sequence-base 0.4.9__tar.gz → 0.4.11__tar.gz

{sequence_base-0.4.9 → sequence_base-0.4.11}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: sequence-base
-Version: 0.4.9
+Version: 0.4.11
 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-0.4.9 → sequence_base-0.4.11}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "sequence-base"
-version = "0.4.9"
+version = "0.4.11"
 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-0.4.9 → sequence_base-0.4.11}/sequence_base/__init__.py

@@ -105,6 +105,10 @@ from sequence_base.contract import (
 from sequence_base.constants.version import CONTRACT_VERSION
 
 
+
+from sequence_base.constants.limits import MAX_ENVS, MAX_USD, TASK_DESCRIPTION_MAX
+
+
 from sequence_base.obs.log import Timer, enabled, log
 from sequence_base.obs.rid import HEADER, clean, clean_text, new_rid, unescape_text
 
@@ -152,6 +156,7 @@ __all__ = [
     "TEMPLATE_SCHEME", "parse_template",
 
     "ErrorCode", "ErrorBody", "WarmingError", "ERRORS", "ErrorSpec", "code_for_status", "CONTRACT_VERSION",
+    "MAX_ENVS", "MAX_USD", "TASK_DESCRIPTION_MAX",
 
     "HEADER", "new_rid", "clean", "clean_text", "unescape_text", "log", "Timer", "enabled",
 

sequence_base-0.4.11/sequence_base/constants/limits.py

@@ -0,0 +1,16 @@
+"""Numeric limits the contract shares with the gateway, defined once here so one bound is enforced everywhere
+rather than drifting between `EvalRequest`, the benchmark register body and the gateway's own routes.
+"""
+
+
+
+MAX_ENVS = 65536
+
+
+
+MAX_USD = 100_000
+
+
+
+TASK_DESCRIPTION_MAX = 300
+

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/constants/version.py

@@ -4,7 +4,8 @@
 
 
 
-CONTRACT_VERSION = "0.5.8"
+CONTRACT_VERSION = "0.5.10"
+
 
 
 

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/contract/benchmarks.py

@@ -16,6 +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.contract.accounts import SecretRef
 from sequence_base.contract.deployments import CodeBundle, DeploymentState, InferParam
 from sequence_base.contract.flags import BENCHMARK_BATCHED, BENCHMARK_BINDABLE, SeqFlags
@@ -125,7 +126,7 @@ class BenchmarkSpec(BaseModel):
     bindings: dict[SeqFlags, str] = Field(..., description="SETUP / RESET / CHECK / SCORE (+ OBSERVE / STEP) → method name")
     gpu: str | list[str] | None = None
 
-    envs: int | None = Field(default=None, ge=1, le=65536)
+    envs: int | None = Field(default=None, ge=1, le=MAX_ENVS)
     assets: list[VolumeRef] = Field(default_factory=list)
     render: RenderSpec | None = None
 
@@ -221,9 +222,10 @@ class BenchmarkRegisterRequest(BaseModel):
     default_adapter_ref: str | None = None
 
 
+
     gpu: str | None = Field(default=None, max_length=32)
 
-    envs: int | None = Field(default=None, ge=1, le=65536)
+    envs: int | None = Field(default=None, ge=1, le=MAX_ENVS)
 
 
 
@@ -253,8 +255,8 @@ class BenchmarkRegisterRequest(BaseModel):
         for task, text in self.task_descriptions.items():
             if task not in declared:
                 raise ValueError(f"task_descriptions names {task!r}, which is not a declared task")
-            if len(text) > 300:
-                raise ValueError(f"task_descriptions[{task!r}] is {len(text)} characters: at most 300")
+            if len(text) > TASK_DESCRIPTION_MAX:
+                raise ValueError(f"task_descriptions[{task!r}] is {len(text)} characters: at most {TASK_DESCRIPTION_MAX}")
         return self
 
 

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/contract/deployments.py

@@ -60,7 +60,12 @@ class Resources(BaseModel):
     memory_limit_gb: float | None = Field(default=None, ge=0.125, le=1024)
 
 
-    region: str | None = Field(default=None, max_length=32)
+
+
+
+    region: str | None = Field(default=None, max_length=32, json_schema_extra={"deprecated": True},
+                               description="Not accepted: a deployment's workers run wherever there are cards, so a "
+                                           "region is refused. Omit it (an older SDK still sends it).")
 
