earakely-scale/BaserowEnv-REA

baserow

baserow

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

Tables

• baserow_tasks: one row per task of the bundle, with its prompt, envs, verifiers and steps.
• baserow_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/baserow.jsonl holds each run's record and native trajectory, and bundles/baserow/ the bundle itself.

Run it

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

Baserow on AgentEnv, from a REA capture

What this is. An environment of Baserow 2.4.1 (MIT), built from one recorded browser session against a self-hosted instance seeded with synthetic data. REA projected the capture into an environment manifest (31 tools) and an environment state, the entities the capture's reads showed, and rea environment-export wrote the bundle you can run from this dataset. Reads the state can render are answered from it, write tools whose effect the capture confirmed change it, and data/reset restores it. Nothing executes Baserow. Tasks are graded deterministically, with no judge.

v0.2.0 rebuilds this environment from a new capture of an instance seeded the same way, recorded so each write's effect is shown: two scratch objects per write target, every write read back at once. A REA build that serves a state, confirmed writes and list queries projects it; the tag v0.1.0 keeps the earlier replay-only version.

What's in it

• 31 tools (28 GET, 1 POST, 1 DELETE, 1 PATCH) over 101 recorded exchanges on http://127.0.0.1:3204; no recorded group excluded. 30 are offered; the other 1 are write tools the environment cannot apply (post_database_rows_table_by_id), left out rather than faked.
• Path parameters: 27 numeric. REA's four format rules (numeric, uuid, hex, iso-date) applied unconditionally; the four opt-in rules (after-uid-literal, opaque-token, sibling-variance, prefix-variance) were requested and are labelled per parameter. With the four format rules alone the manifest has 31 tools.
• State: 32 entities of 12 types. 18 read tools answer from it and 10 replay their recorded responses. 1 list tool also answers filters, search, sorting or paging from it. Served write tools: delete_database_rows_table_by_id_id_2 (delete), patch_database_rows_table_by_id_id_2 (update).
• Unknowns the manifest states explicitly: authorization model, response semantics beyond recorded exchanges, state transitions between tools.
• 12 retrieval tasks (evals/baserow-retrieval.toml): each asks one factual question about a seeded object, with a distinct answer that the app does not ship by default; reward 1 if the response contains the recorded value, else 0. Plus smoke, which deploys the environment and checks it through the MCP endpoint.
• 3 write tasks (evals/baserow-writes.toml): each asks for a change in plain language and is graded from the environment's state through data/get, not from the agent's reply; reward 1 only if every check passes. Their grade is in scores.state; reward holds the retrieval verifier's score, so it is empty for them, as for smoke.
• 24 retrieval episodes and 6 write episodes, plus the smoke run, 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_3555cae2a38133dc990dfb01647e99b0b89787e8135765d14a3241eb09400555 (scoped HAR sha256 6aaee5f742ccd03580ffec96ab2c7e60c54396b988b664e97cce93a09225b117, 195,291 bytes), manifest view ev_894a5b51d6794366bf785936e451b3b365a7c1bd304988c3cae60705b161fff4, confidence inferred, authority historical-reference. The REA build is rea-agents 6.7.0 from main 7091be29 with open pull requests merged: morluto/rea#6385 at 28c3ae97. The bundle installs it from the tarball it ships, so the environment runs without waiting for an npm release.

What the state means, and does not

Reads the state can render, item, list and document views of the entities the capture's responses showed, answer from it. List reads filtered, searched, sorted or paged answer from it too wherever the capture's recorded calls explain what each query parameter does, including values no call recorded; a parameter the recordings do not explain, or a flag's unrecorded value, replays. Other reads replay the recorded response whose path and query values best match the arguments, with replay.exact_match saying whether it matched exactly; a replayed read does not change after a write. A served write was confirmed by the capture itself: replaying the session against the extracted state, a later recorded read reproduced its effect and none contradicted it. A create generates ids and clock values deterministically, so the same calls after a reset give the same state; an update changes the members the call sends; a delete removes the entity. Each takes only the body members the recorded call showed and refuses others with a message that names the accepted ones. Its effect shows in the reads the state renders, lists included wherever a list lists its whole collection or a learned query answers the call; a list call nothing explains keeps its recorded members, dropping a deleted entity but not adding a new one. The data plane at /agentenv takes data/reset (restore the extracted state), data/add and data/get (return it), which is how write tasks are graded.

Not modelled: side effects a write has beyond its own entity (counters, copies in other collections, related entities), preconditions the application enforces before a write, list query parameters the recorded calls do not explain, paging members such as page counts and next links when a call asks for another page size or page, and anything the capture never showed. The manifest's unknowns list what a capture cannot establish.

