. This version is covered by a commercial license and requires a serial number to be entered. There is no open source 'free for personal use' version.

Command Line Options

For enterprise environments, a command line interface allows IT administrators to restrict user capabilities. These restrictions can limit users to searching existing indexes or querying LLMs, without the ability to create new indexes, modify LLM connections, or change configuration settings.

Licensing

The project is open-sourced under a modified Apache 2.0 license. The TARILIO and TARILIO Portable versions are free for personal, non-commercial use. A TARILIO PRO branded version is available for commercial use.