     @model_validator(mode="after")
     def _limits_cover_requests(self) -> "Resources":
@@ -185,6 +190,8 @@ class DeploymentPolicy(BaseModel):
     handler_ref: str
     action: ActionContract | None = None
     streaming: bool = False
+
+
     frames_per_chunk: int | None = None
     secrets: list[SecretRef] = Field(default_factory=list)
 

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/contract/eval.py

@@ -10,9 +10,11 @@ not a bare dict or free text.
 from __future__ import annotations
 
 from enum import Enum
-from typing import Literal
+from typing import Annotated, Literal
 
-from pydantic import BaseModel, ConfigDict, Field, model_validator
+from pydantic import BaseModel, ConfigDict, Field, field_validator, model_validator
+
+from sequence_base.constants.limits import MAX_ENVS, MAX_USD
 
 __all__ = [
     "EvalRequest",
@@ -39,32 +41,57 @@ __all__ = [
 ]
 
 
+
+
+_Trials = Annotated[int, Field(ge=1, le=100000)]
+
+
 class EvalRequest(BaseModel):
-    """`model` and `benchmark` required; the rest are optional knobs. `dry_run` selects the estimate-only
-    response shape below."""
+    """`POST /v1/eval`. A real eval names `model` + `benchmark`; a rerun (`rerun_of` / `rerun_unsuccessful_of`) may
+    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,
+    never silently dropped. The gateway inherits this model rather than keeping its own."""
+
+
 
     model_config = ConfigDict(extra="forbid")
 
-    model: str
-    benchmark: str
 
+    model: str | None = Field(default=None, max_length=200)
+    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: int | dict[str, int] | None = None
+    trials: _Trials | dict[str, _Trials] | None = None
     seeds: list[int] | dict[str, list[int]] | None = None
 
 
 
 
-    policy_instances: int = Field(default=1, ge=1, le=64)
+    policy_instances: int | None = Field(default=None, ge=1, le=64)
     policy_gpu: str | None = Field(default=None, max_length=32)
-    benchmark_instances: int = Field(default=1, ge=1, le=64)
-    benchmark_gpu: str | None = Field(default=None, max_length=32)
-    envs: int | None = Field(default=None, ge=1, le=100000)
-    vlm_max_concurrency: int | None = None
-    adapter: str | None = None
+    benchmark_instances: int | None = Field(default=None, ge=1, le=64)
+
+
+
+    benchmark_gpu: str | None = Field(default=None, max_length=32, json_schema_extra={"deprecated": True},
+                                      description="Not accepted: an eval runs on the card its benchmark was registered "
+                                                  "on, so a card is refused. Omit it (an older SDK still sends it).")
+    envs: int | None = Field(default=None, ge=1)
+
+    gpus: int = Field(default=1, ge=1, le=64, json_schema_extra={"deprecated": True},
+                      description="Deprecated name (sequence-ai <= 0.28) for policy_instances; use policy_instances.")
+    envs_per_gpu: int | None = Field(default=None, ge=1, le=1024, json_schema_extra={"deprecated": True},
+                                     description="Deprecated name (sequence-ai <= 0.28) for envs; use envs.")
+    adapter: str | None = Field(default=None, max_length=200)
 
 
     compare: str | list[str] | None = Field(default=None, description="Other model ids to compare against")
@@ -73,9 +100,15 @@ class EvalRequest(BaseModel):
 
     tasks: list[str] | None = Field(default=None, max_length=1000)
 
-    max_usd: float | None = Field(default=None, gt=0)
 
-    rerun_of: str | None = None
+    max_usd: float | None = Field(default=None, gt=0, le=MAX_USD)
+
+    rerun_of: str | None = Field(default=None, max_length=64)
+
+
+
+    rerun_unsuccessful_of: str | None = Field(default=None, max_length=64)
+    metric: str | None = Field(default=None, max_length=64)
 
 
 
@@ -89,6 +122,21 @@ class EvalRequest(BaseModel):
 
     replay: Literal["3d", "none"] | None = None
 
+
+    region: str | None = Field(default=None, max_length=32, json_schema_extra={"deprecated": True},
+                               description="Not accepted: an eval's machines run wherever there are cards, so a region "
+                                           "is refused. Omit it (an older SDK still sends it).")
+
+    @field_validator("envs")
+    @classmethod
+    def _envs_cap(cls, v: int | None) -> int | None:
+        """One benchmark instance runs at most MAX_ENVS environments; keep the author-facing hint (run more by adding
+        instances) rather than a bare bound, and name the same shared constant as the gateway."""
+        if v is not None and v > MAX_ENVS:
+            raise ValueError(f"a benchmark instance runs at most {MAX_ENVS} environments at once; add more "
+                             "benchmark_instances to run more")
+        return v
+
     @model_validator(mode="after")
     def _selection(self) -> "EvalRequest":
         """The selection's own bounds, as the platform holds them: a count 1-100000, at most 5000 seeds in a list and
@@ -382,11 +430,16 @@ class EvalReport(BaseModel):
         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
-    stop_reason: str | None = None
+
+
+
+
+
+    stop_reason: Literal["canceled", "budget", "balance", "deadline", "mismatch", "refused", "queue_timeout",
+                         "no_gpu_render"] | None = None
 
 
     arm_params: list[ArmParams] | None = None

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/contract/policy.py

@@ -261,6 +261,8 @@ class PolicySpec(BaseModel):
     queue_timeout_s: int = Field(default=120, ge=5, le=3600)
     concurrency: Concurrency | None = None
     streaming: bool = False
+
+
     frames_per_chunk: int | None = None
 
     action: ActionContract | None = None

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/ActRequest.json

@@ -79,5 +79,5 @@
   ],
   "title": "ActRequest",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/ActResponse.json

