Filter

Stream a JSONL file row-by-row through an LLM that decides, based on a natural-language criterion, whether each record should be kept or discarded. Only the matching records are written to the output file.

The model is exposed a matches boolean tool — its return value drives the keep/drop decision per record.

Pre-requisite: Install an AI provider application from Profile > {Organization} > Applications. The provider must support tool/function calling.

Parameters

Provider—REQUIRED

Configured AI application.

Model

Model identifier supporting tool calling. Defaults to the provider's recommended model when left empty.

Mode—REQUIRED

• Batch (default) — submits all records as a single batch job. Cheap, asynchronous, best for large files. - Direct — calls the AI per record synchronously.

Acceptance criteria—REQUIRED

Natural-language condition a record must satisfy to be kept downstream (e.g. "The record is about a technology company"). Records that satisfy it pass; others are dropped.

Input

File—REQUIRED

JSONL file with one record per line.

Output

File

JSONL file containing only the records the model classified as matching.