earakely-scale/HomeboxEnv-REA

homebox

homebox

An AgentEnv bundle with its recorded runs, published with agentenv-hf.

Tables

• homebox_tasks: one row per task of the bundle, with its prompt, envs, verifiers and steps.
• homebox_episodes: one row per recorded run, with its status, model, reward and scores, the prompt and response, and the agent's transcript as chat messages when its trajectory format is one agentenv-hf reads.

raw/homebox.jsonl holds each run's record and native trajectory, and bundles/homebox/ the bundle itself.

Run it

pip install agentenv-framework huggingface_hub
hf download earakely-scale/HomeboxEnv-REA --repo-type dataset --local-dir agentenv-dataset
agent-env run agentenv-dataset/bundles/homebox

Homebox on AgentEnv, from a REA capture

What this is. A replay environment of Homebox 0.27.0 (AGPL-3.0), built from one recorded browser session against a self-hosted instance seeded with synthetic data. REA projected the capture into an environment manifest (28 tools), and rea environment-export wrote the bundle you can run from this dataset. Every tool call replays a retained record from the capture, cites the record and the Evidence it came from, and nothing executes Homebox. Tasks are graded deterministically, with no judge.

What's in it

• 28 tools (24 GET, 2 PUT, 1 POST, 1 DELETE) over 91 recorded exchanges on http://127.0.0.1:3208; no recorded group excluded.
• Path parameters: 7 uuid. REA's four format rules (numeric, uuid, hex, iso-date) applied unconditionally; the three opt-in rules (after-uid-literal, opaque-token, sibling-variance) were requested and are labelled per parameter. With the four format rules alone the manifest has 28 tools.
• Unknowns the manifest states explicitly: authorization model, response semantics beyond recorded exchanges, state transitions between tools.
• 10 retrieval tasks (evals/homebox-retrieval.toml): each asks one factual question about a seeded object, with a distinct answer that the app does not ship by default, answerable through a recorded exchange; reward 1 if the response contains the recorded value, else 0. Plus smoke, which deploys the environment and checks replay through the MCP endpoint.
• 20 recorded episodes by openai/gpt-oss-120b:groq, zai-org/GLM-5.3-Flash:zai-org through the Hugging Face router, with their chat transcripts.
• Provenance: capture ev_cb9cef1bb6902484524658e34cfb3aa220e1f06dd807e8da93c2089899872e1c (scoped HAR sha256 249dbf323faaf79dc6a91d856dc6e7ebc494f2e8bfa3b15af85de7af0bcbf2cf, 256,047 bytes), manifest view ev_12cd79fbad0e4c228684699815fdee9521ba2b9ede6c2b891c504603d1e4e457, confidence inferred, authority historical-reference. The REA build is a prerelease of the environment commands proposed upstream in morluto/rea#4954, #4955, #4975 and #4999; the bundle installs it from the tarball it ships, so the environment runs without waiting for an npm release.

What replay means, and does not

A tool call returns the recorded response whose path and query values best match the arguments, with replay.exact_match telling you whether it matched exactly or fell back to the closest record. Reads of seeded objects behave like the application. Writes return the recorded answer of the one write the session made, and the environment has no state: a created object does not appear in a later list. This is what makes retrieval tasks hard-gradable and write tasks weak; the manifest's unknowns say so.

Tasks and grading

task	tool	arguments	field	answer
qa-01-get_templates_by_uuid	get_templates_by_uuid	uuid=41e97d1f-e29b-4a97-a942-e8b8777e142e	description	Template for parts kept for the lamp room.
qa-02-get_entities_by_uuid	get_entities_by_uuid	uuid=3c7d1d3d-a9b0-4639-ac29-7443dbea61da	children[0].description	Locked stone room under the depot where the audit ledgers are kept.
qa-03-get_entities_maintenance_by_uuid	get_entities_maintenance_by_uuid	uuid=a607103c-56c1-469e-86b3-af4fad212416, status=scheduled	[0].name	Recalibrate tide gauge
qa-04-get_templates	get_templates	no arguments	[0].name	Lighthouse Spare Part
qa-05-get_maintenance	get_maintenance	status=scheduled	[0].description	Zero against the Granite Point benchmark before the spring census.
qa-06-get_entities	get_entities	filterChildren=false, isLocation=true	items[0].entityType.name	Storage Room
qa-07-get_groups	get_groups	no arguments	name	Copperfield Annex Stores
qa-08-get_templates_by_uuid	get_templates_by_uuid	uuid=41e97d1f-e29b-4a97-a942-e8b8777e142e	defaultTags[0].name	Spare Parts
qa-09-get_entities_maintenance_by_uuid	get_entities_maintenance_by_uuid	uuid=a607103c-56c1-469e-86b3-af4fad212416, status=scheduled	[0].itemName	Tide Gauge Logger
qa-10-get_entities	get_entities	filterChildren=false, isLocation=true	items[0].description	Cold pantry next to the pass; brassware lives on the top shelf.

