envloop-cli 1.23.0__tar.gz → 1.27.0__tar.gz

{envloop_cli-1.23.0 → envloop_cli-1.27.0}/.gitignore

@@ -10,3 +10,7 @@ dist/
 # Local runtime output from offline platform API contract fixtures.
 .envplatform/
 .superpowers/
+
+# Isolated build and integration-test fixtures
+.ci-deps/
+.local-runs/

{envloop_cli-1.23.0 → envloop_cli-1.27.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: envloop-cli
-Version: 1.23.0
+Version: 1.27.0
 Summary: EnvLoop CLI and Python SDK for jobs, training, checkpoints and inference
 Project-URL: Homepage, https://envloop.ai
 Project-URL: Documentation, https://docs.envloop.ai
@@ -8,14 +8,11 @@ Project-URL: Repository, https://github.com/EnvLoop/envloop-cli
 Project-URL: Issues, https://github.com/EnvLoop/envloop-cli/issues
 Project-URL: Agent instructions, https://docs.envloop.ai/skill.md
 Requires-Python: >=3.12
-Requires-Dist: envloop-cli-experiment==1.23.0
 Provides-Extra: data
-Requires-Dist: envloop-cli-data==1.23.0; extra == 'data'
+Requires-Dist: envloop-envhub==1.23.0; extra == 'data'
 Provides-Extra: default
 Provides-Extra: experiment
-Requires-Dist: envloop-cli-experiment==1.23.0; extra == 'experiment'
 Provides-Extra: rsi-loop
-Requires-Dist: envloop-cli-rsi-loop==1.23.0; extra == 'rsi-loop'
 Description-Content-Type: text/markdown
 
 # envloop-cli: EnvLoop CLI and Python SDK
@@ -55,7 +52,7 @@ uv pip install --upgrade envloop-cli
 python3.12 -m pip install --upgrade envloop-cli
 ```
 
-In this source candidate the bare installation and `default` include experiment: Jobs, Trials, training, logs and inference calls. Data resources require `data`; RSI Loop controls require `rsi-loop`. This source change is not a claim that the modular packages have been published.
+Every installation includes Jobs, Trials, training, logs, inference calls, Data resources, checkpoints and RSI Loop controls.
 Check that the SDK can be imported:
 
 ```bash
@@ -68,18 +65,26 @@ follow the [managed training guide](https://docs.envloop.ai/training/cli-hello-w
 
 Install the client in your Python environment; the platform provides the execution resources.
 
-## Optional features (source candidate)
+## Optional features
 
-The current source candidate uses independent feature wheels at version 1.23.0. Matching modular distributions and public ANS core must be released before using this contract from a public index:
+One `envloop-cli` wheel and source archive contain all feature code. Extras add dependencies only:
 
 ```bash
-python -m pip install "envloop-cli[data]"        # Dataset/Data/Runner/Checkpoint/Profile/Recipe
-python -m pip install envloop-cli               # Includes Job/Trial/logs/inference
-python -m pip install "envloop-cli[experiment]"  # Explicit spelling of the default installation
-python -m pip install "envloop-cli[rsi-loop]"    # Existing job run paths; includes experiment
+python -m pip install envloop-cli
+python -m pip install "envloop-cli[data]"
+python -m pip install "envloop-cli[data,experiment,rsi-loop]"
 ```
 
-`envloop-cli[default]` equals the bare installation: base plus `envloop-cli-experiment==1.23.0`, delivered as separate wheels. Only base owns the `el`/`envloop` scripts. Feature resources load on access; absent optional features report installation guidance. The default installation uses standard-library verified TLS and system CA configuration, without EnvHub or certifi. The examples below require their corresponding feature packages and an authorized matching platform.
+Every installation includes the existing Job, Trial, Data and RSI Loop commands. The bare installation and `default`, `experiment`, `rsi-loop` groups have no runtime dependencies. `data` adds `envloop-envhub==1.23.0` for ANS parsing and verified directory downloads. Other Data API operations use the bundled client implementation. Missing ANS dependencies produce the `envloop-cli[data]` installation instruction.
+
+Feature modules load lazily. Default imports, help and experiment operations do not require EnvHub or certifi. Only the single `envloop-cli` distribution owns the `el` / `envloop` scripts. The examples below require an authorized matching platform; installation does not grant permissions.
+
+Versions 1.23.0 and 1.24.0 on PyPI depend on feature packages that were never published: their `data` extra cannot be installed and the DATA command group is empty. They also install a separate `envloop-cli-experiment` package that shares files with this release. Remove both before installing; uninstalling `envloop-cli-experiment` after an upgrade deletes files the client needs and leads to `FEATURE_NOT_INSTALLED`, which the same commands repair:
+
+```bash
+python -m pip uninstall -y envloop-cli envloop-cli-experiment
+python -m pip install "envloop-cli[data]"
+```
 
 ## Connect
 
@@ -130,15 +135,14 @@ Error: PROFILE_NOT_FOUND
 ```bash
 el job list --limit 5
 el job list --type train --limit 5
-el trail list --limit 5
-el trail list --type train --job-id JOB_ID
+el trial list --limit 5
+el trial list --type train --job-id JOB_ID
 el checkpoint list --limit 5
 el job list --type inference
 ```
 
 Query a returned Job ID with `el job get JOB_ID` and `el job result JOB_ID`.
-Trial commands use the spelling `trail`, for example
-`el trail artifact list TRIAL_ID`. Use each command's `--help` for options.
+Trial commands use `el trial`, for example `el trial artifact list TRIAL_ID`. Use each command's `--help` for options.
 
