v0.5.0: v5, the lines port, leads the Space

Dockerfile

@@ -6,7 +6,7 @@ USER user
 ENV HOME=/home/user PATH=/home/user/.local/bin:$PATH
 WORKDIR /home/user/app
 RUN pip install --user "huggingface_hub>=1,<2" \
-    "agentenv-portsim @ https://github.com/earakely-scale/agentenv-portsim-plugin/archive/refs/tags/v0.4.5.tar.gz"
+    "agentenv-portsim @ https://github.com/earakely-scale/agentenv-portsim-plugin/archive/refs/tags/v0.5.0.tar.gz"
 COPY --chown=user start.sh .
 EXPOSE 7860
 CMD ["sh", "start.sh"]

README.md

@@ -9,7 +9,7 @@ base_path: /viewer/
 pinned: true
 license: other
 license_name: cc-by-sa-4.0-cc-by-4.0-meteocat-apache-2.0
-license_link: https://github.com/earakely-scale/agentenv-portsim-plugin/blob/v0.4.5/data/LICENSE
+license_link: https://github.com/earakely-scale/agentenv-portsim-plugin/blob/v0.5.0/data/LICENSE
 short_description: Port of Barcelona weeks on AgentEnv, replayed in 3D
 thumbnail: https://huggingface.co/spaces/earakely-scale/PortSimEnv-AgentEnv/resolve/main/thumbnail.jpg
 tags:
@@ -21,6 +21,7 @@ tags:
 - agents
 - weather
 - forecasting
+- multi-agent
 datasets:
 - earakely-scale/PortSimEnv-AgentEnv
 ---
@@ -30,10 +31,17 @@ datasets:
 An agent runs one container quay at the Port of Barcelona for a week of 2024 that has just gone wrong, and decides
 when, where and with how many cranes every ship docks. This Space replays the recorded runs on the
 [AgentEnv Framework](https://www.agentenvframework.com), on a 3D twin of the quay: twelve models on every wind, marine
-and live week, once each, and GPT-6.1 Sol and Claude Sonnet 5.5 on ten v1 weeks. The twelve are Claude Opus, Sonnet
+and live week, once each, seven of them so far on every lines week, and GPT-6.1 Sol and Claude Sonnet 5.5 on ten v1
+weeks. The twelve are Claude Opus, Sonnet
 and Haiku 5.5; GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna; Kimi K3, GLM-5.3, GLM-5.3-Flash, Qwen3.8-2.4T and DeepSeek
 V4.1 Flash on Fireworks; and Qwen3.8-27B on Groq.
 
+- **v5, the lines port:** the wind week with the shipping lines answering back. Each ship's agent, the firm the
+  port's 2024 record names for its call, is its own A2A agent, which the planner messages: the agents report ETAs that
+  firm up as their ships near, and a ship agrees a requested arrival time when it can make it by slowing down. Ships
+  berth first come, first served, a ship slowed on request keeping its turn, and an hour at anchor costs what an hour
+  late does. As in v4, the week is graded against the lower of the hindsight optimum, which knows every arrival, and
+  the best cost of a re-planner that follows the agents' ETAs and asks for arrival times.
 - **v4, the wind port:** the marine week in real Barcelona wind. Each week is moved into a real weather week of 2023
   to 2025, and the wind that blew at Meteocat's XEMA station Y7 (Bocana Sud), standing in for the Dique Sur
   anemometer of the port's traffic ordinance, sets its no-movement windows: above 25 kn ships of 300 m or more may not
@@ -62,7 +70,7 @@ weeks, and a week to replay it. The same board is the dataset's `results` table.
 
