Konnektr Graph SDK for Python

A powerful, Python SDK for Konnektr Graph, fully compatible with the Azure Digital Twins API but optimized for the Konnektr ecosystem.

Features

• Azure-Free: No dependencies on Azure libraries.
• Synchronous & Asynchronous: High-performance clients for both threaded and async workflows.
• Modular Auth: Supports OAuth 2.0 Client Credentials, Device Code Flow, and Static Tokens.
• Auto-Pagination: Seamlessly iterate through large query results and resource lists.
• Data Models: Typed dataclasses for Digital Twins, Models, Relationships, and Jobs.

Installation

pip install konnektr-graph

Quick Start

Synchronous Client

from konnektr_graph import KonnektrGraphClient
from konnektr_graph.auth import ClientSecretCredential

# Authenticate
cred = ClientSecretCredential(
    domain="auth.konnektr.io",
    audience="https://graph.konnektr.io",
    client_id="YOUR_CLIENT_ID",
    client_secret="YOUR_CLIENT_SECRET"
)

# Initialize Client
client = KonnektrGraphClient("https://your-graph-endpoint.konnektr.io", cred)

# Get a Digital Twin
twin = client.get_digital_twin("my-twin-id")
print(twin)

# Query Twins with auto-pagination (Cypher)
query = "MATCH (t:Twin) RETURN t"
for twin in client.query_twins(query):
    print(twin)

# Query with parameters (Cypher + `$name` placeholders, forwarded to the
# server's `parameters` field; only Cypher supports query parameters)
params = {"model": "dtmi:com:example:Room;1", "minTemp": 20}
query = "MATCH (t:Twin) WHERE t.`$metadata`.`$model` = $model AND temperature > $minTemp RETURN t"
for twin in client.query_twins(query, query_parameters=params):
    print(twin)

Asynchronous Client

import asyncio
from konnektr_graph.aio import KonnektrGraphClient
from konnektr_graph.auth import AsyncClientSecretCredential

async def main():
    cred = AsyncClientSecretCredential(
        domain="auth.konnektr.io",
        audience="https://graph.konnektr.io",
        client_id="...",
        client_secret="..."
    )

    async with KonnektrGraphClient("https://your-graph-endpoint.konnektr.io", cred) as client:
        twin = await client.get_digital_twin("my-twin-id")
        print(twin)

asyncio.run(main())

Scoped Memory Search

# Environment/user/privacy scoping is enforced SERVER-SIDE: scope predicates
# (model allow-list, property equality filters, related-twin predicate) are
# applied in the database *before* nearest-neighbour ranking and LIMIT, so the
# ranking never sees out-of-scope records. Never filter broad search results
# locally for authorization — use this dedicated endpoint instead.
results = client.search_memory(
    vector=[0.11, -0.02, 0.58],
    limit=5,
    model_ids=["dtmi:example:MemoryRecord;1"],
    property_filters={"environmentId": "env-1", "userId": "user-9"},
    related_twin_id="session-9",
)
for record in results:
    print(record.id, record.distance, record.excerpt)

# Probe pgvector availability, and create the HNSW index (idempotent) once:
capability = client.get_memory_search_capability()
if capability.vector_search_available:
    client.ensure_memory_search_index(dimension=3)

The async client (konnektr_graph.aio.KonnektrGraphClient) exposes the same three methods (search_memory, ensure_memory_search_index, get_memory_search_capability) with identical signatures and typed results (MemorySearchResult, MemorySearchIndex, MemorySearchCapability).

Error mapping: authorization failures raise AuthenticationError (401/403), server-side validation failures raise ValidationError (400), a missing pgvector extension raises ServiceUnavailableError (503), and transport failures raise HttpResponseError.

Authentication Options

• ClientSecretCredential / AsyncClientSecretCredential: Ideal for server-to-server scenarios.
• DeviceCodeCredential / AsyncDeviceCodeCredential: Best for interactive CLI tools.
• StaticTokenCredential: Use when you already have a valid access token.
• DefaultAzureCredentialAdapter / AsyncDefaultAzureCredentialAdapter: Use Azure Identity credentials through the SDK auth protocol.

Using Azure Identity (DefaultAzureCredential)

If your backend validates Azure AD tokens (e.g., ADT-compatible audience https://digitaltwins.azure.net/), you can adapt DefaultAzureCredential to this SDK's TokenProvider interface.

Install Azure Identity in your app:

pip install "konnektr-graph[azure]"

Or install azure-identity directly if you prefer:

pip install azure-identity

Sync

from azure.identity import DefaultAzureCredential
from konnektr_graph import KonnektrGraphClient
from konnektr_graph.auth import DefaultAzureCredentialAdapter

endpoint = "https://your-graph-endpoint"

azure_cred = DefaultAzureCredential()
cred = DefaultAzureCredentialAdapter(
    azure_cred,
    scope="https://digitaltwins.azure.net/.default",
)

client = KonnektrGraphClient(endpoint, cred)
twin = client.get_digital_twin("my-twin-id")
print(twin)

Async

import asyncio
from azure.identity.aio import DefaultAzureCredential
from konnektr_graph.aio import KonnektrGraphClient
from konnektr_graph.auth import AsyncDefaultAzureCredentialAdapter

async def main():
    endpoint = "https://your-graph-endpoint"

    azure_cred = DefaultAzureCredential()
    cred = AsyncDefaultAzureCredentialAdapter(
        azure_cred,
        scope="https://digitaltwins.azure.net/.default",
    )

    async with KonnektrGraphClient(endpoint, cred) as client:
        twin = await client.get_digital_twin("my-twin-id")
        print(twin)

    await azure_cred.close()

asyncio.run(main())

License

This project is licensed under the MIT License - see the LICENSE file for details.