@@ -89,5 +89,5 @@
   ],
   "title": "ActResponse",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/ActionChunk.json

@@ -99,5 +99,5 @@
   ],
   "title": "ActionChunk",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/ActionContract.json

@@ -116,5 +116,5 @@
   ],
   "title": "ActionContract",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/ActionStep.json

@@ -19,5 +19,5 @@
   ],
   "title": "ActionStep",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/AdapterTransform.json

@@ -38,5 +38,5 @@
   ],
   "title": "AdapterTransform",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/AddLocal.json

@@ -23,5 +23,5 @@
   ],
   "title": "AddLocal",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/Autoscaler.json

@@ -69,5 +69,5 @@
   },
   "title": "Autoscaler",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/BenchmarkConditionSpec.json

@@ -62,5 +62,5 @@
   ],
   "title": "BenchmarkConditionSpec",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/BenchmarkRegisterRequest.json

@@ -991,5 +991,5 @@
   ],
   "title": "BenchmarkRegisterRequest",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/BenchmarkSpec.json

@@ -368,7 +368,7 @@
       ],
       "title": "SeqFlags",
       "type": "integer",
-      "x-contract-version": "0.5.8",
+      "x-contract-version": "0.5.10",
       "x-flag-names": {
         "CHECK": 65536,
         "INFER": 16,
@@ -625,5 +625,5 @@
   ],
   "title": "BenchmarkSpec",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/BenchmarkStatusResponse.json

@@ -40,5 +40,5 @@
   ],
   "title": "BenchmarkStatusResponse",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/BootstrapResponse.json

@@ -79,5 +79,5 @@
   ],
   "title": "BootstrapResponse",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/Camera.json

@@ -72,5 +72,5 @@
   ],
   "title": "Camera",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/ClampInfo.json

@@ -23,5 +23,5 @@
   ],
   "title": "ClampInfo",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/CodeBundle.json

@@ -23,5 +23,5 @@
   ],
   "title": "CodeBundle",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/Concurrency.json

@@ -96,5 +96,5 @@
   },
   "title": "Concurrency",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/Conditions.json

@@ -55,5 +55,5 @@
   ],
   "title": "Conditions",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/ConnectRequest.json

@@ -25,5 +25,5 @@
   ],
   "title": "ConnectRequest",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/ConnectResponse.json

@@ -342,5 +342,5 @@
   ],
   "title": "ConnectResponse",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/DeploymentAccepted.json

@@ -18,5 +18,5 @@
   ],
   "title": "DeploymentAccepted",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/DeploymentPolicy.json

@@ -530,5 +530,5 @@
   ],
   "title": "DeploymentPolicy",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/DeploymentRequest.json

@@ -997,6 +997,8 @@
             }
           ],
           "default": null,
+          "deprecated": true,
+          "description": "Not accepted: a deployment's workers run wherever there are cards, so a region is refused. Omit it (an older SDK still sends it).",
           "title": "Region"
         }
       },
@@ -1224,5 +1226,5 @@
   ],
   "title": "DeploymentRequest",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/DeploymentStatusResponse.json

@@ -58,5 +58,5 @@
   ],
   "title": "DeploymentStatusResponse",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/Embodiment.json

@@ -267,5 +267,5 @@
   ],
   "title": "Embodiment",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/ErrorBody.json

@@ -94,5 +94,5 @@
   ],
   "title": "ErrorBody",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/ErrorCodes.json

@@ -22,7 +22,7 @@
   ],
   "title": "ErrorCode",
   "type": "string",
-  "x-contract-version": "0.5.8",
+  "x-contract-version": "0.5.10",
   "x-errors": {
     "budget_exhausted": {
       "exit_code": 6,

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/Estimate.json

@@ -60,5 +60,5 @@
   ],
   "title": "Estimate",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/EvalDryRunResponse.json

@@ -240,5 +240,5 @@
   ],
   "title": "EvalDryRunResponse",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }

{sequence_base-0.4.9 → sequence_base-0.4.11}/sequence_base/schemas/EvalReport.json

@@ -485,6 +485,16 @@
     "stop_reason": {
       "anyOf": [
         {
+          "enum": [
+            "canceled",
+            "budget",
+            "balance",
+            "deadline",
+            "mismatch",
+            "refused",
+            "queue_timeout",
+            "no_gpu_render"
+          ],
           "type": "string"
         },
         {
@@ -523,5 +533,5 @@
   ],
   "title": "EvalReport",
   "type": "object",
-  "x-contract-version": "0.5.8"
+  "x-contract-version": "0.5.10"
 }