王一帆 bd62f96732
feat(knowledge): add optional platform connections and hierarchical spaces (phase 3) (#7417)
* feat(knowledge): add optional platform connections and hierarchical spaces

* fix(knowledge): return actionable errors for unavailable policy storage

* perf(knowledge): resolve space policy once per snapshot
2026-10-25 19:52:31 +08:00

89 lines
3.3 KiB
Python

"""Extension-owned policy; executable configuration never comes from HTTP."""
import hashlib
import json
from typing import Any, Literal
from deerflow_extension_knowledge_management.contracts import Model, Operation, ProcessingOptions
from pydantic import Field, field_validator, model_validator
OPERATIONS = frozenset({"capabilities", "status", "upload", "url", "manual", "reprocess", "cancel"})
TOOLS = frozenset({"knowledge_search", "knowledge_entries_search", "entity_relations_search"})
Role = Literal["blocked", "viewer", "editor", "manager"]
RANK = {"blocked": 0, "viewer": 1, "editor": 2, "manager": 3}
Tool = Literal["knowledge_search", "knowledge_entries_search", "entity_relations_search"]
def resource_key(provider: str, connection: str, base: str) -> str:
return "kp1_" + hashlib.sha256(json.dumps([provider, connection, base], ensure_ascii=False, separators=(",", ":")).encode("utf-8")).hexdigest()
class Resource(Model):
base_id: str = Field(min_length=1, max_length=256)
dataset_id: str = Field(min_length=1, max_length=256)
operations: frozenset[Operation] = OPERATIONS
tools: frozenset[Tool] = TOOLS
class Connection(Model):
id: str = Field(pattern=r"^[A-Za-z0-9_-]{1,64}$")
provider: str = Field(pattern=r"^[a-z][a-z0-9_-]{0,63}$")
provider_factory: str = "deerflow_extension_knowledge_management.weknora:WeKnoraManagement"
provider_config: dict[str, Any] = Field(default_factory=dict, repr=False)
resources: tuple[Resource, ...] = Field(min_length=1, max_length=200)
@model_validator(mode="after")
def unique(self):
if len({r.base_id for r in self.resources}) != len(self.resources):
raise ValueError("Duplicate native base identity")
return self
class Settings(Model):
enabled: bool = Field(default=False, strict=True)
storage_path: str
connections: tuple[Connection, ...] = Field(min_length=1, max_length=20)
@model_validator(mode="after")
def unique(self):
if len({c.id for c in self.connections}) != len(self.connections):
raise ValueError("Duplicate connection identity")
return self
class Policy(Model):
enabled: bool = Field(default=True, strict=True)
operations: frozenset[Operation] | None = None
tools: frozenset[Tool] | None = None
default_tool: Tool | None = None
processing: ProcessingOptions = Field(default_factory=ProcessingOptions)
class ConnectionPolicy(Policy):
title: str = Field(default="", max_length=200)
class Space(Policy):
title: str = Field(min_length=1, max_length=200)
parent_id: str | None = Field(default=None, pattern=r"^[A-Za-z0-9_-]{1,64}$")
resources: frozenset[str] | None = None
members: dict[str, Role] = Field(default_factory=dict, max_length=200)
@field_validator("members")
@classmethod
def exact_users(cls, members):
if any(not user or user != user.strip() or len(user) > 256 for user in members):
raise ValueError("Use exact bounded host user identities")
return members
class PlatformPolicy(Model):
defaults: Policy = Field(default_factory=Policy)
connections: dict[str, ConnectionPolicy] = Field(default_factory=dict, max_length=20)
spaces: dict[str, Space] = Field(default_factory=dict, max_length=500)
class Update(Model):
revision: int = Field(ge=0, strict=True)
policy: PlatformPolicy