 For the Python SDK:
 
@@ -237,7 +241,7 @@ el job result JOB_ID
 
 Replace `JOB_ID` with the returned value. The example has two independent Trials, each requesting four H100 GPUs: concurrent execution needs eight GPUs. Capacity and budget determine actual concurrency. Startup and execution are billable; check the Runner requirements, project `training.run` permission, and applicable prices and balance in your authorized Console account first. Reduce the Trial or GPU count when your program supports a smaller first run. The [usage guide](https://docs.envloop.ai/manage/usage) explains resource observations.
 
-The request file does not depend on GPU shortcut flags. If `el job submit --help` lists both `--trails` and `--gput-type`, you can instead use the shortcuts in [GPU Runner training](https://docs.envloop.ai/training/runner). Choose one submission method to avoid duplicate runs. The SDK accepts the same dictionary through `client.training_jobs.submit(request)`.
+The request file does not depend on GPU shortcut flags. If `el job submit --help` lists both `--trials` and `--gpu-type`, you can instead use the shortcuts in [GPU Runner training](https://docs.envloop.ai/training/runner). Choose one submission method to avoid duplicate runs. The SDK accepts the same dictionary through `client.training_jobs.submit(request)`.
 
 The platform starts the published Revision’s `main.sh` / `main.py` using the Runner image; it does not upload local code. An optional command can only supply a declared `run_command` parameter. If this mode or GPU type is unavailable, contact your project administrator. Use `el job stop JOB_ID` to stop every Trial, then inspect each Trial’s `cleanup_confirmed` and download the uploaded artifacts you need.
 
@@ -289,6 +293,8 @@ el data download DATA_ID --file FILE_ID --output ./dataset.tgz --resume --json
 
 Unified ANS CRUD supports Dataset, Runner, Recipe and HF model Checkpoint registrations through `el data` and `client.data`. Dataset/Runner/Recipe creation requires a native immutable `--ref`, `--manifest` and exact `--directory`. HF metadata registration uses `el data create --type checkpoint --model namespace/model --revision FULL_COMMIT`; SDK uses `create(kind="checkpoint", model=..., revision=...)`. HF weights are not downloaded, and artifact imports do not establish Model or Job checkpoint provenance.
 
+`el runner register DIRECTORY --revision FULL_COMMIT [--source-ref main]` publishes a Runner package directory through the same Data API with the manifest built for you, then reads the catalog revision back. `el profile create|update|fork ... --secret NAME[=prompt|stdin|env:VAR|file:PATH]` stores write-only Secret parameters whose names follow the platform's credential rule or `secret_params`; values never come from the command line. `el setting model-access grant NAMESPACE/NAME --protocol anthropic [--revision FULL_COMMIT]` (project Admin) lets a published Runner receive the project's metered model key on Fly Trials; `list` and `revoke GRANT_ID` manage grants.
+
 Success follows native storage and original catalog readback and returns an `ans_` identity. Update publishes an explicit new revision. Delete checks active and unknown dependencies; Runner deletion covers all revisions. Historical bytes and quota usage remain. Commands default to the current login base URL/workspace/project, with no `--env`. Both Data grants and the corresponding resource capabilities are required.
 
 Read methods include `list/get/files/open/download`; `client.recipes.list/get` queries the original Recipe catalog. Downloads verify size/SHA-256 and refuse existing destinations, symlinks and storage redirects. Large payloads use multipart with stable idempotency keys. Staging `pub_` IDs and failure phases support retry; they are not native success. Matching client/server support is required. See the [Data reference](https://docs.envloop.ai/reference/cli#data-commands).

{envloop_cli-1.23.0 → envloop_cli-1.27.0}/README.pypi.md

@@ -35,7 +35,7 @@ uv pip install --upgrade envloop-cli
 python3.12 -m pip install --upgrade envloop-cli
 ```
 
-In this source candidate the bare installation and `default` include experiment: Jobs, Trials, training, logs and inference calls. Data resources require `data`; RSI Loop controls require `rsi-loop`. This source change is not a claim that the modular packages have been published.
+Every installation includes Jobs, Trials, training, logs, inference calls, Data resources, checkpoints and RSI Loop controls.
 Check that the SDK can be imported:
 
 ```bash
@@ -48,18 +48,26 @@ follow the [managed training guide](https://docs.envloop.ai/training/cli-hello-w
 
 Install the client in your Python environment; the platform provides the execution resources.
 
-## Optional features (source candidate)
+## Optional features
 
-The current source candidate uses independent feature wheels at version 1.23.0. Matching modular distributions and public ANS core must be released before using this contract from a public index:
+One `envloop-cli` wheel and source archive contain all feature code. Extras add dependencies only:
 
 ```bash
-python -m pip install "envloop-cli[data]"        # Dataset/Data/Runner/Checkpoint/Profile/Recipe
-python -m pip install envloop-cli               # Includes Job/Trial/logs/inference
-python -m pip install "envloop-cli[experiment]"  # Explicit spelling of the default installation
-python -m pip install "envloop-cli[rsi-loop]"    # Existing job run paths; includes experiment
+python -m pip install envloop-cli
+python -m pip install "envloop-cli[data]"
+python -m pip install "envloop-cli[data,experiment,rsi-loop]"
 ```
 
-`envloop-cli[default]` equals the bare installation: base plus `envloop-cli-experiment==1.23.0`, delivered as separate wheels. Only base owns the `el`/`envloop` scripts. Feature resources load on access; absent optional features report installation guidance. The default installation uses standard-library verified TLS and system CA configuration, without EnvHub or certifi. The examples below require their corresponding feature packages and an authorized matching platform.
+Every installation includes the existing Job, Trial, Data and RSI Loop commands. The bare installation and `default`, `experiment`, `rsi-loop` groups have no runtime dependencies. `data` adds `envloop-envhub==1.23.0` for ANS parsing and verified directory downloads. Other Data API operations use the bundled client implementation. Missing ANS dependencies produce the `envloop-cli[data]` installation instruction.
+
+Feature modules load lazily. Default imports, help and experiment operations do not require EnvHub or certifi. Only the single `envloop-cli` distribution owns the `el` / `envloop` scripts. The examples below require an authorized matching platform; installation does not grant permissions.
+
+Versions 1.23.0 and 1.24.0 on PyPI depend on feature packages that were never published: their `data` extra cannot be installed and the DATA command group is empty. They also install a separate `envloop-cli-experiment` package that shares files with this release. Remove both before installing; uninstalling `envloop-cli-experiment` after an upgrade deletes files the client needs and leads to `FEATURE_NOT_INSTALLED`, which the same commands repair:
+
+```bash
+python -m pip uninstall -y envloop-cli envloop-cli-experiment
+python -m pip install "envloop-cli[data]"
+```
 
