leagents 0.0.4__tar.gz

leagents-0.0.4/LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

leagents-0.0.4/PKG-INFO

@@ -0,0 +1,106 @@
+Metadata-Version: 2.4
+Name: leagents
+Version: 0.0.4
+Summary: LeAgents — agentic orchestration for the LeRobot pipeline: collect -> train -> eval -> improve
+Author: ratel
+License: Apache-2.0
+Project-URL: Repository, https://github.com/ratelcode/LeAgents
+Project-URL: Documentation, https://github.com/ratelcode/LeAgents/blob/main/DESIGN.md
+Project-URL: Changelog, https://github.com/ratelcode/LeAgents/releases
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: pydantic>=2.5
+Requires-Dist: PyYAML>=6.0
+Provides-Extra: lerobot
+Requires-Dist: lerobot[libero,smolvla]>=0.5; extra == "lerobot"
+Provides-Extra: dash
+Requires-Dist: fastapi>=0.110; extra == "dash"
+Requires-Dist: uvicorn>=0.29; extra == "dash"
+Provides-Extra: dev
+Requires-Dist: pytest>=8; extra == "dev"
+Requires-Dist: ruff>=0.4; extra == "dev"
+Requires-Dist: fastapi>=0.110; extra == "dev"
+Requires-Dist: uvicorn>=0.29; extra == "dev"
+Requires-Dist: httpx>=0.27; extra == "dev"
+Dynamic: license-file
+
+# LeAgents
+
+Agentic orchestration for the [LeRobot](https://github.com/huggingface/lerobot) robotics pipeline — an orchestrator drives an automated **collect → train → eval → improve** loop over [LeRobotDataset v3.0](https://huggingface.co/docs/lerobot/lerobot-dataset-v3), with a deterministic loop controller, a constitution safety gate, verification gates before promotion, and (M2) a dashboard for visualizing the flow.
+
+<p align="center"><img src="https://raw.githubusercontent.com/ratelcode/LeAgents/main/docs/architecture.svg" alt="LeAgents architecture — a deterministic loop controller dispatches data/train/eval/knowledge agents over LeRobotDataset v3.0 on the Hugging Face Hub, with a constitution safety gate, an OKF knowledge wiki feeding the proposer, and a flow dashboard reading the event log" width="880"></p>
+
+Architecture, research grounding (verified 2023–2026 papers), and roadmap: **[DESIGN.md](DESIGN.md)**.
+
+## Status — v0.0.4
+
+| Milestone | Scope | Status |
+|---|---|---|
+| **M0** | Sim-only loop on LIBERO: seed dataset → SmolVLA fine-tune → `lerobot-eval` gate → promote/iterate/escalate/rollback | ✅ pipeline verified end-to-end on a real GPU (PushT + LIBERO smoke configs); full-scale training run pending |
+| **M1** | DexFlyWheel-style self-improvement, RoboGene-style task curation, policy escalation, OKF knowledge layer (Karpathy-wiki-style, DESIGN.md §3.6) + provider-agnostic LLM proposer | 🚧 knowledge layer + LLM adapter landed |
+| **M2** | Flow dashboard (Rerun episode replay, WandB curves, OTel agent traces) | 🚧 flow view v1 landed: runs → cycles → decisions live, eval chart, event log, knowledge browser (`leagents dash`) |
+| M3 | Real robot: teleop collection, HIL-SERL adapter (requires lerobot ≥ 0.6.0, see CVE note in DESIGN.md §6) | planned |
+
