sequence-base 0.4.11__tar.gz → 0.4.13__tar.gz

{sequence_base-0.4.11 → sequence_base-0.4.13}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: sequence-base
-Version: 0.4.11
+Version: 0.4.13
 Summary: Shared foundation for the Sequences platform — wire contract, correlation, error taxonomy, constants. Used by the robot SDK, gateway, and container harness.
 Project-URL: Changelog, https://generalsequences.com/docs/changelog/
 Requires-Python: >=3.10

{sequence_base-0.4.11 → sequence_base-0.4.13}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "sequence-base"
-version = "0.4.11"
+version = "0.4.13"
 description = "Shared foundation for the Sequences platform — wire contract, correlation, error taxonomy, constants. Used by the robot SDK, gateway, and container harness."
 requires-python = ">=3.10"
 dependencies = ["pydantic>=2.0"]

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/__init__.py

@@ -37,8 +37,6 @@ from sequence_base.contract import (
     ActionContract,
     ActionSpace,
     ActionStep,
-    AdapterKind,
-    AdapterTransform,
     Autoscaler,
     BenchmarkConditionSpec,
     BenchmarkRegisterRequest,
@@ -75,7 +73,6 @@ from sequence_base.contract import (
     ImageStepRun,
     ImageStepUvPip,
     Lease,
-    LossyTransform,
     MetricAgg,
     Observation,
     PolicyRegistration,
@@ -90,7 +87,6 @@ from sequence_base.contract import (
     SecretRef,
     SeqFlags,
     Step,
-    SuiteAgg,
     TaskAgg,
     Timeouts,
     Variant,
@@ -146,8 +142,7 @@ __all__ = [
 
     "Variant", "Conditions", "RenderSpec", "ReplaySpec", "Step", "BenchmarkSpec", "BenchmarkConditionSpec", "BenchmarkRegisterRequest", "BenchmarkStatusResponse",
     "EvalRequest", "ClampInfo", "Estimate", "EvalResponse", "EvalDryRunResponse", "RolloutResult",
-    "EvalStatus", "SuiteAgg", "TaskAgg", "MetricAgg", "LossyTransform", "EvalReport",
-    "AdapterKind", "AdapterTransform",
+    "EvalStatus", "TaskAgg", "MetricAgg", "EvalReport",
 
     "SecretRef", "SecretCreateRequest", "SecretInfo", "BootstrapResponse",
 

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/constants/limits.py

@@ -14,3 +14,8 @@ MAX_USD = 100_000
 
 TASK_DESCRIPTION_MAX = 300
 
+
+
+METRIC_UNITS_MAX = 64
+METRIC_UNIT_MAX = 16
+

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/constants/version.py

@@ -4,7 +4,9 @@
 
 
 
-CONTRACT_VERSION = "0.5.10"
+CONTRACT_VERSION = "0.5.12"
+
+
 
 
 

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/contract/__init__.py

@@ -72,8 +72,6 @@ from sequence_base.contract.benchmarks import (
     Variant,
 )
 from sequence_base.contract.eval import (
-    AdapterKind,
-    AdapterTransform,
     ClampInfo,
     Estimate,
     EvalDryRunResponse,
@@ -81,10 +79,8 @@ from sequence_base.contract.eval import (
     EvalRequest,
     EvalResponse,
     EvalStatus,
-    LossyTransform,
     MetricAgg,
     RolloutResult,
-    SuiteAgg,
     TaskAgg,
 )
 
@@ -134,8 +130,7 @@ __all__ = [
 
     "Variant", "Conditions", "RenderSpec", "ReplaySpec", "Step", "BenchmarkSpec", "BenchmarkConditionSpec", "BenchmarkRegisterRequest", "BenchmarkStatusResponse",
     "EvalRequest", "ClampInfo", "Estimate", "EvalResponse", "EvalDryRunResponse", "RolloutResult",
-    "EvalStatus", "SuiteAgg", "TaskAgg", "MetricAgg", "LossyTransform", "EvalReport",
-    "AdapterKind", "AdapterTransform",
+    "EvalStatus", "TaskAgg", "MetricAgg", "EvalReport",
 
     "SecretRef", "SecretCreateRequest", "SecretInfo", "BootstrapResponse",
 

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/contract/benchmarks.py

@@ -16,7 +16,7 @@ from typing import Literal
 
 from pydantic import BaseModel, ConfigDict, Field, model_validator
 
-from sequence_base.constants.limits import MAX_ENVS, TASK_DESCRIPTION_MAX
+from sequence_base.constants.limits import MAX_ENVS, METRIC_UNIT_MAX, METRIC_UNITS_MAX, TASK_DESCRIPTION_MAX
 from sequence_base.contract.accounts import SecretRef
 from sequence_base.contract.deployments import CodeBundle, DeploymentState, InferParam
 from sequence_base.contract.flags import BENCHMARK_BATCHED, BENCHMARK_BINDABLE, SeqFlags
@@ -167,10 +167,6 @@ class BenchmarkConditionSpec(BaseModel):
     timeout_s: float | None = Field(default=None, gt=0)
 