 ## Connect
 
@@ -110,15 +118,14 @@ Error: PROFILE_NOT_FOUND
 ```bash
 el job list --limit 5
 el job list --type train --limit 5
-el trail list --limit 5
-el trail list --type train --job-id JOB_ID
+el trial list --limit 5
+el trial list --type train --job-id JOB_ID
 el checkpoint list --limit 5
 el job list --type inference
 ```
 
 Query a returned Job ID with `el job get JOB_ID` and `el job result JOB_ID`.
-Trial commands use the spelling `trail`, for example
-`el trail artifact list TRIAL_ID`. Use each command's `--help` for options.
+Trial commands use `el trial`, for example `el trial artifact list TRIAL_ID`. Use each command's `--help` for options.
 
 For the Python SDK:
 
@@ -217,7 +224,7 @@ el job result JOB_ID
 
 Replace `JOB_ID` with the returned value. The example has two independent Trials, each requesting four H100 GPUs: concurrent execution needs eight GPUs. Capacity and budget determine actual concurrency. Startup and execution are billable; check the Runner requirements, project `training.run` permission, and applicable prices and balance in your authorized Console account first. Reduce the Trial or GPU count when your program supports a smaller first run. The [usage guide](https://docs.envloop.ai/manage/usage) explains resource observations.
 
-The request file does not depend on GPU shortcut flags. If `el job submit --help` lists both `--trails` and `--gput-type`, you can instead use the shortcuts in [GPU Runner training](https://docs.envloop.ai/training/runner). Choose one submission method to avoid duplicate runs. The SDK accepts the same dictionary through `client.training_jobs.submit(request)`.
+The request file does not depend on GPU shortcut flags. If `el job submit --help` lists both `--trials` and `--gpu-type`, you can instead use the shortcuts in [GPU Runner training](https://docs.envloop.ai/training/runner). Choose one submission method to avoid duplicate runs. The SDK accepts the same dictionary through `client.training_jobs.submit(request)`.
 
 The platform starts the published Revision’s `main.sh` / `main.py` using the Runner image; it does not upload local code. An optional command can only supply a declared `run_command` parameter. If this mode or GPU type is unavailable, contact your project administrator. Use `el job stop JOB_ID` to stop every Trial, then inspect each Trial’s `cleanup_confirmed` and download the uploaded artifacts you need.
 
@@ -269,6 +276,8 @@ el data download DATA_ID --file FILE_ID --output ./dataset.tgz --resume --json
 
 Unified ANS CRUD supports Dataset, Runner, Recipe and HF model Checkpoint registrations through `el data` and `client.data`. Dataset/Runner/Recipe creation requires a native immutable `--ref`, `--manifest` and exact `--directory`. HF metadata registration uses `el data create --type checkpoint --model namespace/model --revision FULL_COMMIT`; SDK uses `create(kind="checkpoint", model=..., revision=...)`. HF weights are not downloaded, and artifact imports do not establish Model or Job checkpoint provenance.
 
+`el runner register DIRECTORY --revision FULL_COMMIT [--source-ref main]` publishes a Runner package directory through the same Data API with the manifest built for you, then reads the catalog revision back. `el profile create|update|fork ... --secret NAME[=prompt|stdin|env:VAR|file:PATH]` stores write-only Secret parameters whose names follow the platform's credential rule or `secret_params`; values never come from the command line. `el setting model-access grant NAMESPACE/NAME --protocol anthropic [--revision FULL_COMMIT]` (project Admin) lets a published Runner receive the project's metered model key on Fly Trials; `list` and `revoke GRANT_ID` manage grants.
+
 Success follows native storage and original catalog readback and returns an `ans_` identity. Update publishes an explicit new revision. Delete checks active and unknown dependencies; Runner deletion covers all revisions. Historical bytes and quota usage remain. Commands default to the current login base URL/workspace/project, with no `--env`. Both Data grants and the corresponding resource capabilities are required.
 
 Read methods include `list/get/files/open/download`; `client.recipes.list/get` queries the original Recipe catalog. Downloads verify size/SHA-256 and refuse existing destinations, symlinks and storage redirects. Large payloads use multipart with stable idempotency keys. Staging `pub_` IDs and failure phases support retry; they are not native success. Matching client/server support is required. See the [Data reference](https://docs.envloop.ai/reference/cli#data-commands).

{envloop_cli-1.23.0 → envloop_cli-1.27.0}/envloop/_optional.py

@@ -1,4 +1,4 @@
-"""Lazy, static adapters for independently installed client features.
+"""Lazy, static adapters for features bundled in the single client distribution.
 