Episodes

model	tasks	passed	pass rate	mean tool calls	mean input tokens
openai/gpt-oss-120b:groq	10	10	100%	1.0	8,002
zai-org/GLM-5.3-Flash:zai-org	10	10	100%	1.0	9,777

Every episode is in the episodes config with its transcript; rewards are the deterministic verifier scores.

Every recorded episode passed.

Run it yourself

uv tool install agentenv-framework --with 'agentenv-hf @ git+https://github.com/earakely-scale/agentenv-hf-plugin@v0.3.0'
agent-env hf run earakely-scale/HomeboxEnv-REA --bundle homebox --task smoke          # Docker, no model
agent-env hf run earakely-scale/HomeboxEnv-REA --bundle homebox --eval homebox-retrieval --model openai/gpt-oss-120b:groq

The retrieval eval needs a model endpoint: set [model] base_url and api_key in .agentenv/config.toml (the Hugging Face router works with any hf_ token). The bundle's agents/replay-qa is a small OpenAI-compatible tool-calling agent; any A2A agent can replace it.

How it was made

1. Homebox ran in Docker on a throwaway VM, bound to loopback, seeded through its own API with synthetic records (user Rea Admin in the collection Copperfield Annex Stores (EUR); default tags and locations removed; 2 custom entity types (Storage Room, Spare Part); 5 tags (Lighthouse Rota, Calibration Due, Kitchen Brass, Spare Parts with nested Lens Spares); 4 locations (Saltmarsh Depot with nested Granite Ledger Vault, Copperfield Annex Workshop, Blue Heron Kitchen Pantry); 6 items with manufacturer, model, serial, price, notes and custom fields; 3 maintenance entries; 1 item template); people and projects are invented.
2. Playwright drove a login, the main pages and a scripted read, write, read sequence through the app's JSON API; the HAR was scoped to the API, the login exchange dropped, secret-looking body fields scrubbed, headers and cookies removed.
3. rea inspect-web-network-capture retained the exchanges as Evidence; rea inspect-analysis-view projected the environment-manifest view; rea environment-export wrote this bundle; gen_tasks.py derived the retrieval tasks from the manifest and the recorded bodies; agent-env run recorded the episodes; agent-env hf publish wrote this dataset. Scripts: https://github.com/earakely-scale/rea-environments.

Compared with PortSimEnv on AgentEnv

	PortSimEnv-AgentEnv	Homebox (this dataset)
Environment	A simulation: a container quay, ships, cranes, weather, with A2A ship agents	A replay of a real application's API, captured from one session
Observation	Tool results computed by the simulator as the week unfolds	Recorded responses, cited to Evidence
Action space	Planning tools (check_plan, confirm_berths, advance, message_agent)	28 tools mirroring the application's endpoints
State	Full: a virtual clock, triggers, frozen windows, agents that answer	None: every call is independent; writes replay one recorded answer
Reward	Deterministic, against a CP-SAT optimum or a re-planner anchor; continuous 0 to 1	Deterministic, recorded-value match; 0 or 1 per task
What it measures	Planning under uncertainty and negotiation	Tool discovery, parameter inference and faithful retrieval in a real API
Data provenance	Real 2024 port calls and weather; simulated workloads	Synthetic seed data on a self-hosted Homebox; nothing real
How it was built	Hand-written plugin, months of work	Generated from a capture in one scripted pass
Where it is strong	Long-horizon decisions with hard rewards	Breadth: any app you can capture becomes an environment

The two are complementary: PortSimEnv shows what a hand-built simulation can grade; this dataset shows how far a capture alone gets you, and the manifest's unknowns list exactly what a stateful version would still have to specify.

Licence and credits

Bundle code and tasks: Apache-2.0. Recorded content: responses generated by a self-hosted Homebox instance over synthetic data; Homebox is AGPL-3.0 software by the Homebox contributors (sysadminsmedia) and is not redistributed here beyond the recorded JSON responses. REA is MIT software by morluto; AgentEnv Framework is Apache-2.0 by Scale AI.