earakely-scale/MattermostEnv-REA

mattermost

mattermost

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

Tables

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

Run it

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

Mattermost on AgentEnv, from a REA capture

What this is. A replay environment of Mattermost 12.0.0 (Team Edition, MIT-licensed server core), built from one recorded browser session against a self-hosted instance seeded with synthetic data. REA projected the capture into an environment manifest (37 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 Mattermost. Tasks are graded deterministically, with no judge.

What's in it

• 37 tools (28 GET, 6 POST, 2 PUT, 1 DELETE) over 129 recorded exchanges on http://127.0.0.1:3104; 0 recorded groups excluded as UI pages, assets or other origins.
• Path parameters: 32 opaque-token, 1 numeric. 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. Without them the manifest has 58 tools, one per literal path.
• Unknowns the manifest states explicitly: authorization model, response semantics beyond recorded exchanges, state transitions between tools.
• 12 retrieval tasks (evals/mattermost-retrieval.toml): each asks one factual question answerable through exactly one 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.
• 24 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_7cc07ebe56003462bbd033d9c12c3a92d92badddff6f9199083c68e9a60357ba (scoped HAR sha256 e5874c112d0ab28d94c68595073b1e0512f0cadc8a47b5bf40914ef8d4cf470f, 322,821 bytes), manifest view ev_97f3c9f89c92adf208d73e8e8e5de3dcb34c5a19fac09fbb29280d423f3f5342, 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_users_teams_threads_by_uid_uid_2_uid	get_users_teams_threads_by_uid_uid_2_uid_3	uid=hhgpypruffff8mez386q9urrpy, uid_2=en1a6qfcw7yeibjonny5um1xny, uid_3=z5qq993jxjgq8fachqwktxpp6w, extended=true	participants[0].first_name	Ada
qa-02-get_teams_channels_name_granite_ledger_b	get_teams_channels_name_granite_ledger_by_uid	uid=en1a6qfcw7yeibjonny5um1xny	purpose	Ledger reconciliation for the Lighthouse Rota crew
qa-03-get_channels_posts_by_uid	get_channels_posts_by_uid	uid=3wg9nryb87b7dm63amxot84gua, per_page=30	posts.jwhjss3f5t8y7nou4nk1pxiupa.message	The dock receipt counts the display barrel; the ledger is right at 18.
qa-04-get_channels_by_uid	get_channels_by_uid	uid=3wg9nryb87b7dm63amxot84gua	name	granite-ledger
qa-05-get_teams_channels_by_uid	get_teams_channels_by_uid	uid=en1a6qfcw7yeibjonny5um1xny	[0].display_name	Blue Heron Kitchen
qa-06-get_posts_thread_by_uid	get_posts_thread_by_uid	uid=jwhjss3f5t8y7nou4nk1pxiupa, skipFetchThreads=false, collapsedThreads=true, collapsedThreadsExtended=false, direction=down, perPage=60	posts.jwhjss3f5t8y7nou4nk1pxiupa.message	The dock receipt counts the display barrel; the ledger is right at 18.
qa-07-get_teams_by_uid	get_teams_by_uid	uid=en1a6qfcw7yeibjonny5um1xny	display_name	Lighthouse Rota
qa-08-get_teams	get_teams	page=0, per_page=200, include_total_count=false, exclude_policy_constrained=false, for_directory=true	[0].display_name	Lighthouse Rota
qa-09-get_users_username_ada	get_users_username_ada	no arguments	email	ada.finch@example.test
qa-10-get_users_me_channels	get_users_me_channels	no arguments	[0].display_name	Night Shift
qa-11-get_teams_name_lighthouse_rota	get_teams_name_lighthouse_rota	no arguments	email	rea-admin@example.test
qa-12-get_users_username_rui	get_users_username_rui	no arguments	username	rui.castellanos

Episodes

model	tasks	passed	pass rate	mean tool calls	mean input tokens
openai/gpt-oss-120b:groq	12	11	92%	0.9	11,679
zai-org/GLM-5.3-Flash:zai-org	12	12	100%	1.0	14,376

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

openai/gpt-oss-120b:groq missed qa-01-get_users_teams_threads_by_uid_uid_2_uid. The miss is a model error, not a replay error: on qa-01 the model called a tool named get_users_teams_threads_by_uid_uid_3, which does not exist (the three-parameter route is get_users_teams_threads_by_uid_uid_2_uid), the provider rejected the call with a 400 and the episode ended without an answer. Generated names for routes with several same-rule parameters are the hard part of this environment.

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/MattermostEnv-REA --bundle mattermost --task smoke          # Docker, no model
agent-env hf run earakely-scale/MattermostEnv-REA --bundle mattermost --eval mattermost-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. Mattermost ran in Docker on a throwaway VM, bound to loopback, seeded through its own API with synthetic records (see the capture notes); 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	Mattermost (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)	37 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 Mattermost; 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 Mattermost instance over synthetic data; Mattermost is Team Edition, MIT-licensed server core software by Mattermost, Inc. 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.