 No feature code is imported when the base client is imported or constructed.
 Legacy module names resolve to their single feature owner on explicit import.
@@ -24,7 +24,7 @@ FEATURE_COMMANDS = {
     },
     'experiment': {
         'job': 'Submit and manage Agent, Sandbox, Train and Inference Jobs',
-        'trail': 'Inspect and control existing trials',
+        'trial': 'Inspect and control existing Trials',
     },
 }
 
@@ -35,10 +35,10 @@ MODULES: dict[str, str] = {
 }
 CLIENTS = {
     **{name: ('data', 'client') for name in ('ProfileService', 'RunnerService', 'Profiles', 'Runners')},
-    **{name: ('experiment', 'client') for name in ('TaskService', 'RpcService', 'load_task', 'validate_command', 'dataset_selector', 'Tasks', 'runner_selector', 'job_budget', 'Jobs')},
+    **{name: ('experiment', 'client') for name in ('TrialService', 'RpcService', 'load_task', 'validate_command', 'dataset_selector', 'Trials', 'runner_selector', 'job_budget', 'Jobs')},
     **{name: ('base', 'downloads') for name in ('ArtifactError', 'artifact_path')},
 }
-DATA_ENDPOINTS = set('open_download resolve_dataset list_runners profile_path put_profile list_profiles get_profile resolve_profile set_default_profile archive_profile revoke_profile_secret export_profile import_profile open_checkpoint_archive'.split())
+DATA_ENDPOINTS = set('open_download resolve_dataset list_runners get_runner_revision register_runner_revision runner_path profile_path put_profile list_profiles get_profile resolve_profile set_default_profile archive_profile revoke_profile_secret export_profile import_profile open_checkpoint_archive'.split())
 EXPERIMENT_ENDPOINTS = set('submit_job get_job get_job_result get_job_logs list_jobs job_action job_events stop_job list_trials get_task_run get_trial_detail get_trial_training list_trial_metrics list_trial_checkpoints get_task_result get_task_snapshot stop_task list_trial_artifacts open_trial_artifact'.split())
 
 
@@ -48,7 +48,23 @@ class MissingFeatureError(ImportError):
 
     def __init__(self, feature: str):
         self.feature = feature
-        super().__init__(f'Install optional functionality with: pip install "envloop-cli[{feature}]"')
+        super().__init__(self.instruction(feature))
+
+    def instruction(self, feature: str) -> str:
+        return f'Install optional functionality with: pip install "envloop-cli[{feature}]"'
+
+
+class IncompleteInstallationError(MissingFeatureError):
+    """Feature code ships inside envloop-cli, so a missing module means removed files.
+
+    Uninstalling the separate envloop-cli-experiment package of 1.23.0/1.24.0 after an
+    upgrade deletes files it shared with the single distribution; an extra cannot restore them.
+    """
+
+    @override
+    def instruction(self, feature: str) -> str:
+        return ('Client files are missing. Reinstall with: '
+                'pip uninstall -y envloop-cli envloop-cli-experiment && pip install envloop-cli')
 
 
 class CommandExit(Exception):
@@ -60,9 +76,15 @@ def available(feature: str):
     return importlib.util.find_spec(FEATURES[feature]) is not None
 
 
+def require_data_dependencies():
+    """Only ANS operations need the public protocol/download dependency."""
+    if importlib.util.find_spec('envhub') is None:
+        raise MissingFeatureError('data')
+
+
 def feature_module(feature: str, module: str):
     if not available(feature):
-        raise MissingFeatureError(feature)
+        raise IncompleteInstallationError(feature)
     return importlib.import_module(f'{FEATURES[feature]}.{module}')
 
 
@@ -107,7 +129,7 @@ def client_resource(client: Any, name: str):
         return getattr(importlib.import_module(f'envloop.{module}'), cls)(server)
     if name == 'dashboard':
         if available('experiment'):
-            return feature_attribute('experiment', 'navigation_resources', 'Dashboard')(client.project, client.jobs, client.train)
+            return feature_attribute('experiment', 'navigation_resources', 'Dashboard')(client.project, client.jobs, client.training_jobs)
         from .navigation_resources import Dashboard
         return Dashboard(client.project)
     if name in {'profiles', 'runners'}:
@@ -121,17 +143,13 @@ def client_resource(client: Any, name: str):
     if name == 'jobs':
         return feature_attribute('experiment', 'client', 'Jobs')(server)
     if name == 'trials':
-        return client.tasks
-    if name == 'tasks':
-        return feature_attribute('experiment', 'client', 'Tasks')(server)
+        return feature_attribute('experiment', 'client', 'Trials')(server)
     if name == 'training_jobs':
-        return client.train
-    if name == 'train':
         return feature_attribute('experiment', 'train', 'TrainJobs')(server)
     if name == 'decision':
         return feature_attribute('experiment', 'decision', 'DecisionClient')(server)
     if name == 'agent_runs':
-        return feature_attribute('rsi-loop', 'agent_runs', 'AgentRuns')(client.train)
+        return feature_attribute('rsi-loop', 'agent_runs', 'AgentRuns')(client.training_jobs)
     if name in {'endpoints', 'adapters'}:
         return feature_attribute('experiment', 'resources', name.capitalize())(server)
     if name == 'inference':
@@ -140,7 +158,7 @@ def client_resource(client: Any, name: str):
 
 
 def execute_command(args: argparse.Namespace, client: Any, cli: Any):
-    if args.group in {'job', 'trail'}:
+    if args.group in {'job', 'trial'}:
         return getattr(feature_module('experiment', 'cli'), 'execute_' + args.group)(args, client, cli)
     if args.group == 'profile':
         return feature_module('data', 'cli').execute_profile(args, client, cli)

{envloop_cli-1.23.0 → envloop_cli-1.27.0}/envloop/auth_cli.py

@@ -16,7 +16,7 @@ def add_auth_commands(groups: "argparse._SubParsersAction[argparse.ArgumentParse
     login = commands.add_parser("login", help="Open the Clerk sign-in page and authorize CLI/SDK access")
     login.add_argument("--base-url", help="API Base URL; defaults to the effective connection or official API")
     login.add_argument("--no-browser", action="store_true", help="Print the login link without opening a browser")
-    login.add_argument("--timeout", type=int, choices=range(1, 601), metavar="SECONDS", default=300,
+    login.add_argument("--timeout-seconds", type=int, choices=range(1, 601), metavar="SECONDS", default=300,
                        help="Wait up to 1–600 seconds (default: 300)")
     configure = commands.add_parser("config", help="Save Base URL and API key for CLI and SDK")
     configure.add_argument("--no-check", action="store_true", help="Save without API verification (offline; key remains unverified)")

