Implementation:Infiniflow Ragflow Metadata Utils

Knowledge Sources	Infiniflow_Ragflow
Domains	Data_Processing, Search
Last Updated	2026-02-12 06:00 GMT

Overview

Concrete tool for metadata filtering, transformation, and JSON schema generation used in document retrieval provided by the RAGFlow common library.

Description

The metadata_utils module provides functions for filtering document metadata against complex conditions (contains, in, starts with, equality, comparison operators), applying metadata filters with auto/semi-auto/manual resolution modes, deduplicating lists, updating metadata dictionaries, and generating JSON schemas from metadata structures.

Usage

Import these utilities when implementing metadata-based document filtering in retrieval pipelines, when building search queries that incorporate metadata constraints, or when converting metadata definitions to JSON schema for API validation.

Code Reference

Source Location

• Repository: Infiniflow_Ragflow
• File: common/metadata_utils.py
• Lines: 1-344

Signature

def meta_filter(metas: dict, filters: list, logic: str = "and") -> bool:
    """Filter metadata dict against a list of conditions with AND/OR logic."""

async def apply_meta_data_filter(
    meta_data_filter: dict,
    metas: dict,
    question: str,
    chat_mdl=None,
    base_doc_ids: list = None,
    manual_value_resolver=None,
) -> list:
    """Apply metadata filters with auto/semi_auto/manual resolution modes."""

def metadata_schema(metadata: dict) -> dict:
    """Generate JSON schema from metadata structure."""

def turn2jsonschema(obj) -> dict:
    """Convert metadata object to JSON schema format."""

Import

from common.metadata_utils import meta_filter, apply_meta_data_filter, metadata_schema

I/O Contract

Inputs

Name	Type	Required	Description
metas	dict	Yes	Document metadata dictionary to filter
filters	list	Yes	List of filter conditions with field, operator, value
logic	str	No	Logical combinator: "and" or "or" (default: "and")
question	str	Yes	User query for auto-resolution mode
chat_mdl	object	No	Chat model for auto-resolving filter values

Outputs

Name	Type	Description
meta_filter() returns	bool	Whether metadata passes all filter conditions
apply_meta_data_filter() returns	list	List of matching document IDs
metadata_schema() returns	dict	JSON schema representation

Usage Examples

from common.metadata_utils import meta_filter, metadata_schema

# Filter documents by metadata
doc_meta = {"author": "John", "year": 2024, "tags": ["AI", "RAG"]}
filters = [
    {"field": "author", "operator": "=", "value": "John"},
    {"field": "year", "operator": ">", "value": 2020},
]
matches = meta_filter(doc_meta, filters, logic="and")
# Returns True

# Generate JSON schema
schema = metadata_schema({"title": "string", "pages": "integer"})

Related Pages

• Environment:Infiniflow_Ragflow_Python_Runtime