 
-
-    suite: str | None = None
-
-
 class BenchmarkRegisterRequest(BaseModel):
     """POST /v1/benchmarks — register body, same types as the spec (ImageSpec / Conditions / VolumeRef /
     RenderSpec). Carries assets / render / cpu / memory_gb / max_steps so the gateway can build the image,
@@ -194,9 +190,6 @@ class BenchmarkRegisterRequest(BaseModel):
     code: CodeBundle | None = None
 
 
-    suite_of: dict[str, str] = Field(default_factory=dict)
-
-
 
     task_descriptions: dict[str, str] = Field(default_factory=dict)
 
@@ -206,6 +199,14 @@ class BenchmarkRegisterRequest(BaseModel):
     seeds_repeat: bool | None = None
 
 
+
+    success_metric: str | None = Field(default=None, min_length=1, max_length=200)
+
+
+
+    metric_units: dict[str, str] = Field(default_factory=dict, max_length=METRIC_UNITS_MAX)
+
+
     action_contract: ActionContract | None = None
 
 
@@ -245,6 +246,15 @@ class BenchmarkRegisterRequest(BaseModel):
                 raise ValueError(f"setup parameter {name!r} is not a public Python identifier")
         return self
 
+    @model_validator(mode="after")
+    def _metric_units(self) -> "BenchmarkRegisterRequest":
+        for metric, unit in self.metric_units.items():
+            if not metric or len(metric) > 200:
+                raise ValueError("metric_units names a metric of 1 to 200 characters")
+            if not unit.strip() or len(unit) > METRIC_UNIT_MAX:
+                raise ValueError(f"metric_units[{metric!r}] is a unit of 1 to {METRIC_UNIT_MAX} characters")
+        return self
+
     @model_validator(mode="after")
     def _task_descriptions(self) -> "BenchmarkRegisterRequest":
         """Each description names a task the benchmark declares - in `conditions` or `condition_list` - and is at most

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/contract/eval.py

@@ -20,7 +20,6 @@ __all__ = [
     "EvalRequest",
     "ClampInfo",
     "ArmReport",
-    "PairedDiff",
     "EvalTiming",
     "Estimate",
     "EvalResponse",
@@ -31,13 +30,9 @@ __all__ = [
     "EvalQueue",
     "RolloutResult",
     "EvalStatus",
-    "SuiteAgg",
     "TaskAgg",
     "MetricAgg",
-    "LossyTransform",
     "EvalReport",
-    "AdapterKind",
-    "AdapterTransform",
 ]
 
 
@@ -51,7 +46,7 @@ class EvalRequest(BaseModel):
     leave them out and take the earlier eval's. `dry_run` selects the estimate-only response shape below.
 
     The one body the platform accepts: the current knobs plus the legacy names an older SDK still sends — `gpus` /
-    `envs_per_gpu` for the instance counts, `suite`, `region` — each taken and mapped, or refused with the way on,
+    `envs_per_gpu` for the instance counts, `region` — each taken and mapped, or refused with the way on,
     never silently dropped. The gateway inherits this model rather than keeping its own."""
 
 
@@ -63,11 +58,6 @@ class EvalRequest(BaseModel):
     benchmark: str | None = Field(default=None, max_length=200)
 
 
-    suite: str | None = Field(default=None, max_length=200, json_schema_extra={"deprecated": True},
-                              description="Gone: a benchmark is one set of tasks, so there is no suite; a non-null "
-                                          "value is refused. Omit it (an older SDK still sends it).")
-
-
 
 
     trials: _Trials | dict[str, _Trials] | None = None
@@ -108,7 +98,6 @@ class EvalRequest(BaseModel):
 
 
     rerun_unsuccessful_of: str | None = Field(default=None, max_length=64)
-    metric: str | None = Field(default=None, max_length=64)
 
 
 
@@ -289,18 +278,6 @@ class EvalStatus(str, Enum):
     CANCELED = "canceled"
 
 
-class SuiteAgg(BaseModel):
-    """Per-suite aggregate of ONE user-named metric, as an eval from before contract 0.5.8 reported it: kept so those
-    reports stay readable. A benchmark is one set of tasks now, and no eval since reports one.
-"""
-
-    model_config = ConfigDict(extra="forbid")
-
-    suite: str
-    value: float
-    n: int
-
-
 class TaskAgg(BaseModel):
     """Per-task aggregate of ONE user-named metric — structured detail, not a bare dict."""
 
