b60439152b
- harness/es.py: append-only event store + проекции (образец — кооп-портал) - harness/levels.py: уровни 1с/1м/1ч/1д, контекст ∝ масштабу - harness/scheduler.py: уровневые потоки, идемпотентные тики, absorbed-стабы - harness/isolated_runner.py + _llm_worker.py: LLM-тик в сабпроцессе (глобальный asyncio-рантайм Strands не переживает 2+ create_harness) - cli.py: init/task/tasks/status/events + requeue (написан самим минутным агентом) - 9router: отдельный ключ sk-simbiot, модель auto (glm-5.3 первым)
83 lines
3.1 KiB
Python
83 lines
3.1 KiB
Python
"""Checkpoint system for durable agent execution.
|
|
|
|
A ``Checkpoint`` is a pause-point marker emitted at agent cycle boundaries.
|
|
It captures the position (which boundary fired) and the cycle index. It does
|
|
**not** capture conversation state — pair with a ``SessionManager`` for
|
|
cross-process state continuity.
|
|
|
|
Positions per ReAct cycle:
|
|
- ``after_model``: model returned tool_use; tools have not run yet.
|
|
- ``after_tools``: tools finished; the next model call has not happened yet.
|
|
|
|
Per-tool granularity within a cycle is the ``ToolExecutor``'s responsibility.
|
|
|
|
Usage (mirrors interrupts):
|
|
- Pause: ``AgentResult`` with ``stop_reason="checkpoint"`` and ``checkpoint`` populated.
|
|
- Resume: pass back ``{"checkpointResume": {"checkpoint": ckpt.to_dict()}}``.
|
|
|
|
Precedence:
|
|
- Interrupt > checkpoint: an interrupt during a checkpointing cycle returns
|
|
``stop_reason="interrupt"`` and skips ``after_tools``.
|
|
- Cancel > checkpoint: a cancel signal at either boundary returns
|
|
``stop_reason="cancelled"``.
|
|
|
|
Notes:
|
|
- Checkpoints are only emitted on tool_use cycles. A turn with no tool calls
|
|
emits no checkpoint; use a ``SessionManager`` for durability of every turn.
|
|
- ``EventLoopMetrics`` resets per invocation; aggregate yourself if needed.
|
|
- ``BeforeInvocationEvent`` / ``AfterInvocationEvent`` fire on every resume,
|
|
same as interrupts.
|
|
"""
|
|
|
|
import logging
|
|
from dataclasses import asdict, dataclass, field
|
|
from typing import Any, Literal
|
|
|
|
from ...types.exceptions import CheckpointException
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
CHECKPOINT_SCHEMA_VERSION = "1.0"
|
|
|
|
CheckpointPosition = Literal["after_model", "after_tools"]
|
|
|
|
|
|
@dataclass(frozen=True)
|
|
class Checkpoint:
|
|
"""Pause-point marker. Treat as opaque — pass back to resume.
|
|
|
|
Attributes:
|
|
position: Which boundary fired (``after_model`` or ``after_tools``).
|
|
cycle_index: ReAct loop cycle (0-based).
|
|
schema_version: Rejects incompatible checkpoints on resume.
|
|
"""
|
|
|
|
position: CheckpointPosition
|
|
cycle_index: int = 0
|
|
schema_version: str = field(init=False, default=CHECKPOINT_SCHEMA_VERSION)
|
|
|
|
def to_dict(self) -> dict[str, Any]:
|
|
"""Serialize for persistence."""
|
|
return asdict(self)
|
|
|
|
@classmethod
|
|
def from_dict(cls, data: dict[str, Any]) -> "Checkpoint":
|
|
"""Reconstruct from a dict produced by to_dict().
|
|
|
|
Args:
|
|
data: Serialized checkpoint data.
|
|
|
|
Raises:
|
|
CheckpointException: If schema_version doesn't match the current version.
|
|
"""
|
|
version = data.get("schema_version", "")
|
|
if version != CHECKPOINT_SCHEMA_VERSION:
|
|
raise CheckpointException(
|
|
f"Checkpoints with schema version {version!r} are not compatible "
|
|
f"with current version {CHECKPOINT_SCHEMA_VERSION}."
|
|
)
|
|
known_keys = {k for k in cls.__dataclass_fields__ if k != "schema_version"}
|
|
unknown_keys = set(data.keys()) - known_keys - {"schema_version"}
|
|
if unknown_keys:
|
|
logger.warning("unknown_keys=<%s> | ignoring unknown fields in checkpoint data", unknown_keys)
|
|
return cls(**{k: v for k, v in data.items() if k in known_keys})
|