earakely-scale/GrafanaEnv-REA

grafana

grafana

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

Tables

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

Run it

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

Grafana on AgentEnv, from a REA capture

What this is. An environment of Grafana 13.2.4 (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 (46 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 Grafana. 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

• 46 tools (40 GET, 3 POST, 2 DELETE, 1 PUT) over 137 recorded exchanges on http://127.0.0.1:3101; no recorded group excluded. 42 are offered; the other 4 are write tools the environment cannot apply (post_dashboards_db, delete_dashboards_by_uid, post_folders, put_folders_by_folder_id), left out rather than faked.
• Path parameters: 5 after-uid-literal, 4 numeric, 3 sibling-variance, 2 opaque-token. 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 56 tools.
• State: 56 entities of 18 types. 37 read tools answer from it and 4 replay their recorded responses. 5 list tools also answer filters, search, sorting or paging from it. Served write tools: delete_folders_by_folder_id (delete).
• Unknowns the manifest states explicitly: authorization model, response semantics beyond recorded exchanges, state transitions between tools.
• 12 retrieval tasks (evals/grafana-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.
• 2 write tasks (evals/grafana-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 4 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_93fa35400726ef2cd92b9b4d4a076e703d8e9a505a55ddff99a4ebadd13e9bb2 (scoped HAR sha256 ca2c90d96540f46e403ee338e189dc08a09bfd4d54a4a1d65a2b0f13f680cbfe, 262,515 bytes), manifest view ev_b723cdd01dcb820397b18faec73e22eac0e6e555f4b4d6adf6912500d8805c9c, 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_users_by_id	get_users_by_id	id=2	email	ada.finch@example.test
qa-02-get_playlists_by_uid	get_playlists_by_uid	uid=dg20xa0afvawwf	name	Lighthouse Rotation
qa-03-get_users_teams_by_id	get_users_teams_by_id	id=2	[0].email	night-shift@example.test
qa-04-get_dashboards_by_uid	get_dashboards_by_uid	uid=granite-ledger	meta.slug	granite-ledger-overview
qa-05-get_teams_members_by_id	get_teams_members_by_id	id=1	[0].name	Ada Finch
qa-06-get_datasources_by_uid	get_datasources_by_uid	uid=rea-testdata	name	Rea Testdata
qa-07-get_dashboards_versions_by_uid	get_dashboards_versions_by_uid	uid=granite-ledger	versions[0].message	Initial save
qa-08-get_teams_by_id	get_teams_by_id	id=1	name	Night Shift
qa-09-get_org_users	get_org_users	no arguments	[0].login	ada.finch
qa-10-get_dashboard_dashboards_blue_heron_kitc	get_dashboard_dashboards_blue_heron_kitchen_dto	no arguments	spec.title	Blue Heron Kitchen Ops
qa-11-get_dashboard_dashboards_granite_ledger_	get_dashboard_dashboards_granite_ledger_dto	no arguments	spec.timeSettings.from	now-6h
qa-12-get_dashboard	get_dashboard	no arguments	name	dashboard.grafana.app

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-delete	On this Grafana server, delete the folder "Lighthouse Rota".	the folder Lighthouse Rota is gone; no folder remains
write-delete-lookup	On this Grafana server, delete the folder that holds the Granite Ledger Overview dashboard.	the folder holding Granite Ledger Overview, Lighthouse Rota, is gone; no folder remains

Episodes

model	retrieval episodes	passed	pass rate	mean tool calls	mean input tokens
openai/gpt-oss-120b:groq	12	12	100%	1.0	18,130
zai-org/GLM-5.3-Flash:zai-org	12	12	100%	1.0	21,018

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

model	write episodes	passed
zai-org/GLM-5.3-Flash:zai-org	2	2
openai/gpt-oss-120b:groq	2	1

In Grafana, deleting a folder also deletes the dashboards inside it; the served folder delete removes the folder only, since side effects on other entities are not modelled, so both write tasks are graded on the folder alone. In write-delete, openai/gpt-oss-120b:groq never reached the environment: Groq rejected its first tool call, get_folders with limit and page set to null, as not matching the tool's schema (HTTP 400).

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/GrafanaEnv-REA --bundle grafana --task smoke          # Docker, no model
agent-env hf run earakely-scale/GrafanaEnv-REA --bundle grafana --eval grafana-retrieval --model openai/gpt-oss-120b:groq
agent-env hf run earakely-scale/GrafanaEnv-REA --bundle grafana --eval grafana-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. Grafana ran in Docker on a throwaway VM, bound to loopback, seeded through its own API with synthetic records (folders: 1; dashboards: 2; panels: 3; datasources: 1; teams: 1; extra users: 1; annotations: 2; playlists: 1; ids: grafana version: 13.2.4; folder uid: lighthouse-rota; datasource uid: rea-testdata; dashboard uids: granite-ledger, blue-heron-kitchen; team id: 1; team name: Night Shift; user id: 2; user login: ada.finch; playlist name: Lighthouse Rotation; annotation tags: seed); people and projects are invented.
2. Playwright drove a login, the main pages and a scripted read sequence through the app's JSON API, then wrote two scratch objects per write target (created, then updated and deleted where the app allows), reading each write back at once; 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	Grafana (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)	42 tools mirroring the application's endpoints
State	Full: a virtual clock, triggers, frozen windows, agents that answer	A state extracted from the capture (56 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 Grafana; 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 Grafana instance over synthetic data; Grafana is AGPL-3.0 software by Grafana Labs 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.