@@ -312,28 +289,17 @@ class TaskAgg(BaseModel):
 
 
 class MetricAgg(BaseModel):
-    """The aggregate of one user-named metric key across the eval — overall plus per-task breakdowns (and, in a
-    report from before contract 0.5.8, per-suite ones) and the sample count. The platform aggregates whatever the user returned in `Step.metrics`;
+    """The aggregate of one user-named metric key across the eval — overall plus per-task breakdowns and the sample
+    count. The platform aggregates whatever the user returned in `Step.metrics`;
     it neither defines nor injects `success_rate` (success is just one key the user may or may not name)."""
 
     model_config = ConfigDict(extra="forbid")
 
     overall: float
-    per_suite: list[SuiteAgg] = Field(default_factory=list)
     per_task: list[TaskAgg] = Field(default_factory=list)
     n: int
 
 
-class LossyTransform(BaseModel):
-    """A lossy adapter transform recorded in the report. Shape aligns with the lossy element of
-    AdapterTransform (both {transform, reason})."""
-
-    model_config = ConfigDict(extra="forbid")
-
-    transform: str
-    reason: str
-
-
 class ArmReport(BaseModel):
     """One arm of an eval (a model x one combination of swept inference parameters): its own metrics and episode
     counts, so arms are compared side by side and never pooled."""
@@ -346,8 +312,6 @@ class ArmReport(BaseModel):
 
     params: dict[str, int | float | bool | str] = Field(default_factory=dict)
     metrics: dict[str, MetricAgg] = Field(default_factory=dict)
-
-    intervals: dict[str, list[float]] = Field(default_factory=dict)
     episodes: int = 0
     failed: int = 0
     not_run: int = 0
@@ -383,19 +347,6 @@ class ArmParams(BaseModel):
     steps: int | None = None
 
 
-class PairedDiff(BaseModel):
-    """`compare`: one metric's difference between two arms over the seeds both ran (b - a), with its interval."""
-
-    model_config = ConfigDict(extra="forbid")
-
-    metric: str
-    arm_a: int
-    arm_b: int
-    n: int
-    mean_diff: float
-    interval: list[float]
-
-
 class EvalTiming(BaseModel):
     """Where an eval's time went: cold starts, and per step the simulation, rendering, the policy's batched forward and
     the network around it, plus the forward's average batch size."""
@@ -423,15 +374,14 @@ class EvalReport(BaseModel):
     )
     cost_usd: float
     dashboard_url: str
-    adapter_lossy: list[LossyTransform]
     videos: list[str] = Field(default_factory=list, description="mp4 pointers from the platform tool")
     incomplete: bool | None = None
     errors: list[str] = Field(
         default_factory=list, description="Why episodes or the run failed, so an incomplete report says why"
     )
 
+
     arms: list[ArmReport] | None = None
-    paired: list[PairedDiff] | None = None
     timing: EvalTiming | None = None
 
 
@@ -445,22 +395,9 @@ class EvalReport(BaseModel):
     arm_params: list[ArmParams] | None = None
 
 
-class AdapterKind(str, Enum):
-    LOSSLESS = "lossless"
-    LOSSY = "lossy"
-
-
-
 
+    units: dict[str, str] | None = None
 
 
 
-class AdapterTransform(BaseModel):
-    """One adapter transform: its kind (lossless / lossy), the transform's name, and an optional reason. A registered
-    adapter and an older eval report carry these fields; they are kept for compatibility and are no longer acted on."""
-
-    model_config = ConfigDict(extra="forbid")
-
-    kind: AdapterKind
-    transform: str
-    reason: str | None = None
+    success_metric: str | None = None

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/presets/robot_arm.py

@@ -12,7 +12,6 @@ resolving to these same classes.
 """
 from __future__ import annotations
 
-import math
 from enum import Enum
 from typing import Any, Literal
 
@@ -38,12 +37,18 @@ class ImageFrame(BaseModel):
         default=None, description="rgb (the default) or depth: a 16-bit PNG of millimetres, 0 where there is no reading")
 
 
+
+
+
+
+
 class Proprioception(BaseModel):
-    """Proprioceptive state. Only joint_positions is required; the rest is zero-padded per model."""
+    """Proprioceptive state. Every field is optional: a policy's observation contract (proprio) says which it reads,
+    and its worker requires those, each at its length."""
 
