earakely-scale/MemosEnv-REA

memos

memos

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

Tables

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

Run it

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

Memos on AgentEnv, from a REA capture

What this is. A replay environment of Memos 0.31.0 (MIT), built from one recorded browser session against a self-hosted instance seeded with synthetic data. REA projected the capture into an environment manifest (26 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 Memos. Tasks are graded deterministically, with no judge.

What's in it

• 26 tools (23 GET, 1 POST, 1 DELETE, 1 PATCH) over 67 recorded exchanges on http://127.0.0.1:3205; no recorded group excluded.
• Path parameters: 7 uuid, 5 sibling-variance. 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 56 tools.
• Unknowns the manifest states explicitly: authorization model, response semantics beyond recorded exchanges, state transitions between tools.
• 12 retrieval tasks (evals/memos-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.
• 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_c312b72e2468894d9dfbe7bffe1b38ab8a60dfad82d049e36ffd9c199420c31e (scoped HAR sha256 88d94da05d999e68d6b7b3692e0c3bab48b02390921bc66ea2352f39df4cf2de, 113,900 bytes), manifest view ev_9535f9c5f689cd44d904d5b6e3b3462b1bc522be7ae1397e1a2a0eca35296955, 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_by_user_id	get_users_by_user_id	user_id=ada-finch	description	Keeps the tide tables for the winter rota.
qa-02-get_memos_comments_by_memo_id	get_memos_comments_by_memo_id	memo_id=ERVHcKcjsFNbLvBQaFahvB	memos[0].content	I will bring the January tide tables to the handover.
qa-03-get_users_rea_admin_views_by_uuid	get_users_rea_admin_views_by_uuid	uuid=f77a9067-52ff-4017-9a58-2642dd6c752e	title	Granite Ledger notes
qa-04-get_spaces_invitations_by_uuid	get_spaces_invitations_by_uuid	uuid=3d184a5a-0f62-4857-a65c-24f265949f07	spaceInvitations[0].space.title	Saltmarsh Depot Crew
qa-05-get_memos_by_memo_id	get_memos_by_memo_id	memo_id=Fs83ceGx4AHaULNv9zhY2C	content	#blue-heron-kitchen Blue Heron Kitchen serves 36 covers per sitting.
qa-06-get_spaces_by_uuid	get_spaces_by_uuid	uuid=3d184a5a-0f62-4857-a65c-24f265949f07	description	Keepers and quartermasters of the Saltmarsh Depot.
qa-07-get_users	get_users	no arguments	users[0].username	mina-okafor
qa-08-get_users_rea_admin_spaceinvitations	get_users_rea_admin_spaceinvitations	no arguments	spaceInvitations[0].space.title	Blue Heron Kitchen Rota
qa-09-get_users_rea_admin_views	get_users_rea_admin_views	no arguments	memoViews[0].filter	"granite-ledger" in tags
qa-10-get_memos	get_memos	pageSize=20	memos[0].content	#saltmarsh-depot Crew meeting at the Copperfield Annex boathouse, 18:30.
qa-11-get_instance_profile	get_instance_profile	no arguments	admin.description	Keeper of the Lighthouse Rota since the spring census.
qa-12-get_users_by_user_id	get_users_by_user_id	user_id=ada-finch	email	ada.finch@example.test

Episodes

model	tasks	passed	pass rate	mean tool calls	mean input tokens
openai/gpt-oss-120b:groq	12	12	100%	1.0	7,436
zai-org/GLM-5.3-Flash:zai-org	12	12	100%	1.1	8,984

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/MemosEnv-REA --bundle memos --task smoke          # Docker, no model
agent-env hf run earakely-scale/MemosEnv-REA --bundle memos --eval memos-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. Memos ran in Docker on a throwaway VM, bound to loopback, seeded through its own API with synthetic records (users: rea-admin (Rea Admin, ADMIN, first account), ada-finch (Ada Finch), rui-castellanos (Rui Castellanos), mina-okafor (Mina Okafor); memos: rea-admin: winter rota (PROTECTED, pinned, #lighthouse-rota), stone count GL-2291 (PUBLIC, #granite-ledger), Blue Heron Kitchen 36 covers (PUBLIC, location 'Blue Heron Spit boathouse'), lamp-oil order SO-4471 (PRIVATE, #saltmarsh-depot), Granite Ledger checklist (task list, references the stone count), Copperfield Annex beacon (ARCHIVED), crew meeting (SPACE visibility, in space Saltmarsh Depot Crew); ada-finch: January low water 04:12 (PROTECTED);...); 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	Memos (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)	26 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 Memos; 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 Memos instance over synthetic data; Memos is MIT software by the Memos contributors 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.