Filter Dataset Collections in FastGPT Searches

Overview

Collection filtering in FastGPT dataset searches lets you narrow returned results to specific subsets of your uploaded collections. By configuring the Collection Filter field, you can target collections based on their assigned tags and creation timestamps, reducing irrelevant results and streamlining search workflows.

Filter Syntax & Parameters

The filter configuration uses a valid JSON object with two top-level supported fields: tags and createTime. The following table outlines each field and their valid operators:

A complete example filter configuration is provided below:

{
  "tags": {
    "$and": ["Tag 1", "Tag 2"],
    "$or": ["When $and tags are present, $and takes effect and $or is ignored"]
  },
  "createTime": {
    "$gte": "YYYY-MM-DD HH:mm format, matches collections created after this time",
    "$lte": "YYYY-MM-DD HH:mm format, matches collections created before this time. Can be used together with $gte"
  }
}

Each operator functions as defined:

• $and: Requires all listed tags to be present on a collection for it to match the filter.
• $or: Requires at least one listed tag to be present, though this condition is ignored if $and is also configured.
• $gte: Returns collections created on or after the specified timestamp.
• $lte: Returns collections created on or before the specified timestamp, and may be used alongside $gte to define a targeted time range.

Key Usage Guidelines

There are two critical rules for proper filter configuration:

1. Tag value formats: Tag values may be either a string representing a tag name, or null. The null value specifically matches collections that have no tags assigned.
2. Conditional priority: If both $and and $or are included in the tags filter object, only the $and condition will be applied, and the $or condition will be ignored entirely.

Source: FastGPT official source

Applicability and version scope

Use this page for the documented Tutorials scenario. Confirm the FastGPT, dependency, API, and deployment versions in the official source before applying a change.

Safety guardrails

Use [REDACTED_CREDENTIAL] for credentials and private data. Confirm the documented environment and version before review.

Rollback guidance

Restore the prior technical-content authority snapshot. Restore saved configuration and data snapshots, then repeat the smallest verification scenario.