+What works today: the full loop state machine with budgets, the constitution gate, SQLite job store, JSONL event log, subprocess wrappers for `lerobot-train` / `lerobot-eval`, the OKF knowledge layer (`knowledge/` pages with provenance, updated every cycle, linted), the DexFlyWheel data path (success-filtered rollout harvesting → accumulated mix → adaptation training), and a provider-agnostic LLM adapter (`llm: anthropic:*|openai:*[@base_url]`, or none at all — every flow has a deterministic fallback). All covered by tests that run without a GPU or lerobot installed.
+
+## First full-scale autonomous run (2026-07-04)
+
+One M0 run on a single RTX 5070 Ti (16 GB), fully autonomous — 3 cycles, 6.1 GPU-hours, zero human intervention, every decision/event/knowledge-page update logged:
+
+| Cycle | Data | Train | LIBERO spatial eval (100 episodes) | Decision |
+|---|---|---|---|---|
+| 0 | 40 episodes | 20k steps from `smolvla_base` (loss → 0.030) | 0% success — arm reaches targets, never completes | **promote** (baseline) |
+| 1 | 80 | 20k steps, continued from the blessed checkpoint | 0% | **iterate** |
+| 2 | 160 | 20k steps | 0% | **iterate** — the `escalate_floor` guard correctly refused to escalate a 0%-plateau to a bigger policy |
+
+The 0% success is the expected outcome of the data budget, not a pipeline failure: 4→16 episodes per task versus the ~50-per-variation guidance the design research verified for SmolVLA. This run validates the *loop* — budgets held, weights carried over, and the decision function behaved exactly as specified. The next run scales the data schedule to 200→500 episodes (20→50 per task).
+
+## Quickstart
+
+```bash
+pip install -e ".[dev]"
+pytest
+
+# dry run — no GPU, no lerobot; synthetic eval scores exercise the decision logic
+leagents run -c configs/m0_libero.yaml --dry-run
+
+# real run (Linux; needs a GPU and the LIBERO extras)
+pip install -e ".[lerobot]"
+leagents run -c configs/m0_libero.yaml
+
+# inspect runs, cycles, decisions, blessed checkpoints
+leagents status
+
+# flow dashboard — runs, cycle pipeline, decisions, eval chart, events, knowledge
+pip install -e ".[dash]"
+leagents dash            # → http://127.0.0.1:8321
+```
+
+## How the loop decides
+
+Each cycle trains a candidate checkpoint and evaluates it on LIBERO. The decision is a pure function of success-rate deltas vs. the blessed baseline (`leagents/orchestrator/decision.py`):
+
+- **promote** — candidate beats baseline by ≥ `promote_delta`; it becomes the new blessed checkpoint
+- **iterate** — small improvement; collect more data on failing task variations
+- **escalate** — plateaued for `plateau_cycles`; move up the policy ladder (SmolVLA → π0.5)
+- **rollback** — regression; the blessed checkpoint stands
+
+Control flow is deterministic Python persisted to SQLite — LLM agents (task proposal, curation; M1) only make proposals *inside* the gates, never control-flow decisions.
+
+## Layout
+
+```
+leagents/
+├── orchestrator/   # loop controller, decision logic, constitution gate, proposer
+├── agents/         # data / train / eval / improve agents (LeRobot CLI wrappers)
+├── contracts/      # typed records: DatasetRef, CheckpointRecord, EvalReport
+├── events/         # JSONL event bus (dashboard reads this in M2)
+└── store/          # SQLite job store (runs, cycles, checkpoints)
+configs/            # m0_libero.yaml, constitution.yaml
+tests/              # loop e2e with fake runners — no GPU needed
+```
+
+## License
+
+Apache-2.0

