# HugstonOne — Public Product Guide

## NRS table of contents

1. [What HugstonOne is](#1-what-hugstonone-is)
2. [The strongest point](#2-the-strongest-point)
3. [Local-first workflow](#3-local-first-workflow)
4. [Models and runtime choices](#4-models-and-runtime-choices)
5. [Sessions, files, and retrieval](#5-sessions-files-and-retrieval)
6. [Previews and generated workspaces](#6-previews-and-generated-workspaces)
7. [Tools, skills, and agents](#7-tools-skills-and-agents)
8. [Optional online mode](#8-optional-online-mode)
9. [Common uses](#9-common-uses)
10. [Editions](#10-editions)
11. [Practical limits and responsible operation](#11-practical-limits-and-responsible-operation)

## 1. What HugstonOne is

**HugstonOne** is a local-first AI workspace that brings model interaction, project context, files, retrieval, previews, tools, skills, agent workflows, and optional online research into one interface.

It is intended for private knowledge work, research, writing, coding, learning, analysis, prototypes, and repeatable workflows where the user or organization wants clearer control over models and data boundaries.

## 2. The strongest point

HugstonOne's strongest point is **one controlled workspace**:

- Choose the model and runtime.
- Keep local-only work local.
- Add the minimum required files.
- Turn retrieval, tools, skills, agents, or online mode on only when needed.
- Review streaming output, previews, and generated files in the same project context.
- Stop, continue, export, or save work without a forced cloud AI account.

## 3. Local-first workflow

A typical workflow is:

1. Select a compatible local model and runtime mode.
2. Create or reopen a project session.
3. State the objective and constraints.
4. Add only relevant files or folders.
5. Enable only the capabilities the task needs.
6. Review the streamed answer and any preview or generated files.
7. Save, export, continue, or stop the work.

Local-first means that the core workspace is designed to operate on the user's machine or controlled infrastructure. It does not mean every optional feature is offline.

## 4. Models and runtime choices

HugstonOne can work with compatible locally available models and configured runtimes. Depending on the installation, operators can choose CPU, GPU, CLI, server, or combined workflows.

Actual model support and performance depend on model format, runtime compatibility, available memory, hardware acceleration, context size, and configuration.

## 5. Sessions, files, and retrieval

Sessions preserve the working thread of a project. Files can provide task context, while retrieval can locate relevant passages without repeatedly injecting an entire document collection into every request.

Typical materials include documents, source code, structured text, tables, images, and project files. Supported formats and extraction quality vary. Users remain responsible for access rights and confidential-data handling.

## 6. Previews and generated workspaces

HugstonOne can present code or content previews and can save generated project files into an isolated workspace for review and export. Previewed code is not automatically trusted. Users should inspect generated files before running, sharing, or deploying them.

## 7. Tools, skills, and agents

- **Tools** perform a defined operation, such as controlled file or data work.
- **Skills** apply a reusable workflow or instruction pattern.
- **Agents** coordinate a longer multi-step objective.

These capabilities should be explicit, bounded, and reviewable. Agent output is not a substitute for operator approval, backups, or change control.

## 8. Optional online mode

Online mode is for tasks that need current external information. It should be enabled deliberately and kept separate from local-only work.

A reliable online mode uses bounded requests, allowlisted sources, timeouts, partial results, and a fallback path. Outside services can be slow, unavailable, incomplete, or inaccurate. Important findings should be checked against primary sources.

## 9. Common uses

- Private drafting, summarization, and structured writing.
- Code review, debugging, prototypes, and project generation.
- Document and folder research with retrieval.
- Tables, charts, reports, and structured exports.
- Learning, study, and internal knowledge work.
- Image-assisted work with a compatible multimodal model.
- Repeatable tools, skills, and long-running operator-supervised workflows.

## 10. Editions

### HugstonOne Community Edition

- Free.
- Legacy release.
- Not actively updated.
- Suitable for trying the established Community feature set with the understanding that it is not the current maintained edition.

### HugstonOne Enterprise Edition

- Paid.
- Current edition.
- Intended for current professional and organizational use.
- Contact the Hugston.com team for payment details and access.

See [HUGSTONONE-EDITIONS.md](./HUGSTONONE-EDITIONS.md).

## 11. Practical limits and responsible operation

- Output speed depends on the chosen model and hardware.
- Many connected users do not imply that one model can generate at full speed for everyone simultaneously.
- Large contexts and files consume memory and processing time.
- Online sources can fail or return stale information.
- Local-first operation still requires backups, account security, disk protection, and update discipline.
- Model licenses and generated-output rights must be checked for the intended use.
- High-impact decisions require qualified human review.