{envloop_cli-1.23.0 → envloop_cli-1.27.0}/envloop/auth_login.py

@@ -70,13 +70,13 @@ def login(args: argparse.Namespace, store: Any) -> int:
 
     saved, credential = store.workspace_snapshot()
     url = normalize_url(args.base_url or resolve_cli_connection().api_url or DEFAULT_BASE_URL)
-    deadline = time.monotonic() + args.timeout
+    deadline = time.monotonic() + args.timeout_seconds
     key, poll_token = 'epk_' + secrets.token_hex(32), secrets.token_urlsafe(32)
     code = ''.join(secrets.choice('ABCDEFGHJKLMNPQRSTUVWXYZ23456789') for _ in range(8))
     def digest(value: str) -> str:
         return hashlib.sha256(value.encode()).hexdigest()
     response = _request(url, 'start', {'key_hash': digest(key), 'poll_hash': digest(poll_token),
-                                    'code_hash': digest(code)}, min(5, args.timeout))
+                                    'code_hash': digest(code)}, min(5, args.timeout_seconds))
     identifier, uri = _verification_url(url, response)
     interval, expires = response.get('interval'), response.get('expires_in')
     if (isinstance(interval, bool) or not isinstance(interval, (int, float)) or not 0 < interval <= 10 or

{envloop_cli-1.23.0 → envloop_cli-1.27.0}/envloop/cli.py

@@ -14,7 +14,7 @@ from .http import CLI_VERSION
 
 SECTIONS = {
     'OVERVIEW': ('dashboard',),
-    'EXPERIMENT': ('job', 'trail'),
+    'EXPERIMENT': ('job', 'trial'),
     'DATA': ('dataset', 'data', 'runner', 'checkpoint', 'profile'),
     'MANAGE': ('billing', 'limit', 'member', 'setting'),
     'TOOLS': ('auth', 'help'),
@@ -43,6 +43,8 @@ class HelpArgumentParser(argparse.ArgumentParser):
 
     def __init__(self, *args: Any, **kwargs: Any):
         kwargs.setdefault('formatter_class', CommandHelpFormatter)
+        # Prefixes of an option would be unlisted aliases; only full option names are accepted.
+        kwargs.setdefault('allow_abbrev', False)
         super().__init__(*args, **kwargs)
         self.add_argument("--no-color", action="store_true", default=argparse.SUPPRESS,
                           help="Disable terminal highlighting")
@@ -140,7 +142,7 @@ def parser():
     cli = HelpArgumentParser(
         prog=Path(sys.argv[0]).name if Path(sys.argv[0]).name in {"envloop", "el"} else "el",
         usage="%(prog)s [options] [command]", formatter_class=CommandHelpFormatter,
-        description="Manage Agent, Sandbox, Train and Inference Jobs and worker trials with EnvLoop.\nConfigure your connection with 'auth config'.",
+        description="Manage Agent, Sandbox, Train and Inference Jobs and their Trials with EnvLoop.\nConfigure your connection with 'auth config'.",
     )
     cli._optionals.title = "Options"
     cli.set_defaults(api_url=None, workspace_id=None, no_color=False)
@@ -222,18 +224,18 @@ def _main(argv: list[str] | None = None):
         disable_color()
     if any(value == "--api-key" or value.startswith("--api-key=") for value in argv):
         cli.error("API keys must be entered with auth config or supplied through ENVLOOP_API_KEY")
-    from ._optional import MissingFeatureError, available
+    from ._optional import IncompleteInstallationError, available
     requested = next((value for value in argv if not value.startswith('-')), None)
-    feature = cast(dict[str | None, str], {'job': 'experiment', 'trail': 'experiment', 'data': 'data', 'dataset': 'data',
+    feature = cast(dict[str | None, str], {'job': 'experiment', 'trial': 'experiment', 'data': 'data', 'dataset': 'data',
                'runner': 'data', 'checkpoint': 'data', 'profile': 'data'}).get(requested)
     if feature is not None and not available(feature):
         from .cli_errors import format_error
-        print(format_error(MissingFeatureError(feature)), file=sys.stderr)
+        print(format_error(IncompleteInstallationError(feature)), file=sys.stderr)
         return 1
     job_action = next((value for value in argv[argv.index('job') + 1:] if not value.startswith('-')), None) if requested == 'job' else None
     if job_action == 'run' and not available('rsi-loop'):
         from .cli_errors import format_error
-        print(format_error(MissingFeatureError('rsi-loop')), file=sys.stderr)
+        print(format_error(IncompleteInstallationError('rsi-loop')), file=sys.stderr)
         return 1
     args = cli.parse_args(argv)
     if args.group == "help":
@@ -254,7 +256,7 @@ def _main(argv: list[str] | None = None):
 
             print(format_error(exc), file=sys.stderr)
             return 1
-    if args.group in {'job', 'trail'}:
+    if args.group in {'job', 'trial'}:
         from envloop_experiment.cli import validate
         status = validate(args, cli)
         if status is not None:
@@ -308,6 +310,11 @@ def _main(argv: list[str] | None = None):
             print(json.dumps(operation_output(exc.operation), ensure_ascii=False))
             print(format_error(exc), file=sys.stderr)
             return 1
+        from .http import WaitTimeoutError
+        if isinstance(exc, WaitTimeoutError) and isinstance(exc.last, dict):
+            # The work continues on the server; keep its ID so it can be queried or waited on again.
+            last = operation_output(exc.last) if 'operation_id' in exc.last else exc.last
+            print(json.dumps(last, ensure_ascii=False))
         # Do not emit arbitrary exception messages, file content or credentials.
         print(format_error(exc), file=sys.stderr)
         return 1

