File size: 4,906 Bytes
e5dd13c | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 | # Tellico computer app
Exported by REA's `export-desktop-application` from a capture of the `tellico` package
(capture Evidence `ev_fdb48b142e6ac091f6afd3ee62cfc396189707de8d57e85b0ade720218c0b43e`). It was checked against agentenv-framework
0.9.1431.
- `envs/tellico/` is a computer app on the `ubuntu-24.04` base with the `reference` driver, for
`amd64` and `arm64`. It installs `tellico` from the distribution archive, launches `sh -c cd "$HOME" 2>/dev/null; exec 'tellico' '-qwindowtitle' 'Tellico' "$HOME"/'Documents/my-books.tc'` at boot and is ready
when a window of class `tellico` appears.
- At launch the app opens `~/Documents/my-books.tc`, the seed's one document.
- Owned paths: `~/.config/tellicorc`, `~/.local/share/tellico`, `~/Documents`.
`data/get` returns their files and a typed JSON view of each; `data/reset` restores the seed.
- Seed: 14 file(s) made for the task, as saved at checkpoint `seed`.
Text files and dconf keys spell the capture host's name as `localhost`, `user@host` as `user@localhost`
and the home as `/home/user`, which the computer respells to its own at boot.
- Neither owned nor seeded, so the app starts without what it saved there: files changed during the capture in
`~/.config/QtProject.conf`, `~/.local/share/RecentDocuments`.
To own them, capture again naming them with `state_paths` on `start_build_capture`, or `--state-path` on `capture-desktop-application`.
- Seeded as captured, though the app keeps them for its own running, so a stale one can change how it starts:
WebKit or Chromium storage `~/.local/share/tellico/QtWebEngine/Default/Network Persistent State`; caches `~/.local/share/tellico/QtWebEngine/Default/GPUCache`.
- Left out of the seed: lock files the app holds while it runs, since a copied one can make it find its profile or document
in use and keep it from starting: `~/.local/share/tellico/QtWebEngine/Default/Platform Notifications/LOCK`.
- `tasks/smoke.json` deploys the app.
Run the task with `agent-env run <this folder>`, or only the deploy with `--task smoke`.
The seed was made for this task, not taken from a user's own data; read `envs/tellico/seed/` before publishing the bundle.
## Task `add-glass-orchard`
`tasks/add-glass-orchard.json` deploys the app, has the agent work through the computer from the prompt below, and grades
the outcome with `artifacts/add-glass-orchard-check/verify.py`, which reads the app's `data/get` views and needs
no model. It grades once the app's files have held for 3 s, so a save the app makes on its own within that
time of the agent stopping counts; each result's `settled` is false when they were still changing 15 s after the
first read. `evals/tellico.toml` names it, so a run of the bundle runs it.
> Add the book The Glass Orchard by Aurelio Benn to my book collection in Documents/my-books.tc, with Northwind House as the publisher and 2022 as the publication year, and save it.
Checks, each of which must pass:
- `title`: `{"kind":"text_equals","file":"Documents/my-books.tc","pointer":"/entries/[name=tellico.xml]/view/root/children/[name=collection]/children/[name=entry][@id=4]/children/[name=title]","value":"The Glass Orchard","normalize_whitespace":true}`
- `author`: `{"kind":"text_equals","file":"Documents/my-books.tc","pointer":"/entries/[name=tellico.xml]/view/root/children/[name=collection]/children/[name=entry][@id=4]/children/[name=authors]","value":"Aurelio Benn","normalize_whitespace":true}`
- `publisher`: `{"kind":"text_equals","file":"Documents/my-books.tc","pointer":"/entries/[name=tellico.xml]/view/root/children/[name=collection]/children/[name=entry][@id=4]/children/[name=publisher]","value":"Northwind House","normalize_whitespace":true}`
- `pub-year`: `{"kind":"text_equals","file":"Documents/my-books.tc","pointer":"/entries/[name=tellico.xml]/view/root/children/[name=collection]/children/[name=entry][@id=4]/children/[name=pub_year]","value":"2022","normalize_whitespace":true}`
- `keeps-first-book`: `{"kind":"text_equals","file":"Documents/my-books.tc","pointer":"/entries/[name=tellico.xml]/view/root/children/[name=collection]/children/[name=entry][@id=1]/children/[name=title]","value":"The Lantern Keeper"}`
- `keeps-third-book`: `{"kind":"text_equals","file":"Documents/my-books.tc","pointer":"/entries/[name=tellico.xml]/view/root/children/[name=collection]/children/[name=entry][@id=3]/children/[name=title]","value":"A Field Guide to Quiet Rivers"}`
REA kept the task because its checks fail on the seed (`title`, `author`, `publisher`, `pub-year`) and all
pass on the captured end state.
To measure a solve rate, run the task k times (3 by default) with the agent and model to measure, and count the runs
whose summary line says `passed`; a run that says `failed` did not finish and is run again rather than counted:
agent-env run <this folder> --task add-glass-orchard --agent claude-code --model MODEL
|