leagents-0.0.4/README.md

@@ -0,0 +1,79 @@
+# LeAgents
+
+Agentic orchestration for the [LeRobot](https://github.com/huggingface/lerobot) robotics pipeline — an orchestrator drives an automated **collect → train → eval → improve** loop over [LeRobotDataset v3.0](https://huggingface.co/docs/lerobot/lerobot-dataset-v3), with a deterministic loop controller, a constitution safety gate, verification gates before promotion, and (M2) a dashboard for visualizing the flow.
+
+<p align="center"><img src="https://raw.githubusercontent.com/ratelcode/LeAgents/main/docs/architecture.svg" alt="LeAgents architecture — a deterministic loop controller dispatches data/train/eval/knowledge agents over LeRobotDataset v3.0 on the Hugging Face Hub, with a constitution safety gate, an OKF knowledge wiki feeding the proposer, and a flow dashboard reading the event log" width="880"></p>
+
+Architecture, research grounding (verified 2023–2026 papers), and roadmap: **[DESIGN.md](DESIGN.md)**.
+
+## Status — v0.0.4
+
+| Milestone | Scope | Status |
+|---|---|---|
+| **M0** | Sim-only loop on LIBERO: seed dataset → SmolVLA fine-tune → `lerobot-eval` gate → promote/iterate/escalate/rollback | ✅ pipeline verified end-to-end on a real GPU (PushT + LIBERO smoke configs); full-scale training run pending |
+| **M1** | DexFlyWheel-style self-improvement, RoboGene-style task curation, policy escalation, OKF knowledge layer (Karpathy-wiki-style, DESIGN.md §3.6) + provider-agnostic LLM proposer | 🚧 knowledge layer + LLM adapter landed |
+| **M2** | Flow dashboard (Rerun episode replay, WandB curves, OTel agent traces) | 🚧 flow view v1 landed: runs → cycles → decisions live, eval chart, event log, knowledge browser (`leagents dash`) |
+| M3 | Real robot: teleop collection, HIL-SERL adapter (requires lerobot ≥ 0.6.0, see CVE note in DESIGN.md §6) | planned |
+
+What works today: the full loop state machine with budgets, the constitution gate, SQLite job store, JSONL event log, subprocess wrappers for `lerobot-train` / `lerobot-eval`, the OKF knowledge layer (`knowledge/` pages with provenance, updated every cycle, linted), the DexFlyWheel data path (success-filtered rollout harvesting → accumulated mix → adaptation training), and a provider-agnostic LLM adapter (`llm: anthropic:*|openai:*[@base_url]`, or none at all — every flow has a deterministic fallback). All covered by tests that run without a GPU or lerobot installed.
+
+## First full-scale autonomous run (2026-07-04)
+
+One M0 run on a single RTX 5070 Ti (16 GB), fully autonomous — 3 cycles, 6.1 GPU-hours, zero human intervention, every decision/event/knowledge-page update logged:
+
+| Cycle | Data | Train | LIBERO spatial eval (100 episodes) | Decision |
+|---|---|---|---|---|
+| 0 | 40 episodes | 20k steps from `smolvla_base` (loss → 0.030) | 0% success — arm reaches targets, never completes | **promote** (baseline) |
+| 1 | 80 | 20k steps, continued from the blessed checkpoint | 0% | **iterate** |
+| 2 | 160 | 20k steps | 0% | **iterate** — the `escalate_floor` guard correctly refused to escalate a 0%-plateau to a bigger policy |
+
+The 0% success is the expected outcome of the data budget, not a pipeline failure: 4→16 episodes per task versus the ~50-per-variation guidance the design research verified for SmolVLA. This run validates the *loop* — budgets held, weights carried over, and the decision function behaved exactly as specified. The next run scales the data schedule to 200→500 episodes (20→50 per task).
+
+## Quickstart
+
+```bash
+pip install -e ".[dev]"
+pytest
+
+# dry run — no GPU, no lerobot; synthetic eval scores exercise the decision logic
+leagents run -c configs/m0_libero.yaml --dry-run
+
+# real run (Linux; needs a GPU and the LIBERO extras)
+pip install -e ".[lerobot]"
+leagents run -c configs/m0_libero.yaml
+
+# inspect runs, cycles, decisions, blessed checkpoints
+leagents status
+
+# flow dashboard — runs, cycle pipeline, decisions, eval chart, events, knowledge
+pip install -e ".[dash]"
+leagents dash            # → http://127.0.0.1:8321
+```
+
+## How the loop decides
+
+Each cycle trains a candidate checkpoint and evaluates it on LIBERO. The decision is a pure function of success-rate deltas vs. the blessed baseline (`leagents/orchestrator/decision.py`):
+
+- **promote** — candidate beats baseline by ≥ `promote_delta`; it becomes the new blessed checkpoint
+- **iterate** — small improvement; collect more data on failing task variations
+- **escalate** — plateaued for `plateau_cycles`; move up the policy ladder (SmolVLA → π0.5)
+- **rollback** — regression; the blessed checkpoint stands
+
+Control flow is deterministic Python persisted to SQLite — LLM agents (task proposal, curation; M1) only make proposals *inside* the gates, never control-flow decisions.
+
+## Layout
+
+```
+leagents/
+├── orchestrator/   # loop controller, decision logic, constitution gate, proposer
+├── agents/         # data / train / eval / improve agents (LeRobot CLI wrappers)
+├── contracts/      # typed records: DatasetRef, CheckpointRecord, EvalReport
+├── events/         # JSONL event bus (dashboard reads this in M2)
+└── store/          # SQLite job store (runs, cycles, checkpoints)
+configs/            # m0_libero.yaml, constitution.yaml
+tests/              # loop e2e with fake runners — no GPU needed
+```
+
+## License
+
+Apache-2.0