{envloop_cli-1.23.0 → envloop_cli-1.27.0}/envloop/cli_errors.py

@@ -17,7 +17,7 @@ from collections.abc import Iterable
 from datetime import datetime
 from typing import Any, cast
 
-from .http import RemoteError
+from .http import GATEWAY_TIMEOUT_STATUSES, RemoteError
 
 _MESSAGES: dict[str, str] = {}
 
@@ -45,6 +45,13 @@ _register("EDGE_REJECTED", "The network edge (Cloudflare) rejected the request b
 _register("HUMAN_SESSION_REQUIRED WORKSPACE_LIST_REQUIRES_SESSION PROJECT_LIST_REQUIRES_SESSION", "This operation requires a browser session authorized by the server.")
 _register("CONFLICT VERSION_CONFLICT RESOURCE_VERSION_CONFLICT PRECONDITION_FAILED IDEMPOTENCY_CONFLICT", "The request conflicts with the current resource version or an earlier request.")
 _register("BAD_REQUEST VALIDATION_ERROR VALIDATION_FAILED INVALID_REQUEST INVALID_ARGUMENT INVALID_CURSOR INVALID_PROFILE INVALID_PROFILE_SECRETS", "The request or input is invalid.")
+_register("PROFILE_SECRET_NAME_INVALID", "A Secret parameter name must look like a credential, as the platform decides: it ends with API_KEY or _TOKEN, or contains PASSWORD, SECRET, CREDENTIAL, AUTHORIZATION, ACCESS_TOKEN, AUTH_TOKEN, BEARER_TOKEN or PRIVATE_KEY. Otherwise list the name in the Profile's secret_params. Names use letters, digits and underscores.")
+_register("PROFILE_SECRET_VALUE_INVALID", "A Secret parameter value is missing or unusable: it must be non-empty UTF-8 text of at most 64 KiB from a prompt, stdin, an environment variable or a file.")
+_register("PROFILE_SECRET_CONFLICT", "Give each Secret parameter once, by --secret only (not also in the Profile file or patch), and read at most one value from stdin.")
+_register("RUNNER_PACKAGE_INVALID", "The Runner package directory is invalid. It needs runner.toml whose [runner] name matches the Runner and has no version, the [execution] entrypoint main.py or main.sh, at most 100 regular files, no symlinks, and path parts of letters, digits, '_', '.' or '-' that do not start with '.'.")
+_register("RUNNER_REVISION_INVALID", "Use the full lowercase 40- or 64-character source commit as --revision and a NAMESPACE/NAME Runner name.")
+_register("RUNNER_SOURCE_REF_INVALID", "Use a branch or tag name such as main or v1 with --source-ref-kind branch, tag or commit; HEAD, latest, default, release, '..', spaces and ~^:?*[\\ are not allowed.")
+_register("RUNNER_REVISION_CONFLICT", "The platform catalog holds different package or index hashes for this Runner revision. Publish a new revision instead of changing an existing one.")
 _register("DATASET_REQUIRED", "Missing Dataset input (--dataset or --source) for this Agent Job.")
 _register("DATASET_CONFIG_REQUIRED", "An explicit Dataset configuration (--config) is required.")
 _register("RUNNER_REQUIRED", "Missing Runner input (--runner) for this Job.")
@@ -54,6 +61,7 @@ _register("RATE_LIMITED RATE_LIMIT_EXCEEDED TOO_MANY_REQUESTS", "The API rate li
 _register("SERVICE_UNAVAILABLE INTERNAL_SERVER_ERROR EXECUTOR_UNAVAILABLE PROVIDER_UNAVAILABLE LOG_STORAGE_UNAVAILABLE", "The service is unavailable or encountered an internal error.")
 _register("INTERNAL_ERROR", "The command encountered an unexpected error.")
 _register("NETWORK_ERROR", "A connection to the API or download service could not be completed.")
+_register("SOURCE_UNAVAILABLE", "The API gateway did not get a valid answer from the service, for example because it stopped waiting. A write may still have been processed: read the Job or resource before retrying it.")
 _register("REQUEST_TIMEOUT", "The API did not answer before the client timeout. The request may still have been processed: read the Job or resource before retrying a write.")
 _register("RESPONSE_INTERRUPTED", "The connection closed before the API finished answering; the service may have failed or restarted. The request may have been processed: read the Job or resource before retrying a write, and report the time if it repeats.")
 _register("INVALID_RESPONSE RESPONSE_TOO_LARGE INVALID_LOG_PAGE INVALID_PAGINATION TRAIN_JOB_CURSOR_REPEATED", "The service returned an invalid or unsupported response.")
@@ -114,6 +122,8 @@ _register("INVALID_RSI_REQUEST INVALID_STEER INVALID_DIRECTIVE", "The RSI servic
 _register("RSI_CONFLICT RSI_RUN_NOT_ACTIVE", "The run id exists, the run is not in a state that allows this action, or no loop machine is free.")
 _register("RSI_CONTROL_UNCONFIGURED RSI_CONTROL_UNAVAILABLE RSI_CONTROL_TIMEOUT", "The RSI loop service is not available through this platform right now.")
 
+_register("UNSUPPORTED_ADAPTER_SOURCE", 'Register a training checkpoint with {"serving_profile": "NAME", "source": {"type": "train_checkpoint", "uri": "envloop://..."}}. '
+          "List serving profiles with `el job serving-profile list` and checkpoint paths with `el checkpoint list`.")
 _register("JOB_ACTION_UNSUPPORTED", "This Job type does not support the requested action.")
 _register("INFERENCE_ROUTE_RETIRED", "The global inference route has been retired.")
 
