earakely-scale/KomgaEnv-REA

komga

komga

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

Tables

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

Run it

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

Komga on AgentEnv, from a REA capture

What this is. An environment of Komga 1.29.1 (MIT), built from one recorded browser session against a self-hosted instance seeded with synthetic data. REA projected the capture into an environment manifest (46 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 Komga. 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

• 46 tools (37 GET, 5 POST, 2 DELETE, 2 PATCH) over 116 recorded exchanges on http://127.0.0.1:3231; no recorded group excluded. 43 are offered; the other 3 are write tools the environment cannot apply (post_readlists, delete_readlists_by_readlist_id, patch_readlists_by_readlist_id), left out rather than faked.
• Path parameters: 11 sibling-variance, 7 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 64 tools.
• State: 45 entities of 13 types. 26 read tools answer from it and 14 replay their recorded responses. Served write tools: post_collections (create), delete_collections_by_collection_id (delete), patch_collections_by_collection_id (update).
• Unknowns the manifest states explicitly: authorization model, response semantics beyond recorded exchanges, state transitions between tools.
• 12 retrieval tasks (evals/komga-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.
• 4 write tasks (evals/komga-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 8 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_5d1edee16d3181a6cb765a258edf8a455e0f669980a1b3943b5a465d859db5a8 (scoped HAR sha256 4486c491d956d4a28b83a739ed6dd788f6bbb8f3790f73186583a35075870d9e, 254,388 bytes), manifest view ev_e7fb1c64b18223b23787a62a5b529fe31467620f24f036832803bb08297d020e, 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. 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, except that a list the capture saw only filtered or in part keeps its recorded members: it drops a deleted entity but does not add 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, 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. Replayed against a live Komga 1.29.1 seeded by the same script, 9 of 12 calls agreed: creating, renaming, reordering and deleting collections and the reads after them matched apart from the generated id. The collection list after a create omits the new collection, and a deleted collection's 404 repeats the recorded body.

Tasks and grading

Retrieval

task	tool	arguments	field	answer
qa-01-get_collections_by_collection_id	get_collections_by_collection_id	collection_id=0RZ51KS97A6Q5	name	Quarry and Depot Files
qa-02-get_series_books_by_sery_id	get_series_books_by_sery_id	sery_id=0RZ51K8GQACTA	content[0].name	Lighthouse Rota 001
qa-03-get_readlists_books_0rz51k8gka231_next_b	get_readlists_books_0rz51k8gka231_next_by_readlist_id	readlist_id=0RZ51KSZKA53J	name	Lighthouse Rota 002
qa-04-get_readlists_books_by_readlist_id	get_readlists_books_by_readlist_id	readlist_id=0RZ51KSZKA53J	content[0].name	Granite Ledger 001
qa-05-get_series_by_sery_id	get_series_by_sery_id	sery_id=0RZ51K8GFA1KA	metadata.summary	Quarry audits and a ledger that never quite balances, told by its auditor.
qa-06-get_readlists_books_0rz51k8gva36d_next_b	get_readlists_books_0rz51k8gva36d_next_by_readlist_id	readlist_id=0RZ56NR37AC1T	name	Blue Heron Kitchen 002
qa-07-get_collections_series_by_collection_id	get_collections_series_by_collection_id	collection_id=0RZ51KRY3A8P5	content[0].metadata.summary	Recipes that follow the tides in a small fishing kitchen.
qa-08-get_books_readlists_by_book_id	get_books_readlists_by_book_id	book_id=0RZ51K8GFA1KB	[0].name	Granite Ledger audit order
qa-09-get_books_pages_by_book_id	get_books_pages_by_book_id	book_id=0RZ51K8GVA36D	[0].fileName	001.png
qa-10-get_books_next_by_book_id	get_books_next_by_book_id	book_id=0RZ51K8GQACTB	metadata.title	Shift Three
qa-11-get_books_by_book_id	get_books_by_book_id	book_id=0RZ51K8GQACTF	metadata.title	Kelp Broth Night
qa-12-get_readlists_by_readlist_id	get_readlists_by_readlist_id	readlist_id=0RZ51KSZKA53J	summary	Read both quarry issues before the rota winter, as Rui Castellanos filed them.

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 Komga library, create a new collection named "Tide Table Shelf" that contains the series Granite Ledger and Blue Heron Kitchen, and is not ordered.	a collection named Tide Table Shelf exists; it holds Granite Ledger and Blue Heron Kitchen; it is not ordered; the two seeded collections are still there
write-update	In this Komga library, rename the collection "Night Shift Reading" to "Night Watch Reading" and make it ordered.	the collection is now called Night Watch Reading; it is the same collection; it is ordered; no collection was added or removed
write-delete	In this Komga library, delete the collection "Quarry and Depot Files", and only that collection.	Quarry and Depot Files is gone; Night Shift Reading remains
write-create-then-update	In this Komga library, create a collection named "Saltmarsh Shelf" with the series Lighthouse Rota, not ordered. Then rename that new collection to "Saltmarsh Depot Shelf".	the new collection is called Saltmarsh Depot Shelf; it holds Lighthouse Rota; exactly one collection was added

Episodes

model	retrieval episodes	passed	pass rate	mean tool calls	mean input tokens
openai/gpt-oss-120b:groq	12	12	100%	1.1	20,084
zai-org/GLM-5.3-Flash:zai-org	12	12	100%	1.0	22,769

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	4	3
zai-org/GLM-5.3-Flash:zai-org	4	4

gpt-oss-120b missed write-update: the model provider rejected a tool call the model generated (null for optional query arguments), so the episode ended without an answer.

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/KomgaEnv-REA --bundle komga --task smoke          # Docker, no model
agent-env hf run earakely-scale/KomgaEnv-REA --bundle komga --eval komga-retrieval --model openai/gpt-oss-120b:groq
agent-env hf run earakely-scale/KomgaEnv-REA --bundle komga --eval komga-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. Komga ran in Docker on a throwaway VM, bound to loopback, seeded through its own API with synthetic records (server claimed by rea-admin@example.test; library Copperfield Annex Shelf with 3 series (Lighthouse Rota, Granite Ledger, Blue Heron Kitchen) and 7 generated CBZ issues whose ComicInfo.xml names invented writers, artists and publishers (Ada Finch, Rui Castellanos, Mina Okafor, Tomas Vey, Saltmarsh Depot Press, Copperfield Annex Comics); series summaries, statuses, age ratings and tags; collections Night Shift Reading and Quarry and Depot Files; read list Granite Ledger audit order; users Ada Finch and Rui Castellanos; read progress on 2 books); 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; 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	Komga (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)	43 tools mirroring the application's endpoints
State	Full: a virtual clock, triggers, frozen windows, agents that answer	A state extracted from the capture (45 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 Komga; 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 Komga instance over synthetic data; Komga is MIT software by Gauthier Roebroeck and the Komga 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.