leagents-0.0.4/leagents/__init__.py

@@ -0,0 +1,3 @@
+"""leagents — agentic orchestration for the LeRobot pipeline."""
+
+__version__ = "0.0.4"

leagents-0.0.4/leagents/agents/__init__.py

@@ -0,0 +1,21 @@
+from leagents.agents.base import RunResult, Runner, dry_runner, subprocess_runner
+from leagents.agents.data_agent import DataAgent
+from leagents.agents.eval_agent import EvalAgent, EvalError
+from leagents.agents.improve_agent import ImproveAgent, ImproveError
+from leagents.agents.knowledge_agent import KnowledgeAgent
+from leagents.agents.train_agent import TrainAgent, TrainError
+
+__all__ = [
+    "RunResult",
+    "Runner",
+    "dry_runner",
+    "subprocess_runner",
+    "DataAgent",
+    "EvalAgent",
+    "EvalError",
+    "ImproveAgent",
+    "ImproveError",
+    "KnowledgeAgent",
+    "TrainAgent",
+    "TrainError",
+]

leagents-0.0.4/leagents/agents/base.py

@@ -0,0 +1,42 @@
+"""Shared subprocess runner for agents that wrap LeRobot CLIs.
+
+Runners are injectable so tests (and --dry-run) exercise the full loop
+without lerobot installed or a GPU present.
+"""
+
+from __future__ import annotations
+
+import subprocess
+import time
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Callable, Sequence
+
+
+@dataclass
+class RunResult:
+    cmd: list[str]
+    exit_code: int
+    duration_s: float
+    log_path: Path | None = None
+
+
+Runner = Callable[[Sequence[str], Path], RunResult]
+
+
+def subprocess_runner(cmd: Sequence[str], log_path: Path) -> RunResult:
+    """Run a command, teeing stdout+stderr to a log file.
+
+    Training jobs stream output for hours — never buffer it in memory.
+    """
+    log_path.parent.mkdir(parents=True, exist_ok=True)
+    start = time.monotonic()
+    with log_path.open("w") as log:
+        proc = subprocess.run(list(cmd), stdout=log, stderr=subprocess.STDOUT)
+    return RunResult(list(cmd), proc.returncode, time.monotonic() - start, log_path)
+
+
+def dry_runner(cmd: Sequence[str], log_path: Path) -> RunResult:
+    log_path.parent.mkdir(parents=True, exist_ok=True)
+    log_path.write_text("DRY RUN: " + " ".join(cmd) + "\n")
+    return RunResult(list(cmd), 0, 0.0, log_path)

leagents-0.0.4/leagents/agents/data_agent.py

@@ -0,0 +1,38 @@
+"""Data Agent — M0 scope (DESIGN.md §3.2, §8).
+
+M0 collects no new data: the agent resolves the seed dataset from the
+proposal and records provenance for the cycle. RoboGene-style curation,
+MimicGen-style amplification, and GenAug-style augmentation land in M1.
+"""
+
+from __future__ import annotations
+
+from leagents.contracts import DatasetRef, Proposal
+from leagents.events import Event, EventBus
+
+
+class DataAgent:
+    def __init__(self, bus: EventBus):
+        self.bus = bus
+
+    def run(self, *, run_id: str, cycle: int, proposal: Proposal) -> DatasetRef:
+        ref = DatasetRef(
+            repo_id=proposal.dataset,
+            num_episodes=proposal.num_episodes,
+            notes=proposal.notes,
+        )
+        self.bus.emit(
+            Event(
+                run_id=run_id,
+                stage="collect",
+                kind="dataset_resolved",
+                cycle=cycle,
+                payload={
+                    "action": proposal.action,
+                    "repo_id": ref.repo_id,
+                    "num_episodes": ref.num_episodes,
+                    "notes": ref.notes,
+                },
+            )
+        )
+        return ref