@@ -139,14 +149,14 @@ for _code, _resource in {
 _KNOWN_REMOTE_CODES = set("""
 ADAPTER_FILES_MISSING ADAPTER_FORMAT_UNSUPPORTED ADAPTER_RANK_EXCEEDS_PROFILE ADAPTER_TARGET_UNSUPPORTED
 DATASET_SELECTION_UNSUPPORTED INVALID_ADAPTERS INVALID_ADAPTER_ALIAS INVALID_AMOUNT INVALID_ATTEMPTS
-INVALID_BILLING_WINDOW INVALID_CHECKPOINT_PATH INVALID_CLEANUP_RESOURCES INVALID_COMMAND INVALID_DATASET_DELIVERY
+INVALID_BILLING_WINDOW INVALID_CHECKPOINT_PATH INVALID_CHECKPOINT_URI INVALID_CLEANUP_RESOURCES INVALID_COMMAND INVALID_DATASET_DELIVERY
 INVALID_DATASET_INPUT INVALID_DATASET_SELECTOR INVALID_ENDPOINT_NAME INVALID_ENFORCEMENT INVALID_ENTRY_ID
 INVALID_ENTRY_KIND INVALID_ENVHUB_REQUEST INVALID_GROUP_BY INVALID_JOB INVALID_JOB_ID INVALID_JOB_MEMBER
 INVALID_JOB_MODE INVALID_JOB_SIZE INVALID_JOB_STATUS INVALID_JOB_TYPE INVALID_LIMIT INVALID_LOG_CURSOR
 INVALID_LOG_PROCESS INVALID_LOG_QUERY INVALID_LOG_SOURCE INVALID_LOG_STREAM INVALID_PRIORITY
 INVALID_PROFILE_COMBINATION INVALID_PROJECT INVALID_QUOTE INVALID_REPOSITORY_COMMIT INVALID_REPOSITORY_ROOT
 INVALID_REQUEST_ID INVALID_RESOLVED_JOB INVALID_RUNNER INVALID_RUNNER_ENVIRONMENT INVALID_SANDBOX_RUNNER
-INVALID_SCALING INVALID_SCOPE_TTL INVALID_TAG INVALID_TINKER_PATH INVALID_TRAINING_LINEAGE INVALID_TRIALS
+INVALID_SCALING INVALID_SCOPE_TTL INVALID_TAG INVALID_TRAINING_LINEAGE INVALID_TRIALS
 INVALID_TRIAL_STATUS INVALID_WORKSPACE JOB_REQUIRED LEADERBOARD_PROVENANCE_REQUIRED LEADERBOARD_SCORE_REQUIRED
 PROFILE_IDENTITY_MISMATCH QUEUE_NOT_FOUND REPOSITORY_BINDING_REQUIRED RUNNER_GPU_UNAVAILABLE RUNNER_HASH_MISMATCH
 TOO_MANY_ADAPTERS UNSUPPORTED_ADAPTER_SOURCE UNSUPPORTED_CAPABILITY UNSUPPORTED_COMMAND UNSUPPORTED_HARDWARE
@@ -209,6 +219,9 @@ def _http_message(status: int | None) -> str:
         return _MESSAGES["INVALID_REQUEST"]
     if status == 429:
         return _MESSAGES["RATE_LIMITED"]
+    if status in GATEWAY_TIMEOUT_STATUSES:
+        return ("The API gateway stopped waiting for the service. The request may still have been processed: "
+                "read the Job or resource before retrying a write.")
     if status is not None and status >= 500:
         return _MESSAGES["SERVICE_UNAVAILABLE"]
     return _MESSAGES["HTTP_ERROR"]
@@ -294,7 +307,9 @@ def format_error(error: BaseException) -> str:
     extra: dict[str, Any] = {"hint": remote.hint, "details": remote.details, "request_id": remote.request_id} if remote else {}
     if remote and remote.message:
         # Local text only adds guidance for codes the CLI documents.
-        guidance = local if code in _MESSAGES and code != "HTTP_ERROR" and local != remote.message else None
+        # A gateway timeout always adds that the request may have been processed.
+        documented = (code in _MESSAGES and code != "HTTP_ERROR") or status in GATEWAY_TIMEOUT_STATUSES
+        guidance = local if documented and local != remote.message else None
         rendered = error_block(code, remote.message, status=status, guidance=guidance, **extra)
     else:
         rendered = error_block(code, local, status=status, **extra)

{envloop_cli-1.23.0 → envloop_cli-1.27.0}/envloop/cli_output.py

@@ -49,7 +49,7 @@ def next_page(args: argparse.Namespace, cursor: object, program: str) -> str:
     if getattr(args, 'no_color', False):
         command.append('--no-color')
     from ._optional import feature_module
-    feature = 'data' if args.group in {'data', 'dataset', 'runner', 'checkpoint', 'profile'} else 'experiment' if args.group in {'job', 'trail'} else None
+    feature = 'data' if args.group in {'data', 'dataset', 'runner', 'checkpoint', 'profile'} else 'experiment' if args.group in {'job', 'trial'} else None
     if feature is not None:
         feature_module(feature, 'cli_output').pagination(command, args, cursor)
     else:
@@ -59,15 +59,31 @@ def next_page(args: argparse.Namespace, cursor: object, program: str) -> str:
     return "Next page: " + shlex.join(command)
 