     model_config = ConfigDict(extra="forbid")
 
-    joint_positions: list[float] = Field(..., description="Joint angles, in radians")
+    joint_positions: list[float] | None = Field(default=None, description="Joint angles, in radians")
     joint_velocities: list[float] | None = None
 
     end_effector_pose: list[float] | None = Field(default=None, min_length=7, max_length=7)
@@ -124,22 +129,8 @@ class ActionContract(BaseModel):
     bounds: list[tuple[float, float]] | None = Field(default=None, max_length=256)
     max_delta: float | list[float] | None = None
 
-
-    on_violation: Literal["clip", "refuse"] | None = None
-
-
-
-
-    binarize: dict[int, float] | None = Field(
-        default=None, description="dimension -> threshold: the client gives the robot 1.0 above it, 0.0 otherwise")
-
     @model_validator(mode="after")
     def _limits_fit(self) -> "ActionContract":
-        for d, threshold in (self.binarize or {}).items():
-            if not 0 <= d < self.action_dim:
-                raise ValueError(f"binarize names dimension {d}; action_dim is {self.action_dim}")
-            if not math.isfinite(threshold):
-                raise ValueError(f"binarize[{d}] is {threshold}, not a finite threshold")
         if self.bounds is not None:
             if len(self.bounds) != self.action_dim:
                 raise ValueError(f"bounds has {len(self.bounds)} [low, high] pairs; action_dim is {self.action_dim}")

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/schema_export.py

@@ -67,17 +67,14 @@ from sequence_base.contract.volumes import (
     VolumeUploadResponse,
 )
 from sequence_base.contract.eval import (
-    AdapterTransform,
     ClampInfo,
     Estimate,
     EvalDryRunResponse,
     EvalReport,
     EvalRequest,
     EvalResponse,
-    LossyTransform,
     MetricAgg,
     RolloutResult,
-    SuiteAgg,
     TaskAgg,
 )
 from sequence_base.contract.flags import seqflags_json_schema
@@ -167,12 +164,9 @@ WIRE_MODELS: dict[str, type[BaseModel]] = {
     "EvalResponse": EvalResponse,
     "EvalDryRunResponse": EvalDryRunResponse,
     "RolloutResult": RolloutResult,
-    "SuiteAgg": SuiteAgg,
     "TaskAgg": TaskAgg,
     "MetricAgg": MetricAgg,
-    "LossyTransform": LossyTransform,
     "EvalReport": EvalReport,
-    "AdapterTransform": AdapterTransform,
 
     "SecretRef": SecretRef,
     "SecretCreateRequest": SecretCreateRequest,

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/schemas/ActRequest.json

@@ -79,5 +79,5 @@
   ],
   "title": "ActRequest",
   "type": "object",
-  "x-contract-version": "0.5.10"
+  "x-contract-version": "0.5.12"
 }

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/schemas/ActResponse.json

@@ -89,5 +89,5 @@
   ],
   "title": "ActResponse",
   "type": "object",
-  "x-contract-version": "0.5.10"
+  "x-contract-version": "0.5.12"
 }

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/schemas/ActionChunk.json

@@ -99,5 +99,5 @@
   ],
   "title": "ActionChunk",
   "type": "object",
-  "x-contract-version": "0.5.10"
+  "x-contract-version": "0.5.12"
 }

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/schemas/ActionContract.json

@@ -23,22 +23,6 @@
     "action_space": {
       "$ref": "#/$defs/ActionSpace"
     },
-    "binarize": {
-      "anyOf": [
-        {
-          "additionalProperties": {
-            "type": "number"
-          },
-          "type": "object"
-        },
-        {
-          "type": "null"
-        }
-      ],
-      "default": null,
-      "description": "dimension -> threshold: the client gives the robot 1.0 above it, 0.0 otherwise",
-      "title": "Binarize"
-    },
     "bounds": {
       "anyOf": [
         {
@@ -87,22 +71,6 @@
       "default": null,
       "title": "Max Delta"
     },
-    "on_violation": {
-      "anyOf": [
-        {
-          "enum": [
-            "clip",
-            "refuse"
-          ],
-          "type": "string"
-        },
-        {
-          "type": "null"
-        }
-      ],
-      "default": null,
-      "title": "On Violation"
-    },
     "robot": {
       "title": "Robot",
       "type": "string"
@@ -116,5 +84,5 @@
   ],
   "title": "ActionContract",
   "type": "object",
-  "x-contract-version": "0.5.10"
+  "x-contract-version": "0.5.12"
 }