leagents-0.0.4/leagents/agents/eval_agent.py

@@ -0,0 +1,105 @@
+"""Eval Agent — `lerobot-eval` LIBERO gate (DESIGN.md §3.4)."""
+
+from __future__ import annotations
+
+import json
+from pathlib import Path
+from typing import Any
+
+from leagents.agents.base import Runner, subprocess_runner
+from leagents.config import EvalConfig
+from leagents.contracts import CheckpointRecord, EvalReport
+from leagents.events import Event, EventBus
+from leagents.orchestrator.constitution import Constitution, ConstitutionError
+
+
+class EvalError(Exception):
+    pass
+
+
+class EvalAgent:
+    def __init__(
+        self,
+        cfg: EvalConfig,
+        constitution: Constitution,
+        bus: EventBus,
+        runner: Runner = subprocess_runner,
+    ):
+        self.cfg = cfg
+        self.constitution = constitution
+        self.bus = bus
+        self.runner = runner
+
+    def build_command(self, checkpoint: CheckpointRecord, output_dir: Path) -> list[str]:
+        return [
+            "lerobot-eval",
+            f"--policy.path={checkpoint.path}",
+            f"--env.type={self.cfg.env_type}",
+            f"--env.task={self.cfg.task_suite}",
+            f"--eval.n_episodes={self.cfg.n_episodes}",
+            # lerobot-eval rejects batch_size > n_episodes (default batch is 50)
+            f"--eval.batch_size={min(self.cfg.batch_size, self.cfg.n_episodes)}",
+            f"--output_dir={output_dir}",
+            *self.cfg.extra_args,
+        ]
+
+    def run(
+        self, *, run_id: str, cycle: int, checkpoint: CheckpointRecord, workdir: Path
+    ) -> EvalReport:
+        output_dir = workdir / f"cycle_{cycle}" / "eval"
+        cmd = self.build_command(checkpoint, output_dir)
+
+        for verdict in (self.constitution.check_eval(self.cfg.env_type, self.cfg.n_episodes),
+                        self.constitution.check_command(cmd)):
+            if not verdict.allowed:
+                self.bus.emit(Event(run_id, "eval", "constitution_denied", cycle,
+                                    {"rule": verdict.rule, "reason": verdict.reason}))
+                raise ConstitutionError(verdict)
+
+        self.bus.emit(Event(run_id, "eval", "job_started", cycle, {"cmd": cmd}))
+        # log lives outside output_dir: lerobot CLIs refuse a pre-existing output_dir
+        result = self.runner(cmd, output_dir.parent / "eval.log")
+        self.bus.emit(Event(run_id, "eval", "job_finished", cycle,
+                            {"exit_code": result.exit_code, "duration_s": result.duration_s,
+                             "log": str(result.log_path)}))
+        if result.exit_code != 0:
+            raise EvalError(f"lerobot-eval exited {result.exit_code}, see {result.log_path}")
+
+        success_rate, per_task, raw = self._parse_eval_info(output_dir / "eval_info.json")
+        report = EvalReport(
+            checkpoint=checkpoint.path,
+            env_type=self.cfg.env_type,
+            task_suite=self.cfg.task_suite,
+            n_episodes=self.cfg.n_episodes,
+            success_rate=success_rate,
+            per_task=per_task,
+            raw=raw,
+        )
+        self.bus.emit(Event(run_id, "eval", "report", cycle,
+                            {"success_rate": success_rate, "task_suite": self.cfg.task_suite}))
+        return report
+
+    @staticmethod
+    def _parse_eval_info(path: Path) -> tuple[float, dict[str, float], dict[str, Any]]:
+        """Parse lerobot-eval's eval_info.json.
+
+        lerobot 0.5 schema: {"overall": {"pc_success": <0-100>, ...},
+        "per_group": {<group>: {"pc_success": ...}}, "per_task": [...]}.
+        Returns (success fraction, per-group success fractions, raw data).
+        """
+        if not path.exists():
+            raise EvalError(f"eval output missing: {path}")
+        data: dict[str, Any] = json.loads(path.read_text())
+        aggregate = data.get("overall") or data.get("aggregated") or data
+        if "pc_success" in aggregate:  # percentage 0-100 by definition
+            success = float(aggregate["pc_success"]) / 100.0
+        elif "success_rate" in aggregate:  # fraction 0-1
+            success = float(aggregate["success_rate"])
+        else:
+            raise EvalError(f"no success metric found in {path}")
+        per_task = {
+            group: float(metrics["pc_success"]) / 100.0
+            for group, metrics in (data.get("per_group") or {}).items()
+            if isinstance(metrics, dict) and "pc_success" in metrics
+        }
+        return success, per_task, data