Metadata-Version: 2.4
Name: localagentbrowser-cleanroom
Version: 1.67
Summary: Clean Room, a private edition of LocalAgentBrowser: an autonomous web research agent with checked answers, on a local LLM (LM Studio) only, and Playwright.
Author: SnowFX
License-Expression: MIT
Project-URL: Homepage, https://github.com/SnowFXXX/LocalAgentBrowser
Project-URL: Website, https://snowfxxx.github.io/LocalAgentBrowser/
Project-URL: Issues, https://github.com/SnowFXXX/LocalAgentBrowser/issues
Keywords: ai-agent,autonomous-agent,llm,local-llm,lm-studio,playwright,browser-automation,web-agent,web-research,qwen,local-first,privacy
Classifier: Development Status :: 4 - Beta
Classifier: Environment :: Console
Classifier: Intended Audience :: Developers
Classifier: Intended Audience :: Science/Research
Classifier: Operating System :: OS Independent
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Programming Language :: Python :: 3.13
Classifier: Topic :: Internet :: WWW/HTTP :: Browsers
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Requires-Python: >=3.10
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: openai>=2.0
Requires-Dist: playwright>=1.50
Provides-Extra: dev
Requires-Dist: pytest>=8; extra == "dev"
Requires-Dist: ruff>=0.15; extra == "dev"
Dynamic: license-file

# LocalAgentBrowser Clean Room

**Clean Room is a private edition of [LocalAgentBrowser](https://github.com/SnowFXXX/LocalAgentBrowser):** an
autonomous web research agent with checked answers, run by a language model on your own PC in
[LM Studio](https://lmstudio.ai), with a Chromium browser you can watch.

The full edition can also use hosted AI models over the internet, if you turn that on. Clean Room doesn't have that
option at all: the code for it is removed before the edition is built, not just switched off, and the build fails if
any of it is left. What Clean Room connects to:

- **The model server:** LM Studio on this PC, or another machine on your own network. Clean Room refuses any other
  address, whatever `LMSTUDIO_BASE_URL` says.
- **The search engine** (DuckDuckGo) and **the pages the agent reads**, the same as a person browsing.
- During setup only: the Chromium browser download, and the model you pick (from Hugging Face, through LM Studio).

Your objective, the pages read and the report stay on your PC. Each run writes its research log and report to its
own folder.

## Install

### Windows: the installer

1. Install [LM Studio](https://lmstudio.ai) (free).
2. Download **`LocalAgentBrowser-CleanRoom-<version>-setup.exe`** from the
   [latest release](https://github.com/SnowFXXX/LocalAgentBrowser/releases/latest) and run it.
   - Windows may say **"Windows protected your PC"**, because the installer isn't code-signed yet. Click
     **More info → Run anyway**.
   - It installs for your Windows account only, with no administrator rights, and brings its own Python.
   - It installs beside the full edition, in its own folder (`LocalAgentBrowser Clean Room`) with its own
     shortcuts, so you can have both.
3. Setup asks which of the tested models to download through LM Studio, then downloads the Chromium browser the agent
   uses (about 150 MB, once).
4. Open **LocalAgentBrowser Clean Room** from the Start menu or the desktop.

To remove it: **Start menu → LocalAgentBrowser Clean Room → Uninstall**, or Windows Settings → Apps. Your reports in
Documents are kept.

### Any system: pip

You need Python 3.10 or later and LM Studio.

```bash
python -m venv venv
venv\Scripts\activate
pip install localagentbrowser-cleanroom
python -m playwright install chromium
```

On macOS or Linux, activate with `source venv/bin/activate` instead. The package installs the same `localagentbrowser`
command and module as the full edition, so install one or the other in a given environment, not both.

## First run

1. In LM Studio, load your model with a **context length of 8192** and start the local server (Developer tab → Start
   Server).
2. Start LocalAgentBrowser Clean Room: the Start-menu shortcut on Windows, or `localagentbrowser` in a terminal.
3. Type your objective, for example *What is the tallest mountain in Africa, and how tall is it in meters and in
   feet?* A Chromium window opens and the agent starts working. The first step takes a minute or two on a CPU.
4. When it finishes, the report opens in your browser, or read `research_report.md` in the folder you ran it from.
   Runs from the shortcut each get their own folder under **Documents\LocalAgentBrowser runs**.

A run takes about 10 to 60 minutes on a CPU. Press Ctrl+C at any time to stop; the report is still written with
everything found so far.

## Settings

| Environment variable | Default | Meaning |
|---|---|---|
| `LMSTUDIO_BASE_URL` | `http://localhost:1234/v1` | LM Studio's address: this PC or your own network only |
| `LMSTUDIO_MODEL` | `qwen2.5-7b-instruct` | Model identifier as LM Studio lists it |
| `LMSTUDIO_API_KEY` | `lm-studio` | A key, if your LM Studio server requires one |
| `LMSTUDIO_CONTEXT_LENGTH` | `8192` | Context length the model is loaded with |
| `LAB_SIDE_BY_SIDE` | on | On Windows, `0` keeps the console and the browser from being placed side by side |

## More

How the agent works, what it was tested on and its limitations are in the
[project README](https://github.com/SnowFXXX/LocalAgentBrowser#readme) and
[How it works](https://github.com/SnowFXXX/LocalAgentBrowser/blob/main/docs/how-it-works.md). Clean Room runs the
same agent and the same checks as the full edition, from the same code.

MIT License; see `LICENSE`.