 The dataset [earakely-scale/PortSimEnv-AgentEnv](https://huggingface.co/datasets/earakely-scale/PortSimEnv-AgentEnv)
 holds the weeks, the references, the runs and their transcripts in two kinds of tables: PortSim's own, one family per
-version (`v4_tasks`, `v4_episodes`, `v4_references`, …), and agentenv-hf's standard tables, one pair per bundle, beside
+version (`v5_tasks`, `v5_episodes`, `v5_references`, …), and agentenv-hf's standard tables, one pair per bundle, beside
 the bundles as agent-env tasks. The environment is the plugin
 [earakely-scale/agentenv-portsim-plugin](https://github.com/earakely-scale/agentenv-portsim-plugin), built on
 [PortSimEnv](https://github.com/adithya-s-k/FineEnvs/tree/main/07-simulation-environments/portsim-v1), Adithya S
@@ -71,7 +79,7 @@ Kolavi's environment in FineEnvs; this viewer is PortSimEnv's.
 | | PortSimEnv in FineEnvs | PortSimEnv on AgentEnv |
 |---|---|---|
 | Environment and code | [FineEnvs/PortSimEnv](https://huggingface.co/spaces/FineEnvs/PortSimEnv) (OpenEnv Space), [code](https://github.com/adithya-s-k/FineEnvs/tree/main/07-simulation-environments/portsim-v1) | [earakely-scale/agentenv-portsim-plugin](https://github.com/earakely-scale/agentenv-portsim-plugin) |
-| Eval | [FineEnvs/PortSimEnv-Eval](https://huggingface.co/spaces/FineEnvs/PortSimEnv-Eval) | [the `results` table](https://huggingface.co/datasets/earakely-scale/PortSimEnv-AgentEnv#results), [model board](https://github.com/earakely-scale/agentenv-portsim-plugin/blob/v0.4.5/results/board.md) |
+| Eval | [FineEnvs/PortSimEnv-Eval](https://huggingface.co/spaces/FineEnvs/PortSimEnv-Eval) | [the `results` table](https://huggingface.co/datasets/earakely-scale/PortSimEnv-AgentEnv#results), [model board](https://github.com/earakely-scale/agentenv-portsim-plugin/blob/v0.5.0/results/board.md) |
 | Dataset | [FineEnvs/PortSimEnv](https://huggingface.co/datasets/FineEnvs/PortSimEnv) | [earakely-scale/PortSimEnv-AgentEnv](https://huggingface.co/datasets/earakely-scale/PortSimEnv-AgentEnv) |
 | Replays | | this Space |
 | Article | [Simulation RL Environments, part 1](https://huggingface.co/spaces/FineEnvs/simulation-rl-environments) | |
@@ -86,62 +94,64 @@ the dataset:
 
 ```bash
 uv tool install agentenv-framework \
-    --with "agentenv-portsim @ git+https://github.com/earakely-scale/agentenv-portsim-plugin@v0.4.5"
+    --with "agentenv-portsim @ git+https://github.com/earakely-scale/agentenv-portsim-plugin@v0.5.0"
 agent-env portsim setup --agent        # build the env and agent images and register them (a few minutes)
 agent-env run portsim --task smoke     # no model: grades a known optimal plan, prints "passed"
 
 export HF_TOKEN=hf_...                  # with "Make calls to Inference Providers", on an account with credits
 export LITELLM_BASE_URL=https://router.huggingface.co/v1 LITELLM_API_KEY=$HF_TOKEN
-agent-env hf run earakely-scale/PortSimEnv-AgentEnv@v0.4.5 --task dock-36A-w06x1-standard-0-e07 \
-    --model zai-org/GLM-5.3-Flash:baseten          # the default bundle: dock-v1-eval-wind (v4)
+agent-env hf run earakely-scale/PortSimEnv-AgentEnv@v0.5.0 --task dock-36A-w06x1-standard-0-e07 \
+    --model zai-org/GLM-5.3-Flash:baseten          # the default bundle: dock-v1-eval-lines (v5)
 ```
 
 `agent-env hf run` downloads the card and the bundles at that tag, checks that the plugins the card names are
 installed, shows what it will run and asks first (`--yes` skips the question). `dock-36A-w06x1-standard-0-e07` is the
-shortest wind week, 4 watches, and each task stops an episode before it could spend $5. To replay your own runs on the
-same viewer, locally, play them as a sweep:
+shortest lines and wind week, 4 watches, and each task stops an episode before it could spend $5. To replay your own
+runs on the same viewer, locally, play them as a sweep:
 
 ```bash
-agent-env portsim sweep run --wind --name mine --models zai-org/GLM-5.3-Flash:baseten \
+agent-env portsim sweep run --lines --name mine --models zai-org/GLM-5.3-Flash:baseten \
     --tasks dock-36A-w06x1-standard-0-e07 --episode-cap-usd 1 --cap-usd 3
 agent-env portsim view mine            # at http://127.0.0.1:8237/viewer/
 ```
 
 The [dataset card](https://huggingface.co/datasets/earakely-scale/PortSimEnv-AgentEnv#play-a-week-yourself) has more:
-the v3, v2 and v1 bundles and the other open models. If a step fails, see the plugin README's
-[troubleshooting](https://github.com/earakely-scale/agentenv-portsim-plugin/blob/v0.4.5/README.md#run-it-yourself).
+the v4, v3, v2 and v1 bundles and the other open models. If a step fails, see the plugin README's
+[troubleshooting](https://github.com/earakely-scale/agentenv-portsim-plugin/blob/v0.5.0/README.md#run-it-yourself).
 
 ## How this Space runs
 
-`start.sh` downloads the recorded sweeps (`runs/`) from the dataset at its `v0.4.5` tag (`--revision v0.4.5`), and
-`agent-env portsim view`, installed from the plugin's `v0.4.5` tag, serves them, replaying each v2, v3 and v4 run
+`start.sh` downloads the recorded sweeps (`runs/`) from the dataset at its `v0.5.0` tag (`--revision v0.5.0`), and
+`agent-env portsim view`, installed from the plugin's `v0.5.0` tag, serves them, replaying each v2 to v5 run
 through its week from the recorded tool calls. The 3D twin is downloaded from PortSimEnv's public bucket at start.
 
