SDK

Python SDK

The Python SDK is suited for data pipelines, evaluation harnesses, and backend services.

Client setup

Use one reusable client instance and avoid constructing clients inside hot loops.

Initialize client
from aletheia import AletheiaClient

client = AletheiaClient(
    base_url="http://127.0.0.1:3000",
    api_key="XXX1111AAA",
    timeout_s=10,
)

Batch ingest pattern

For very large batches, parallelize by entity partition to reduce lock contention and preserve ordering semantics within each session.

Batch ingestion
batch = [
    {"entity_id": "user-123", "textual_content": "I moved to LA."},
    {"entity_id": "user-123", "textual_content": "I now prefer tea."},
]

for item in batch:
    client.ingest(**item)

Evaluation usage

  • Keep deterministic seeds for benchmark comparability.
  • Record model and policy versions in run metadata.
  • Persist raw hit lists, not only aggregate metrics.