earakely-scale/SonarrEnv-REA

sonarr

sonarr

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

Tables

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

Run it

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

Sonarr on AgentEnv, from a REA capture

What this is. An environment of Sonarr 4.0.20.3014 (GPL-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 (50 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 Sonarr. 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

• 50 tools (44 GET, 2 POST, 2 DELETE, 2 PUT) over 146 recorded exchanges on http://127.0.0.1:3221; no recorded group excluded. 49 are offered; the other one is a write tool the environment cannot apply (post_tag), left out rather than faked.
• Path parameters: 16 numeric. 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 50 tools.
• State: 123 entities of 15 types. 43 read tools answer from it and 1 replays its recorded response. 1 list tool also answers filters, search, sorting or paging from it. Served write tools: post_releaseprofile (create), delete_releaseprofile_by_id (delete), put_releaseprofile_by_id (update), delete_tag_by_id (delete), put_tag_by_id (update).
• Unknowns the manifest states explicitly: authorization model, response semantics beyond recorded exchanges, state transitions between tools.
• 12 retrieval tasks (evals/sonarr-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.
• 3 write tasks (evals/sonarr-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 6 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_abf64d509b36565e8a7e7eaf3fa5590eb8f51efd48b3b1ec5d8faf885377eb8a (scoped HAR sha256 c9a30b253e7ffda01db9923bc2e027d9305b5aa9b631a3195bfb3a68e030d26f, 397,195 bytes), manifest view ev_d54f5ce7754846fc5f69bf7a38cd5cc9310154bbfb027bfd29540ef8d5442b11, confidence inferred, authority historical-reference. The REA build is rea-agents 6.7.0 from main 3e7e102a with open pull requests merged: morluto/rea#6385 at 342933ea and morluto/rea#6897 at 08c2532a. 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_releaseprofile_by_id	get_releaseprofile_by_id	id=2	name	Tide Table Filter
qa-02-get_customformat_by_id	get_customformat_by_id	id=2	name	Blue Heron Kitchen Subtitles
qa-03-get_qualityprofile_by_id	get_qualityprofile_by_id	id=8	name	Granite Ledger Archive SD
qa-04-get_notification_by_id	get_notification_by_id	id=2	name	Night Shift Pager
qa-05-get_autotagging_by_id	get_autotagging_by_id	id=1	specifications[0].name	Drama genre
qa-06-get_importlistexclusion_by_id	get_importlistexclusion_by_id	id=2	title	The Saltmarsh Depot Hour
qa-07-get_notification	get_notification	no arguments	[0].name	Copperfield Annex Script
qa-08-get_importlistexclusion	get_importlistexclusion	no arguments	[0].title	Granite Ledger Chronicles
qa-09-get_qualityprofile	get_qualityprofile	no arguments	[0].formatItems[0].name	Saltmarsh Depot x265
qa-10-get_autotagging	get_autotagging	no arguments	[0].name	Night Shift Dramas
qa-11-get_customformat	get_customformat	no arguments	[0].name	Granite Ledger Remux
qa-12-get_importlistexclusion_paged	get_importlistexclusion_paged	page=1, pageSize=10, sortKey=title, sortDirection=ascending	records[0].title	Blue Heron Kitchen Nights

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-create	In this Sonarr instance, add an enabled release profile named "Night Shift Repacks" that requires the term "repack" and ignores nothing.	a release profile named Night Shift Repacks exists; it requires the term repack; both seeded release profiles are still there
write-update	In this Sonarr instance, change the release profile "Tide Table Filter" so that it also ignores the term "teaser", keeping the terms it already ignores.	Tide Table Filter now ignores teaser; it still ignores sample and trailer; no release profile was added or removed
write-delete-lookup	In this Sonarr instance, one release profile requires the term "proper". Find it and delete it.	the Copperfield Annex Preferred profile, which requires proper, is gone; the Tide Table Filter profile is still there

Episodes

model	retrieval episodes	passed	pass rate	mean tool calls	mean input tokens
openai/gpt-oss-120b:groq	12	12	100%	1.0	21,174
zai-org/GLM-5.3-Flash:zai-org	12	12	100%	1.0	24,307

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

model	write episodes	passed
openai/gpt-oss-120b:groq	3	3
zai-org/GLM-5.3-Flash:zai-org	3	3

Tag create stays refused (side-effect): Sonarr's UI re-reads /api/v3/tag/detail after each tag create, and the new tag appears in that second collection, which REA does not model. Seeded tags are all in use (notifications, release profiles, delay profile, auto-tagging), so no tag delete task: the environment does not model Sonarr's in-use conflicts.

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/SonarrEnv-REA --bundle sonarr --task smoke          # Docker, no model
agent-env hf run earakely-scale/SonarrEnv-REA --bundle sonarr --eval sonarr-retrieval --model openai/gpt-oss-120b:groq
agent-env hf run earakely-scale/SonarrEnv-REA --bundle sonarr --eval sonarr-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. Sonarr ran in Docker on a throwaway VM, bound to loopback, seeded through its own API with synthetic settings objects (4 tags, 2 root folders, 3 custom formats, 2 quality profiles, 2 release profiles, 1 delay profile, 2 Custom Script connections, 3 import-list exclusions with invented titles, 1 auto-tagging rule); no series, since adding one needs Sonarr's online metadata service.
2. Playwright drove a forms login, the main pages and a scripted read sequence through the app's JSON API, then wrote two scratch tags and two scratch release profiles (created, edited and deleted), edited and deleted a seeded release profile and renamed a seeded tag, reading each write back at once by item and by list; the HAR was scoped to the API, the login exchange and the host settings (which carry the API key) 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	Sonarr (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)	49 tools mirroring the application's endpoints
State	Full: a virtual clock, triggers, frozen windows, agents that answer	A state extracted from the capture (123 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 Sonarr; 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 Sonarr instance over synthetic data; Sonarr is GPL-3.0 software by Team Sonarr 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.