earakely-scale/GiteaEnv-REA

gitea

gitea

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

Tables

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

Run it

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

Gitea on AgentEnv, from a REA capture

What this is. A replay environment of Gitea 28.1.0 (MIT), built from one recorded browser session against a self-hosted instance seeded with synthetic data. REA projected the capture into an environment manifest (77 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 Gitea. Tasks are graded deterministically, with no judge.

What's in it

• 77 tools (73 GET, 2 POST, 1 DELETE, 1 PATCH) over 97 recorded exchanges on http://127.0.0.1:3102; 0 recorded groups excluded as UI pages, assets or other origins.
• Path parameters: 19 numeric, 2 hex. 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 77 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/gitea-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_1a2a26cf486e359cd4ee757d73a47f3544f937fc2877a815de9b3416349c2556 (scoped HAR sha256 40801721c583336d92f7bd882767095a8c682bf1b9880ea62c6dce72fc1c186b, 310,109 bytes), manifest view ev_989ab6936d49511fae9956a43c67254d91e48eb263fd558d04133a50cf86344b, 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_repos_lighthouse_rota_granite_ledger	get_repos_lighthouse_rota_granite_ledger_git_commits_by_hash	hash=22bd3ff64b6db3b468148e078db6b9d77d2e7982	author.full_name	Ada Finch
qa-02-get_repos_lighthouse_rota_blue_heron_kit	get_repos_lighthouse_rota_blue_heron_kitchen_issues_by_id	id=1	user.login	rea-admin
qa-03-get_repos_lighthouse_rota_granite_ledger	get_repos_lighthouse_rota_granite_ledger_issues_comments_by	id=1	[0].user.login	rea-admin
qa-04-get_repos_lighthouse_rota_granite_ledger	get_repos_lighthouse_rota_granite_ledger_issues_labels_by_id	id=1	[0].name	high-tide
qa-05-get_teams_members_by_id	get_teams_members_by_id	id=2	[0].username	ada.finch
qa-06-get_repos_lighthouse_rota_granite_ledger	get_repos_lighthouse_rota_granite_ledger_pulls_files_by_id	id=4	[0].filename	docs/spring-tide-shifts.md
qa-07-get_repos_lighthouse_rota_granite_ledger	get_repos_lighthouse_rota_granite_ledger_issues_timeline_by	id=2	[0].milestone.title	Spring Tide
qa-08-get_repos_lighthouse_rota_blue_heron_kit	get_repos_lighthouse_rota_blue_heron_kitchen_issues_comments	id=2	[0].body	Counted 14 tins of chopped tomatoes.
qa-09-get_teams_repos_by_id	get_teams_repos_by_id	id=2	[0].full_name	lighthouse-rota/blue-heron-kitchen
qa-10-get_repos_lighthouse_rota_granite_ledger	get_repos_lighthouse_rota_granite_ledger_pulls_by_id	id=4	base.repo.full_name	lighthouse-rota/granite-ledger
qa-11-get_repos_lighthouse_rota_granite_ledger	get_repos_lighthouse_rota_granite_ledger_issues_by_id	id=1	repository.full_name	lighthouse-rota/granite-ledger
qa-12-get_repos_lighthouse_rota_granite_ledger	get_repos_lighthouse_rota_granite_ledger_labels_by_id	id=1	description	Blocks the next lamp-room handover

Episodes

model	tasks	passed	pass rate	mean tool calls	mean input tokens
openai/gpt-oss-120b:groq	12	12	100%	1.1	18,321
zai-org/GLM-5.3-Flash:zai-org	12	11	92%	1.1	21,569

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

zai-org/GLM-5.3-Flash:zai-org missed qa-08-get_repos_lighthouse_rota_blue_heron_kit.

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/GiteaEnv-REA --bundle gitea --task smoke          # Docker, no model
agent-env hf run earakely-scale/GiteaEnv-REA --bundle gitea --eval gitea-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. Gitea ran in Docker on a throwaway VM, bound to loopback, seeded through its own API with synthetic records (users ada.finch, rui.castellanos; org lighthouse-rota (team keepers); repos granite-ledger (README, ledger/2026-q1.csv, docs/rota.md) and blue-heron-kitchen; 2 labels; milestone Spring Tide; 4 issues (2 labelled, 1 assigned, 1 closed); 2 comments; 1 branch + 1 PR on granite-ledger); 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	Gitea (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)	77 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 Gitea; 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 Gitea instance over synthetic data; Gitea is MIT software by the Gitea 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.