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/schemas/ActionStep.json

@@ -19,5 +19,5 @@
   ],
   "title": "ActionStep",
   "type": "object",
-  "x-contract-version": "0.5.10"
+  "x-contract-version": "0.5.12"
 }

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/schemas/AddLocal.json

@@ -23,5 +23,5 @@
   ],
   "title": "AddLocal",
   "type": "object",
-  "x-contract-version": "0.5.10"
+  "x-contract-version": "0.5.12"
 }

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/schemas/Autoscaler.json

@@ -69,5 +69,5 @@
   },
   "title": "Autoscaler",
   "type": "object",
-  "x-contract-version": "0.5.10"
+  "x-contract-version": "0.5.12"
 }

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/schemas/BenchmarkConditionSpec.json

@@ -19,18 +19,6 @@
       "title": "Seed",
       "type": "integer"
     },
-    "suite": {
-      "anyOf": [
-        {
-          "type": "string"
-        },
-        {
-          "type": "null"
-        }
-      ],
-      "default": null,
-      "title": "Suite"
-    },
     "task": {
       "title": "Task",
       "type": "string"
@@ -62,5 +50,5 @@
   ],
   "title": "BenchmarkConditionSpec",
   "type": "object",
-  "x-contract-version": "0.5.10"
+  "x-contract-version": "0.5.12"
 }

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/schemas/BenchmarkRegisterRequest.json

@@ -11,22 +11,6 @@
         "action_space": {
           "$ref": "#/$defs/ActionSpace"
         },
-        "binarize": {
-          "anyOf": [
-            {
-              "additionalProperties": {
-                "type": "number"
-              },
-              "type": "object"
-            },
-            {
-              "type": "null"
-            }
-          ],
-          "default": null,
-          "description": "dimension -> threshold: the client gives the robot 1.0 above it, 0.0 otherwise",
-          "title": "Binarize"
-        },
         "bounds": {
           "anyOf": [
             {
@@ -75,22 +59,6 @@
           "default": null,
           "title": "Max Delta"
         },
-        "on_violation": {
-          "anyOf": [
-            {
-              "enum": [
-                "clip",
-                "refuse"
-              ],
-              "type": "string"
-            },
-            {
-              "type": "null"
-            }
-          ],
-          "default": null,
-          "title": "On Violation"
-        },
         "robot": {
           "title": "Robot",
           "type": "string"
@@ -164,18 +132,6 @@
           "title": "Seed",
           "type": "integer"
         },
-        "suite": {
-          "anyOf": [
-            {
-              "type": "string"
-            },
-            {
-              "type": "null"
-            }
-          ],
-          "default": null,
-          "title": "Suite"
-        },
         "task": {
           "title": "Task",
           "type": "string"
@@ -888,6 +844,14 @@
       "default": null,
       "title": "Memory Gb"
     },
+    "metric_units": {
+      "additionalProperties": {
+        "type": "string"
+      },
+      "maxProperties": 64,
+      "title": "Metric Units",
+      "type": "object"
+    },
     "name": {
       "title": "Name",
       "type": "string"
@@ -954,12 +918,19 @@
       "default": null,
       "title": "Seeds Repeat"
     },
-    "suite_of": {
-      "additionalProperties": {
-        "type": "string"
-      },
-      "title": "Suite Of",
-      "type": "object"
+    "success_metric": {
+      "anyOf": [
+        {
+          "maxLength": 200,
+          "minLength": 1,
+          "type": "string"
+        },
+        {
+          "type": "null"
+        }
+      ],
+      "default": null,
+      "title": "Success Metric"
     },
     "task_descriptions": {
       "additionalProperties": {
@@ -991,5 +962,5 @@
   ],
   "title": "BenchmarkRegisterRequest",
   "type": "object",
-  "x-contract-version": "0.5.10"
+  "x-contract-version": "0.5.12"
 }

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/schemas/BenchmarkSpec.json

@@ -368,7 +368,7 @@
       ],
       "title": "SeqFlags",
       "type": "integer",
-      "x-contract-version": "0.5.10",
+      "x-contract-version": "0.5.12",
       "x-flag-names": {
         "CHECK": 65536,
         "INFER": 16,
@@ -625,5 +625,5 @@
   ],
   "title": "BenchmarkSpec",
   "type": "object",
-  "x-contract-version": "0.5.10"
+  "x-contract-version": "0.5.12"
 }

{sequence_base-0.4.11 → sequence_base-0.4.13}/sequence_base/schemas/BenchmarkStatusResponse.json

@@ -40,5 +40,5 @@
   ],
   "title": "BenchmarkStatusResponse",
   "type": "object",
-  "x-contract-version": "0.5.10"
+  "x-contract-version": "0.5.12"
 }