Set Up Conditional Loop AI Copy Refinement Workflows

Overview

This documentation details configuring a FastGPT Conditional Loop node to automate iterative AI copy refinement until a predefined quality threshold is met. The workflow leverages per-iteration scoring to evaluate revised copy, eliminating manual rework by automatically initiating additional refinement rounds until the output meets required standards.

Step-by-Step Implementation

Follow these concrete, source-aligned steps to build the copy refinement loop:

1. Set Loop Type: Select Conditional Loop as the loop type for the target Loop node.
2. Configure Loop Body Sub-Workflow:
   • Deploy an AI Chat node labeled Copy Refinement to accept initial draft inputs.
   • Deploy a second AI Chat node labeled Evaluation to generate a numeric quality score for the refined copy.
   • Deploy a Condition node with two distinct routing paths:
     • Passing score: Route to an Assigned Reply node to format the final copy for user output, then connect this node to a Loop Break node to terminate the loop immediately.
     • Failing score: Do not establish any downstream connections; the loop will automatically initiate a new iteration using the polished draft generated in the current round.
3. Configure Loop Outputs:
   • Navigate to the Loop node’s Outputs configuration panel, then add a custom output variable named final_text.
   • Reference the reply output of the Copy Refinement AI Chat node for this variable. All downstream workflow nodes can use the final_text variable to retrieve the final approved copy once the loop exits.

Execution Flow Details

Post-execution, full granular details of each iteration are accessible in the "Complete Response" panel, as referenced in the included workflow execution screenshot. The standard execution sequence follows two core phases:

1. First Iteration: The loop runs Loop Start → Copy Refinement → Evaluation → Condition. If the generated score fails to meet requirements, no break trigger occurs, and the system automatically starts a new iteration with the refined draft from the first round.
2. Second Iteration: The sub-workflow runs again using the updated draft. If the score meets the predefined threshold, the workflow routes to Assigned Reply, triggers the Loop Break node, and the loop terminates safely.

Source: FastGPT official source

Applicability and version scope

Use this page for the documented Workflow nodes 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.