 
+# Empty-list text names the listed resource, not the command group (`job adapter list` lists adapters).
+EMPTY_RESOURCES = {
+    ('job', 'adapter'): 'adapters', ('job', 'serving-profile'): 'serving profiles',
+    ('trial',): 'trials', ('trial', 'artifact'): 'artifacts',
+    ('dataset',): 'datasets', ('dataset', 'revision'): 'dataset revisions', ('dataset', 'task'): 'dataset tasks',
+    ('data',): 'resources', ('data', 'files'): 'files',
+    ('billing', 'usage'): 'usage records', ('billing', 'ledger'): 'ledger entries', ('billing', 'price'): 'prices',
+    ('member',): 'members', ('member', 'candidate'): 'member candidates',
+    ('setting', 'registry'): 'registry credentials',
+}
+
+
+def empty_message(args: argparse.Namespace) -> str:
+    path = getattr(args, '_command_path', [args.group, 'list'])
+    resource = EMPTY_RESOURCES.get(tuple(path[:-1] if path[-1:] == ['list'] else path))
+    return f'No {resource or args.group} found.'
+
+
 def generic_layout(items: list[dict[str, Any]], args: argparse.Namespace, layouts: dict[str, tuple[str, ...]]) -> tuple[tuple[str, ...], Callable[[dict[str, Any]], tuple[object, ...]], str]:
     fields = tuple(field for field in layouts.get(args.group, ()) if any(field in item for item in items))
     fields = fields or tuple(dict.fromkeys(key for item in items for key in item))[:6]
     headers = tuple(field.replace('_', ' ').upper() for field in fields)
     render: Callable[[dict[str, Any]], tuple[object, ...]] = lambda item: tuple(timestamp(item.get(field)) if field.endswith('_at') or field == 'at'
                                 else item.get(field) for field in fields)
-    empty = ('No registry credentials found.' if args.group == 'setting'
-             else f'No {getattr(args, "resource", args.group)} found.')
-    return headers, render, empty
+    return headers, render, empty_message(args)
 
 
 def list_output(result: Any, args: argparse.Namespace, program: str) -> str:
@@ -79,7 +95,7 @@ def list_output(result: Any, args: argparse.Namespace, program: str) -> str:
         items = [{'name': key, 'value': value} for key, value in items.items()]
     from ._optional import feature_module
     feature = ('rsi-loop' if args.group == 'job' and args.action == 'run' else
-               'experiment' if args.group in {'job', 'trail'} else
+               'experiment' if args.group in {'job', 'trial'} else
                'data' if args.group in {'data', 'dataset', 'runner', 'checkpoint', 'profile'} else None)
     if feature is not None:
         headers, render, empty = feature_module(feature, 'cli_output').list_layout(items, args)

{envloop_cli-1.23.0 → envloop_cli-1.27.0}/envloop/client.py

@@ -33,9 +33,10 @@ class Client:
         raise RuntimeError(INTERACTIVE_TRAINING_REMOVED)
 
     @classmethod
-    def remote(cls, api_url: str | None = None, *, timeout: float = 30, api_key: str | None = None, workspace_id: str | None = None, project_id: str | None = None, config: bool = True):
+    def remote(cls, api_url: str | None = None, *, timeout: float = 90, api_key: str | None = None, workspace_id: str | None = None, project_id: str | None = None, config: bool = True):
         from .http import HttpService
 
+        # The default matches http.DEFAULT_TIMEOUT_SEC (longer than the hosted proxy's 60 s wait).
         return cls(HttpService(api_url, timeout, api_key=api_key, workspace_id=workspace_id, project_id=project_id, config=config))
 
     @classmethod

{envloop_cli-1.23.0 → envloop_cli-1.27.0}/envloop/http.py

@@ -27,6 +27,14 @@ def validate_public_execution(value: Any) -> None:
         validate_public_execution(value.get(key))
 
 
+class WaitTimeoutError(TimeoutError):
+    """A wait ended before the terminal state; ``last`` is the last resource or Operation read."""
+
+    def __init__(self, message: str, last: dict[str, Any] | None = None):
+        super().__init__(message)
+        self.last = last
+
+
 class RemoteError(RuntimeError):
     """Platform failure. Server-supplied text lives in bounded attributes, never in str()."""
 
@@ -63,6 +71,12 @@ def retry_after_seconds(value: str | None) -> float | None:
         return None
 
 
+# The hosted API proxy waits up to 60 s for the platform. Wait longer by default so a slow answer, or the
+# proxy's own gateway timeout, reaches the caller instead of a client-side timeout. Explicit values are kept.
+DEFAULT_TIMEOUT_SEC = 90.0
+# The proxy or edge stopped waiting while the platform was still working: the request may have been processed.
+GATEWAY_TIMEOUT_STATUSES = frozenset({504, 524})
+
 # Transport failures that are not an HTTP answer. A request that was sent and then timed out or lost its
 # response may still have been processed, unlike one that never connected.
 TRANSIENT_CODES = frozenset({"NETWORK_ERROR", "REQUEST_TIMEOUT", "RESPONSE_INTERRUPTED"})
@@ -213,7 +227,7 @@ def encode_json(body: Any) -> bytes:
 
 
 class HttpService:
-    def __init__(self, base_url: str | None = None, timeout: float = 30, *, api_key: str | None = None, workspace_id: str | None = None, project_id: str | None = None, config: bool = True, environment: bool = True):
+    def __init__(self, base_url: str | None = None, timeout: float = DEFAULT_TIMEOUT_SEC, *, api_key: str | None = None, workspace_id: str | None = None, project_id: str | None = None, config: bool = True, environment: bool = True):
         from .config import require_remote, resolve_connection
 
         connection = resolve_connection(base_url, api_key, workspace_id, project_id, config=config, environment=environment)
@@ -233,9 +247,9 @@ class HttpService:
         self.opener = build_client_opener()
 
     def remote_error(self, error: HTTPError, body: Any = None, **options: Any) -> RemoteError:
-        submitted = body.get("secrets") if isinstance(body, dict) else None
-        submitted = [value for value in submitted.values() if isinstance(value, str)] if isinstance(submitted, dict) else []
-        fields = error_fields(error, (self.secret, *submitted))
+        from .profile_secrets import submitted_secrets
+
+        fields = error_fields(error, (self.secret, *submitted_secrets(body)))
         return RemoteError(fields.pop("code"), error.code, **fields, **options)
 
     def request(self, method: str, path: str, body: Any = None, *, headers: dict[str, str] | None = None) -> Any: