earakely-scale/ShlinkEnv-REA

shlink

shlink

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

Tables

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

Run it

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

Shlink on AgentEnv, from a REA capture

What this is. An environment of Shlink 5.1.7 (MIT), built from one recorded browser session against a self-hosted instance seeded with synthetic data. REA projected the capture into an environment manifest (17 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, no write tool is offered because the capture confirmed none, and data/reset restores it. Nothing executes Shlink. Tasks are graded deterministically, with no judge.

v0.2.0 rebuilds this environment from the same capture with a REA build that serves a state and confirmed writes; the tag v0.1.0 keeps the earlier replay-only version.

What's in it

• 17 tools (15 GET, 1 POST, 1 PATCH) over 66 recorded exchanges on http://127.0.0.1:3206; 1 recorded group excluded (1 as non-JSON replies such as bodiless 204 writes). 15 are offered; the other 2 are write tools the environment cannot apply (post_rest_short_urls, patch_rest_short_urls_by_short_url_id), left out rather than faked.
• Path parameters: 4 sibling-variance, 1 prefix-variance. 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 43 tools.
• State: 8 entities of 1 types. 5 read tools answer from it and 10 replay their recorded responses. No write tool is served.
• Unknowns the manifest states explicitly: authorization model, response semantics beyond recorded exchanges, state transitions between tools.
• 7 retrieval tasks (evals/shlink-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.
• 14 retrieval episodes, plus the smoke run, by zai-org/GLM-5.3-Flash:zai-org, openai/gpt-oss-120b:groq through the Hugging Face router, with their chat transcripts.
• Provenance: capture ev_4779d529d62305de670c032f0cdd9188c95cf86f0b549b9a0fac8554db69142b (scoped HAR sha256 97a0390d535b7d6f2555d14a7402f5785ac85825b7077236bcfcc8ccbbc1714c, 117,350 bytes), manifest view ev_d3754c8da879ed80d0caf8688871eca4bcfbf0bb1e5bfaea6d1dbffe8210169f, confidence inferred, authority historical-reference. The REA build is rea-agents 6.7.0 from main 1629266e with open pull requests merged: morluto/rea#5600 at 370b9bef (which carries #5351 and #5352) and #5664 at 930c96d0. 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. 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. No write tool is offered: the app returned a short URL's tags sorted rather than in the order they were sent, so later reads contradicted the modelled create and update, so the capture confirmed no write's effect and the environment is read-only. 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, new entities in lists whose recorded calls were filtered, 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_rest_short_urls_visits_by_short_url_	get_rest_short_urls_visits_by_short_url_id	short_url_id=granite-ledger	visits.data[0].userAgent	KeeperConsole/2.1 (Granite Point Light)
qa-02-get_rest_short_urls_by_short_url_id	get_rest_short_urls_by_short_url_id	short_url_id=granite-ledger	title	Granite Ledger quarterly stone count
qa-03-get_rest_tags_visits_by_tag_id	get_rest_tags_visits_by_tag_id	tag_id=blue-heron-kitchen	visits.data[0].userAgent	GalleyTablet/1.4 (Blue Heron Kitchen)
qa-04-get_rest_visits_non_orphan	get_rest_visits_non_orphan	no arguments	visits.data[0].userAgent	TideTableReader/3.2 (Copperfield Annex)
qa-05-get_rest_visits_orphan	get_rest_visits_orphan	no arguments	visits.data[0].userAgent	LostLantern/1.1 (mistyped short code)
qa-06-get_rest_short_urls	get_rest_short_urls	orderBy=title-ASC	shortUrls.data[0].title	Blue Heron Kitchen spring-tide menu
qa-07-get_rest_domains_links_visits	get_rest_domains_links_visits	no arguments	visits.data[0].userAgent	DepotScanner/0.9 (Saltmarsh Depot quay)

Episodes

model	retrieval episodes	passed	pass rate	mean tool calls	mean input tokens
zai-org/GLM-5.3-Flash:zai-org	7	7	100%	1.0	8,407
openai/gpt-oss-120b:groq	7	7	100%	1.0	7,009

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

The gpt-oss-120b episodes are its second retrieval run; the first missed qa-04-get_rest_visits_non_orphan and qa-05-get_rest_visits_orphan because the model provider rejected tool calls the model generated (undefined arguments).

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/ShlinkEnv-REA --bundle shlink --task smoke          # Docker, no model
agent-env hf run earakely-scale/ShlinkEnv-REA --bundle shlink --eval shlink-retrieval --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. Shlink ran in Docker on a throwaway VM, bound to loopback, seeded through its own API with synthetic records (api key: one key named rea-capture from shlink api-key:generate (value only in .admin, deleted at teardown); short urls: granite-ledger: Granite Ledger quarterly stone count (tags granite-ledger, lighthouse-rota; maxVisits 500), tide-tables: January tide tables for Granite Point (crawlable; redirect rules device=android and query shift=night), blue-heron-menu: Blue Heron Kitchen spring-tide menu, depot-manifest on links.saltmarsh.example.test: Saltmarsh Depot lamp-oil manifest SO-4471, annex-rebuild: Copperfield Annex beacon rebuild notes (validUntil 2027-03-31), rota-binder: Lighthouse...); 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 and the state it serves; 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	Shlink (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)	15 tools mirroring the application's endpoints
State	Full: a virtual clock, triggers, frozen windows, agents that answer	A state extracted from the capture (8 entities); read-only; 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; 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 Shlink; 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 Shlink instance over synthetic data; Shlink is MIT software by Alejandro Celaya and the Shlink 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.