Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.
Python agent-contextagent-memoryagent-pluginsagentic-ragcontext-databasedsh-pluginknowledge-basememorymemory-managementself-evolvingvector-database
# Copyright (c) 2026 Beijing Volcano Engine Technology Co., Ltd.
# SPDX-License-Identifier: AGPL-3.0
from pathlib import Path
from typing import Optional
from pydantic import BaseModel, Field, model_validator
class ParseOutputConfig(BaseModel):
"""Where parsers write intermediate artifacts.
``agfs`` (default) writes to the shared AGFS ``viking://temp`` space. ``local``
keeps parser intermediates on the current worker; add-resources commits the
required bytes to the formal AGFS tree before enqueueing downstream work.
"""
mode: str = Field(
default="agfs",
description="Parse artifact backend: 'agfs' (shared temp) or 'local' (local dir).",
)
local_root: Optional[str] = Field(
default=None,
description=(
"Directory root for the 'local' backend. Defaults to a dedicated subdir "
"under the system temp dir when unset."
),
)
model_config = {"extra": "forbid"}
@model_validator(mode="after")
def validate_mode(self):
if self.mode not in {"agfs", "local"}:
raise ValueError("storage.parse_output.mode must be 'agfs' or 'local'")
return self
def resolved_local_root(self) -> str:
"""Return the local root, falling back to a dedicated system-temp subdir."""
if self.local_root:
return str(Path(self.local_root).expanduser())
import tempfile
return str(Path(tempfile.gettempdir()) / "openviking-parse")