Class BaseKnowledge

quantmind/knowledge/_base.py:66–132

Root of every quantmind knowledge type. All three shapes (Flatten / Tree / Graph) share this field set; subclasses add domain-specific payload. Subclasses MUST override `embedding_text()` to declare what string the store layer should embed.

    run_id: UUID | None = None
    extracted_at: datetime


class ArtifactMeta(BaseModel):
    """Minimal provenance for a derived, self-contained artifact.

    A derived artifact (for example a page-preserving structure tree) is not
    canonical ``BaseKnowledge``: it is rebuildable from its source. It still
    needs the light provenance carried here to be stored and time-queried on its
    own — an information cutoff (``as_of``), a typed light source reference, and
    the exact content hash of the source revision it was derived from. Kept
    deliberately small so several artifact shapes can mix it in without
    inheriting the full canonical-knowledge field set.

    This provenance is metadata, never identity: it must stay out of an
    artifact's ``id`` and ``content_hash`` so a rebuild at a different wall-clock
    time yields the identical artifact.
    """

    model_config = ConfigDict(extra="forbid", frozen=True)

    as_of: datetime = Field(..., description="Information cutoff time.")
    source: SourceRef
    source_title: str | None = None
    source_content_hash: str = Field(pattern=r"^[0-9a-f]{64}$")


class BaseKnowledge(BaseModel):
    """Root of every quantmind knowledge type.

    All three shapes (Flatten / Tree / Graph) share this field set; subclasses
    add domain-specific payload. Retrieval-specific text selection belongs to
    the library indexing boundary.
    """

    model_config = ConfigDict(extra="forbid", frozen=True)

    # Identity
    id: UUID = Field(default_factory=uuid4)
    item_type: str
    schema_version: str = "1.0"

    # Time (financial mandate)
    as_of: datetime = Field(..., description="Information cutoff time.")
    available_at: datetime | None = Field(
        default=None,
        description="Time the source became observable to researchers.",
    )
    created_at: datetime = Field(
        default_factory=lambda: datetime.now(timezone.utc)
    )

    # Provenance (no bare strings)
    source: SourceRef
    extraction: ExtractionRef | None = None

    # Trust
    confidence: Literal["low", "medium", "high"] = "medium"