dxrpy 0.4.0

A client library for DXR

Data X-Ray Python Library

Unofficial Python library for the Data X-Ray API.

Warning This library is unofficial and unsupported and may change at any time. Not for production use.

Installation

pip install dxrpy

Usage

Initializing the Client

from dxrpy import DXRClient

client = DXRClient(
    api_url="https://your-dxr-instance.example.com",
    api_key="your_api_key",
)

Credentials can also be loaded from environment variables (or a .env file):

DXR_BASE_URL=https://your-dxr-instance.example.com
DXR_API_KEY=your_api_key

---

On-Demand Classifier

Run files through the on-demand classification pipeline.

from dxrpy.utils import File

files = [File("path/to/file1.txt"), File("path/to/file2.txt")]
hits = client.on_demand_classifier.run_job(files, datasource_ids=[123])

for hit in hits:
    print(hit.labels)

---

Datasources

Create and manage datasources programmatically.

# List all datasources
datasources = client.datasources.list()

# Create a new datasource
ds = client.datasources.create(
    name="benchmark-experiment-1",
    connector_type_id=7,          # on-demand / file upload connector
    settings_profile_id=50217,
)
print(ds.id)

# Find by name
ds = client.datasources.find_by_name("benchmark-experiment-1")

# Update
client.datasources.update(ds.id, name="benchmark-experiment-1-v2")

# Delete
client.datasources.delete(ds.id)

---

Smart Labels

Smart labels (tags) automatically tag documents matching a query condition. They are defined globally and reference one or more datasources.

from dxrpy.index.json_search_query import JsonSearchQueryItem

# Create a label that fires whenever an SSN annotation exists
label = client.smart_labels.create(
    name="Has-SSN",
    datasource_ids=[50148],
    query_items=[
        JsonSearchQueryItem(
            parameter="annotators",
            value="annotation.42",
            type="text",
            match_strategy="exists",
        ),
    ],
    color="#FF5733",
)

# List / find
labels = client.smart_labels.list()
label = client.smart_labels.find_by_name("Has-SSN")

# Update conditions
client.smart_labels.update(label.id, query_items=[...])

# Delete
client.smart_labels.delete(label.id)

---

Extractors

LLM metadata extractors define a prompt and optional target data types.

extractor = client.extractors.create(
    name="PII extractor",
    prompt="Extract all personal information from the text. Return a JSON object with keys: name, email, phone.",
    data_types=["PERSON", "EMAIL_ADDRESS", "PHONE_NUMBER"],
)
print(extractor.id)

# CRUD
extractors = client.extractors.list()
extractor = client.extractors.find_by_name("PII extractor")
client.extractors.update(extractor.id, prompt="Updated prompt...")
client.extractors.delete(extractor.id)

---

Settings Profiles & Extraction Workflows

Settings profiles bundle classification configuration. Use set_extraction_workflow to attach an extractor (unconditionally or with conditions) to a profile.

from dxrpy import WorkflowStep
from dxrpy.index.json_search_query import JsonSearchQueryItem

# Extractor that always runs
client.settings_profiles.set_extraction_workflow(
    profile_id=50217,
    enabled=True,
    steps=[WorkflowStep(extractor_id=594)],
)

# Extractor that only runs when an SSN annotation already exists
client.settings_profiles.set_extraction_workflow(
    profile_id=50217,
    enabled=True,
    steps=[
        WorkflowStep(
            extractor_id=594,
            condition=[
                JsonSearchQueryItem(
                    parameter="annotators",
                    value="annotation.16",
                    type="text",
                    match_strategy="exists",
                )
            ],
        )
    ],
)

# Disable extraction
client.settings_profiles.set_extraction_workflow(
    profile_id=50217,
    enabled=False,
)

---

Searching the Index

from dxrpy.index.json_search_query import JsonSearchQuery, JsonSearchQueryItem

query = JsonSearchQuery(
    datasource_ids=["123"],
    query_items=[
        JsonSearchQueryItem(parameter="match_all", value={}, type="match")
    ],
)
results = client.index.search(query)

print(f"Total hits: {results.total_hits}")
for hit in results.hits:
    print(hit.file_name, hit.labels)

---

Changelog

See CHANGELOG.md.