CSV

File

The CSV connector is a file connector that imports claims from local CSV or TSV files. It uses Python’s built-in csv.DictReader with zero external dependencies.

How it works

Maps CSV columns to claim fields via a configurable mapping dictionary
The CSV file must have a header row — column names are used for field resolution
Supports TSV files by setting delimiter="\t"
Confidence can come from a mapped column or a default value; the column takes precedence
Source ID format: csv:{path}:row{num}
Zero external dependencies — uses only Python standard library

Mapping fields

Field	Required	Description
`subject`	Yes	Column name for the claim subject
`predicate`	Yes	Column name for the claim predicate
`object`	Yes	Column name for the claim object
`subject_type`	No	Static value for subject entity type
`predicate_type`	No	Static value for predicate type
`object_type`	No	Static value for object entity type
`confidence`	No	Column name for per-row confidence score

1 No external dependencies required

The CSV connector uses only Python’s standard library csv module. No additional packages need to be installed.

2 Prepare your CSV file

Ensure your CSV file has a header row. Identify which columns contain the subject, predicate, and object for each claim.

gene,relation,target
BRCA1,inhibits,TP53
EGFR,activates,MAPK1

3 Connect

conn = db.connect("csv",
    path="data/interactions.csv",
    mapping={"subject": "gene", "predicate": "relation", "object": "target"},
)
result = conn.run(db)

Parameter	Required	Default	Description
`path`	Yes	—	Path to the CSV or TSV file
`mapping`	Yes	—	Column-to-claim field mapping dictionary
`delimiter`	No	`,`	Field delimiter. Use `\t` for TSV files.
`confidence`	No	`None`	Default confidence for all rows
`source_type`	No	`csv_import`	Provenance source type
`save`	No	`False`	N/A (no credentials to save)

Basic CSV

conn = db.connect("csv",
    path="data/interactions.csv",
    mapping={"subject": "gene", "predicate": "relation", "object": "target"},
)
result = conn.run(db)

TSV with default confidence

conn = db.connect("csv",
    path="data/edges.tsv",
    mapping={"subject": "src", "predicate": "rel", "object": "dst"},
    delimiter="\t",
    confidence=0.8,
)
result = conn.run(db)

With confidence column and types

conn = db.connect("csv",
    path="data/knowledge.csv",
    mapping={
        "subject": "entity_a",
        "predicate": "relationship",
        "object": "entity_b",
        "confidence": "score",
        "subject_type": "gene",
        "object_type": "disease",
    },
)
result = conn.run(db)

CSV