 ## Licence and attribution
 
-- **Tasks and transcripts:** CC BY-SA 4.0, except v4's weather (below). They carry PortSimEnv's dock-v1 task text,
-  which Adithya S Kolavi built from the Port of Barcelona's 2024 container calls; the other traffic in v3 and v4 is
-  derived from `Barcelona_2024.csv` in
+- **Tasks and transcripts:** CC BY-SA 4.0, except the weather of v4 and v5 (below). They carry PortSimEnv's dock-v1
+  task text, which Adithya S Kolavi built from the Port of Barcelona's 2024 container calls; the other traffic in v3 to
+  v5, and v5's ships' agents, are derived from `Barcelona_2024.csv` in
   [alberto-santini/berth-allocation-problems](https://github.com/alberto-santini/berth-allocation-problems) (CC BY-SA
   4.0). Contains data from the [Port de Barcelona open data portal](https://opendata.portdebarcelona.cat/).
-- **v4's forecasts**, in its tasks, transcripts and replays, are modified ECMWF open data. Copyright: © 2023-2026
-  European Centre for Medium-Range Weather Forecasts (ECMWF). Source: [www.ecmwf.int](https://www.ecmwf.int). Licence
-  Statement: This data is published under a Creative Commons Attribution 4.0 International
+- **v5's distances between ports** come from NGA Pub. 151, *Distances Between Ports* (11th edition, 2001), a work of
+  the United States government on which no copyright is claimed.
+- **The forecasts of v4 and v5**, in their tasks, transcripts and replays, are modified ECMWF open data. Copyright:
+  © 2023-2026 European Centre for Medium-Range Weather Forecasts (ECMWF). Source: [www.ecmwf.int](https://www.ecmwf.int).
+  Licence Statement: This data is published under a Creative Commons Attribution 4.0 International
   ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)). Disclaimer: ECMWF does not accept any liability
   whatsoever for any error or omission in the data, their availability, or for any loss or damage arising from their
   use. This data is based on data and products of the European Centre for Medium-Range Weather Forecasts (ECMWF).
   Modified: each run's 10 m wind at one grid point, every 3 hours to 72 hours, mapped to the Y7 anemometer by
   quantiles and rounded to whole knots. [Terms of use](https://apps.ecmwf.int/datasets/licences/general/).
-- **v4's observed wind windows**, in its tasks and replays, are derived from the readings of the Servei Meteorològic de
-  Catalunya's (Meteocat) XEMA station Y7, Port de Barcelona – Bocana Sud, in the Generalitat de Catalunya's
-  [open data portal](https://analisi.transparenciacatalunya.cat/d/nzvn-apee) (dataset nzvn-apee); no reading is
-  included, and they keep the terms published there
+- **The observed wind windows of v4 and v5**, in their tasks and replays, are derived from the readings of the Servei
+  Meteorològic de Catalunya's (Meteocat) XEMA station Y7, Port de Barcelona – Bocana Sud, in the Generalitat de
+  Catalunya's [open data portal](https://analisi.transparenciacatalunya.cat/d/nzvn-apee) (dataset nzvn-apee); no
+  reading is included, and they keep the terms published there
   ([Meteocat's legal notice](https://www.meteo.cat/wpweb/avis-legal/)).
   Font: Servei Meteorològic de Catalunya (Meteocat), estació XEMA Y7; dades extretes el 2026-10-08; finestres derivades.
 - **The 3D twin:** © OpenStreetMap contributors (ODbL) · terrain: Terrain Tiles (AWS).
 - **Code:** the plugin and PortSimEnv's viewer are under the Apache License 2.0.
 
-The plugin's [NOTICE](https://github.com/earakely-scale/agentenv-portsim-plugin/blob/v0.4.5/NOTICE) and
-[data/LICENSE](https://github.com/earakely-scale/agentenv-portsim-plugin/blob/v0.4.5/data/LICENSE) set out each part.
-This Space is independent: neither the Port de Barcelona, PortSimEnv's author, ECMWF nor Meteocat is affiliated with
-it or endorses it.
+The plugin's [NOTICE](https://github.com/earakely-scale/agentenv-portsim-plugin/blob/v0.5.0/NOTICE) and
+[data/LICENSE](https://github.com/earakely-scale/agentenv-portsim-plugin/blob/v0.5.0/data/LICENSE) set out each part.
+This Space is independent: neither the Port de Barcelona, PortSimEnv's author, ECMWF, Meteocat, the NGA nor any
+shipping line or agent named in it is affiliated with it or endorses it.

start.sh

@@ -1,4 +1,4 @@
 set -e
-[ -d results/runs ] || hf download earakely-scale/PortSimEnv-AgentEnv --repo-type dataset --revision v0.4.5 --include "runs/*" \
+[ -d results/runs ] || hf download earakely-scale/PortSimEnv-AgentEnv --repo-type dataset --revision v0.5.0 --include "runs/*" \
     --local-dir results --quiet
 exec agent-env portsim view --host 0.0.0.0 --port 7860