Tasks and grading

Retrieval

task	tool	arguments	field	answer
qa-01-get_applications_by_id	get_applications_by_id	id=19	workspace.name	Saltmarsh Depot Cooperative
qa-02-get_database_views_table_by_id	get_database_views_table_by_id	id=55, include=filters,sortings,group_bys,decorations,default_row_values	[0].name	Keeper roster
qa-03-get_database_tables_by_id	get_database_tables_by_id	id=56	name	Supply Orders
qa-04-get_database_views_grid_by_id	get_database_views_grid_by_id	id=193, limit=80, offset=0, include=field_options,row_metadata, limit_linked_items=20	results[0].field_495	Keeps the tide tables for January.
qa-05-get_database_views_by_id	get_database_views_by_id	id=213	name	Request supplies
qa-06-get_database_views_gallery_by_id	get_database_views_gallery_by_id	id=212, limit=100, offset=0, include=field_options,row_metadata, limit_linked_items=20	results[0].field_493	Granite tower on the north shoal; the oldest station on the rota.
qa-07-get_database_rows_table_by_id_id_2	get_database_rows_table_by_id_id_2	id=56, id_2=3, user_field_names=true	Supplier	Granite Ledger Chandlery
qa-08-get_applications_workspace_by_id	get_applications_workspace_by_id	id=11	[0].name	Lighthouse Rota
qa-09-get_database_rows_table_by_id	get_database_rows_table_by_id	id=55, user_field_names=true	results[0].Email	ada.finch@example.test
qa-10-get_database_rows_names	get_database_rows_names	table__54=3,4,5,6	54.4	Copperfield Annex Beacon
qa-11-get_database_views_grid_by_id	get_database_views_grid_by_id	id=193, limit=80, offset=0, include=field_options,row_metadata, limit_linked_items=20	results[0].field_500.value	Head keeper
qa-12-get_database_views_by_id	get_database_views_by_id	id=213	description	Orders reach the Saltmarsh Depot quartermaster within two tides.

Writes

Graded from the environment's state through data/get (scores.state in the episodes table; reward 1 only if every check passes).

task	prompt	checks
write-update	In this Baserow database, supply order SO-4473 (the fog horn diaphragm) now needs 3 units instead of 2. Change its quantity to 3 and leave everything else as it is.	SO-4473 now has a quantity of 3; its item is still the Fog horn diaphragm; the station and keeper that share its row id are untouched; every station, keeper and order is still there
write-update-lookup	In this Baserow database, one supply order is for lens polishing cloths. Raise its quantity to 36.	SO-4472, the lens polishing cloths, now has a quantity of 36; the other orders keep their quantities; every station, keeper and order is still there
write-delete	In this Baserow database, delete supply order SO-4474 (the tide gauge batteries), but keep the station and keeper it links to.	SO-4474 is gone; the Copperfield Annex Beacon and Tomas Vey are still there; the station and keeper that share its row id are still there; every other station, keeper and order is still there

Episodes

model	retrieval episodes	passed	pass rate	mean tool calls	mean input tokens
openai/gpt-oss-120b:groq	12	12	100%	1.0	14,060
zai-org/GLM-5.3-Flash:zai-org	12	12	100%	1.0	16,548

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

model	write episodes	passed
openai/gpt-oss-120b:groq	3	3
zai-org/GLM-5.3-Flash:zai-org	3	3

The row create (post_database_rows_table_by_id) is withheld: Baserow gives each new row a decimal-string position (order, 5.00000000000000000000 for the first scratch row and 6.00000000000000000000 for the second), REA has no sequence rule for that spelling, so it learned the first create's position as a constant and the second create's read-back contradicted it. The rows list (get_database_rows_table_by_id) replays its recorded lines rather than rendering from the state, so a list call after a write still shows the seeded rows; the row item read (get_database_rows_table_by_id_id_2) renders from the state and shows the change.

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/BaserowEnv-REA --bundle baserow --task smoke          # Docker, no model
agent-env hf run earakely-scale/BaserowEnv-REA --bundle baserow --eval baserow-retrieval --model openai/gpt-oss-120b:groq
agent-env hf run earakely-scale/BaserowEnv-REA --bundle baserow --eval baserow-writes --model openai/gpt-oss-120b:groq

The evals need 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. Baserow ran in Docker on a throwaway VM, bound to loopback, seeded through its own API with synthetic records (owner Rea Admin, the first sign-up; workspace Saltmarsh Depot Cooperative; database Lighthouse Rota with three tables: Stations (4 rows), Keepers (5 rows, linked to Stations, with a Rank single select) and Supply Orders (4 rows SO-4471 to SO-4474, linked to Stations and Keepers, with a Stage single select); one keeper edit so its row history has an entry; grid views Certified keepers (filter and sort) and Open orders (filter), gallery view Station cards and form view Request supplies; an empty second database, Granite Ledger Archive); people and projects are invented.
2. Playwright drove a login, the main pages and a scripted read sequence through the app's JSON API (every table's rows read with field names first, grid-view and include variants after), then created two scratch supply orders, updated and deleted each, and edited one seeded order, reading every write back at once by row and by the whole order list (deletes read back as 404); the HAR was scoped to the API, the login exchange dropped, secret-looking string 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 and the state it serves; gen_tasks.py derived the retrieval tasks from the manifest and the recorded bodies; the write tasks were written by hand against the served write tools and the seeded names; 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	Baserow (this dataset)
Environment	A simulation: a container quay, ships, cranes, weather, with A2A ship agents	An application's API rebuilt from one captured session
Observation	Tool results computed by the simulator as the week unfolds	Answers rendered from the extracted state, or recorded responses, cited to Evidence
Action space	Planning tools (check_plan, confirm_berths, advance, message_agent)	30 tools mirroring the application's endpoints
State	Full: a virtual clock, triggers, frozen windows, agents that answer	A state extracted from the capture (32 entities); served writes change it; data/reset restores it
Reward	Deterministic, against a CP-SAT optimum or a re-planner anchor; continuous 0 to 1	Deterministic: recorded-value match for retrieval, state checks for writes; 0 or 1 per task
What it measures	Planning under uncertainty and negotiation	Tool discovery, parameter inference and faithful retrieval in a real API, and changing its state
Data provenance	Real 2024 port calls and weather; simulated workloads	Synthetic seed data on a self-hosted Baserow; 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 and the state's limits above say what it still leaves out.

Licence and credits

Bundle code and tasks: Apache-2.0. Recorded content: responses generated by a self-hosted Baserow instance over synthetic data; Baserow is MIT software by Baserow B.V. 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.