Journal-entry testing

client.search.je_testing runs journal-entry testing over a workspace’s ledger package and returns a Job whose result is a JetResult. A sibling of client.search.deep, not a preset of it. Deep search retrieves and cites; this performs a test: the run orients itself in the ingested ledger tables, writes SQL over them itself, and answers with sql_receipts — every statement it executed — so a figure and the query behind it travel together.

from ndi_sdk import NdiClient

with NdiClient() as client:
    job = client.search.je_testing(
        workspace_id,
        query="test whether any posted journal line was approved by its own preparer",
        paths=["ledger/journal.xlsx"],
    )
    job = client.jobs.wait(job.job_id)
    result = job.result
    print(result.answer)
    for figure in result.figures:
        print(figure.criterion, figure.count, "of", figure.population)
        print("  readable out of", figure.exception_listing_result_id)
    for ground_id in result.grounding_job_ids or []:
        grounded = client.jobs.wait(ground_id, timeout=300)
        print(grounded.result)

Method signature

def je_testing(
    workspace_id: UUID | str,
    *,
    query: str,
    context: str | None = None,
    reasoning_effort: "none" | "minimal" | "low" | "medium" | "high" | None = None,
    path_prefix: str | None = None,
    paths: list[str] | None = None,
    session_id: UUID | None = None,
    wait_seconds: int = 0,
    idempotency_key: str | None = None,
) -> Job

Parameters

Parameter	Type	Required	Description
`workspace_id`	`UUID` or `str`	Yes	Workspace that already holds an ingested ledger package
`query`	`str`	Yes	The procedure to perform, in your own words
`context`	`str`	No	Background the run should treat as given
`reasoning_effort`	`"none"` / `"minimal"` / `"low"` / `"medium"` / `"high"`	No	Thinking budget; omitted uses the service default
`path_prefix`	`str`	No	Limit the run to files under this prefix
`paths`	`list[str]`	No	The files of the package under test (max 64). Intersects with `path_prefix`
`session_id`	`UUID`	No	Continue an earlier run’s thread
`wait_seconds`	`int`	No	Hold the submission request open (max 300)
`idempotency_key`	`str`	No	Override the key the SDK normally mints

Returns

A Job whose result is a JetResult (result_type is je_testing).

Reasoning budget

reasoning_effort changes how much the fixed JET engine thinks. Omit it to use the service default, or select a larger budget for a more thorough run. This does not change the model. JET no longer accepts effort or tier; new JET jobs have job.effort = null. There is no top_k (the run authors its own citation list) and no allow_clarification (a run that cannot proceed says so in its answer).

Reading the result

Field	What it is
`answer`	The prose answer. `None` only on a run that exhausted its budget first
`figures`	One entry per criterion: `definition`, `population`, `count`, `value`, and the `result_ids` behind it
`sql_receipts`	Every statement executed, with its table bindings and a bounded preview
`binding`	The definitional layer the run committed to — period column, posted encoding, clock
`discrepancies`	Where the data contradicted the request’s premise. Empty is an answer, not an omission
`population_statement`	The population every figure was computed over, in the run’s own words
`quality`	How well a performed run went: fail-open checks, tool failures, missing artifacts
`execution_status`	Whether the run tested what it was asked to test
`exhausted`	Budget ran out before finishing — a partial answer, not a failure
`degraded`	The run broke before finalizing an answer; the working is published, `figures` is empty
`grounding_job_ids`	Ground jobs planned for this answer’s evidences. Poll each with `jobs.wait` after this result

A figure’s cannot_be_performed means the workspace holds no material for that criterion; result_ids then name the probes that establish the absence.

result = job.result
if result.degraded:
    raise RuntimeError(result.degraded)

for receipt in result.sql_receipts:
    print(receipt.result_id, receipt.sql)
    # rows is the bounded preview; result_row_count is the stored result's size
    print(receipt.result_row_count, "rows", "(truncated)" if receipt.truncated else "")

Follow-ups

Pass an earlier result’s session_id to ask a follow-up in the same thread. Only the API key that started a thread may continue it, and a thread whose turn is still running returns 409 session_busy. See Search methods and the je-testing API reference.

Get started

Core methods

Async and utilities

Method signature

Parameters

Returns

Reasoning budget

Reading the result

Follow-ups

​Method signature

​Parameters

​Returns

​Reasoning budget

​Reading the result

​Follow-ups

Method signature

Parameters

Returns

Reasoning budget

Reading the result

Follow-ups