|
Download README.md from earakely-scale/TreeLineFromRequestComputerEnv-REA: direct link, hf CLI and curl.
- Browser
- Download file 6.5 kB
-
https://huggingface.co/datasets/earakely-scale/TreeLineFromRequestComputerEnv-REA/resolve/main/README.md
- Command line
-
hf download hf://datasets/earakely-scale/TreeLineFromRequestComputerEnv-REA/README.md
-
curl -L -o README.md https://huggingface.co/datasets/earakely-scale/TreeLineFromRequestComputerEnv-REA/resolve/main/README.md
6.5 kB
| license: apache-2.0 | |
| pretty_name: TreeLine computer-use environment from a one-line request (REA) | |
| tags: | |
| - computer-use | |
| - cua | |
| - gui-agent | |
| - rl-environment | |
| - agentenv | |
| - desktop | |
| - linux | |
| - rea | |
| - outliner | |
| - tree | |
| - personal-database | |
| - from-request | |
| # TreeLine: a computer-use environment built from a one-line request | |
|  | |
| **The real TreeLine 3.1 outliner running on an Ubuntu 24.04 desktop, as an [agentenv](https://pypi.org/project/agentenv-framework/) computer environment. An agent completes the task only through screenshots, mouse and keyboard, and a verifier grades the saved files with no model.** | |
| Claude Code built this environment from a one-line request in about 34 minutes, with only REA's build skill and MCP tools; [How it was made](#how-it-was-made) has the request and the steps. TreeLine opens a reading-list outline, `Documents/reading-list.trln`, at launch: three categories, Fiction, History and Science, holding seven books with their authors and years. The agent has to add The Silk Roads by Peter Frankopan, published in 2015, under History, and save. The checks count the new book under History but do not check its title, author or year; see Limits. It sees only the screen and acts only with the mouse and keyboard. | |
| ## Watch | |
| - Agent runs (Claude Code with Claude Sonnet 5.5): [run 1](media/solve-1.mp4), [run 2](media/solve-2.mp4), [run 3](media/solve-3.mp4) | |
| - Witness, which proves the task can be done through the computer: [witness.mp4](media/witness.mp4) | |
| Each video is the computer's own screen recording at 10 frames per second. | |
| ## The task | |
| > In TreeLine, add the book The Silk Roads by Peter Frankopan, published in 2015, under the History category of my reading list in Documents/reading-list.trln, and save it. | |
| `bundle/artifacts/add-history-book-check/verify.py` reads the app's files through the computer's `data/get` view and passes only if every check holds: | |
| | Check | Kind | File | | |
| | --- | --- | --- | | |
| | `history-gains-a-book` | `member_count` | `Documents/reading-list.trln` | | |
| | `one-more-book` | `member_count` | `Documents/reading-list.trln` | | |
| | `history-kept` | `value_equals` | `Documents/reading-list.trln` | | |
| | `no-new-category` | `member_count` | `Documents/reading-list.trln` | | |
| A `.trln` file is gzip-compressed JSON, and REA reads the JSON inside: a flat list of `nodes`, each with a generated `uid`, a `format` (CATEGORY or BOOK), its `data` fields and its children's uids. The checks find History by its uid, the seed's own: it must now have three children, the file must hold eight BOOK nodes and still three CATEGORY nodes (`member_count`), and History must keep its name. **They do not check the new book's title, author or year**: REA can't yet select a new JSON record by a nested field. REA's draft suggested seven checks, all by position in the node list, which TreeLine orders by ids it generates afresh in every run; Claude Code discarded all seven and wrote these. | |
| ## Results | |
| 3 of 3 runs passed, about $0.75 per run. | |
| | Run | Outcome | Turns | Cost | Video | | |
| | --- | --- | --- | --- | --- | | |
| | 1 | passed | 17 | $1.06 | [video](media/solve-1.mp4) | | |
| | 2 | passed | 19 | $0.65 | [video](media/solve-2.mp4) | | |
| | 3 | passed | 16 | $0.53 | [video](media/solve-3.mp4) | | |
| ## Run it | |
| You need Docker. `agentenv-framework` builds the computer: Ubuntu 24.04 with the app installed from Ubuntu's archive (`treeline` 3.1.5). | |
| ```bash | |
| pip install agentenv-framework==0.9.1431 | |
| hf download earakely-scale/TreeLineFromRequestComputerEnv-REA --repo-type dataset --local-dir treeline-request | |
| agent-env run treeline-request/bundle --task add-history-book --agent claude-code --model <model> | |
| ``` | |
| The agent needs a model provider configured for agentenv. `--task smoke` only boots the computer, with no agent. `bundle/README.md` describes the bundle in detail. | |
| ## How it was made | |
| Claude Code built this environment from a one-line request, given only [REA](https://github.com/morluto/rea)'s build skill and REA's MCP tools: | |
| > Build me a TreeLine environment where the agent adds a new book under one of the categories in a reading-list outline and saves it. | |
| It asked only the skill's upfront questions, wrote no app-specific scripts, and finished in about 34 minutes. The steps were REA's usual ones: | |
| 1. **Capture.** REA launched TreeLine on a Linux desktop and recorded its package, windows and opened files while the edit was made. It also recorded the app's state folders before and after. | |
| 2. **Export.** REA turned the capture into an agentenv `computer_app`. It installs the app, seeds the starting files, starts the app, and waits for its main window. It also serves typed views of the app's files for verifiers. | |
| 3. **Task.** The task's checks were validated against the capture: they fail on the starting state and pass on the captured end state. | |
| 4. **Witness.** In a fresh computer, the book was added through computer actions alone, and the checks were read at four stages. They held on the first try: | |
| | Stage | Checks | | |
| | --- | --- | | |
| | `seed` | fail, as expected: History holds two books, and there are seven; the kept checks pass | | |
| | `after_edit` | pass | | |
| | `near_miss` | fail, as expected: the new book was moved out of History | | |
| | `after_reset` | fail, as expected: `data/reset` restored the seed | | |
| The agent runs came after the build: Claude Code with Claude Sonnet 5.5 (`--model anthropic/claude-sonnet-5-5 --no-emulate`, on an amd64 host) solved the task 3 times, each run recorded. | |
| ## What's inside | |
| ``` | |
| bundle/ the agentenv bundle: envs/ (app module, seed files), tasks/, artifacts/ (verifier), evals/ | |
| media/ witness and agent-run recordings (MP4) with short GIF previews | |
| results.json the runs above, with each check's result | |
| ``` | |
| ## Limits | |
| - **The new book's title, author and year are not checked.** The checks count that one BOOK node was added under History, so a book with any details, or with none, passes. | |
| - History is found by its uid, the seed's own id, which holds in every computer. | |
| - The computer installs whatever `treeline` Ubuntu 24.04's archive holds at build time (3.1.5 for these runs); it is not pinned. | |
| - Three runs is a small sample: the solve rate shows the task is solvable, not how hard it is. | |
| ## Licences | |
| The bundle code, the task and the synthetic seed files are Apache-2.0. TreeLine itself is not included: the computer installs it from Ubuntu's archive (`treeline` 3.1.5) at build time, under its own licence (GPL-2.0-or-later). | |