Nodaro DocsQA Check
Score a text from 0 to 1 with an AI model for content, quality, consistency or safety, and get an approved or rejected decision with a short reason.
The QA Check node asks an AI model to score a text from 0.0 to 1.0 against one check type, such as quality or safety. The text is approved when the score reaches your threshold, and the model explains its decision in one short reason. Use QA Check as an automatic quality gate, for example to catch a weak script or caption before anyone publishes it.
When to use it
- Check that a generated script is complete and well structured before you turn it into audio or video.
- Check captions for clarity and grammar before they are posted.
- Screen a text for harmful or offensive content before it goes public.
- Compare the scores of several drafts to see which prompt writes better text.
Quick start
Add the node
Press Tab on the canvas and choose Automate › Logic & Data › QA Check.
Connect the text
Wire the text to check into the Input input, for example from a Generate Text - Any LLM node, a Generate Script node or a Text node.
Choose the check
Open the settings panel. Choose a Check Type and a Threshold, and a different AI Model if you want one.
Run it
The node shows the score, the decision and the model's reason. The reason also leaves through the Approved or the Rejected output, whichever matches the decision.
Inputs and outputs
| Input or output | Accepts or sends | What it does |
|---|---|---|
| Input | Text | The text to evaluate. Required. |
| Approved | Text | The model's reason, when the score is at or above the threshold. |
| Rejected | Text | The model's reason, when the score is below the threshold. |
On each run, only one of the two outputs carries a value. Both carry the reason, not the checked text. To use the checked text in later steps, connect those steps to the node that wrote it.
Check types
| Check Type | What the model evaluates |
|---|---|
| Content (default) | Completeness, coherence and structure. |
| Quality | Clarity, grammar and professionalism. |
| Consistency | Internal consistency: no contradictions and no logical errors. |
| Safety | Whether the text is safe and appropriate, with no harmful or offensive material. |
Settings
| Setting | What it does |
|---|---|
| AI Model | The model that scores the text. The default is Gemini 3.6 Flash. |
| Reasoning Effort | How much the model reasons before it answers, on models that support it. A level has no price of its own: a higher level usually writes more tokens, and those tokens are charged. |
| Advanced mode | Available on Gemini and Claude models. It unlocks Temperature, Max Tokens and the full reasoning range, which appear once it is on. The check then runs on the maker's own service and is charged at that service's rate. The estimate on the node updates at once. On other models the switch is disabled and says why. |
| Check Type | What to evaluate: Content, Quality, Consistency or Safety. |
| Threshold | The lowest score that counts as approved, from 0 to 1. The default is 0.7. |
Credits
A check is charged for the tokens its call actually used, at the selected model's rate. The tokens are the text, your criteria and the verdict.
- Before the run, the Run button shows an estimate (
≈and a number) of a typical check on the selected model. - A hold comes first. When the run starts, Nodaro holds enough credits for the longest answer the model may write. After the call, the check is charged what it used.
- Advanced mode and any Reasoning Effort on a Claude model run the check on the maker's own service, at that service's rate.
- Failures. A failed check that the model's maker billed is charged for that usage. A failure that was not billed is refunded.
The full rules are in How LLM calls are charged.
Tips
- Match the check to the risk. Use Safety for anything that is published automatically, and Content or Quality for drafts.
- Tune the threshold on real examples. Run the check on a few texts you already judged yourself, then raise or lower Threshold until the decisions match yours.
- Pick between drafts instead. To keep the best of several texts rather than score one, use Choose Best with AI picks the best.
- Check images with Image Critic. QA Check reads text only. For pictures, use Image Critic.
Frequently asked questions
What does QA Check return?
A score from 0.0 to 1.0, a decision, and a short reason written by the model. The text is approved when its score is at or above the Threshold, which is 0.7 by default.
Which check types does QA Check offer?
Content checks completeness, coherence and structure. Quality checks clarity, grammar and professionalism. Consistency looks for contradictions and logical errors. Safety checks that the text is free of harmful or offensive material.
What do the Approved and Rejected outputs carry?
The model's reason. Approved carries it when the text passes, and Rejected carries it when the text fails. Only one of the two has a value on each run, and neither carries the text you checked.
How much does QA Check cost?
A check is charged for the tokens its call used, at the rate of the selected model. Before the run, the Run button shows an estimate of a typical check. A longer text costs more, and a premium model costs more per token than an economy model.
Can QA Check score images?
No. QA Check reads text. To score an image for realism, prompt adherence or consistency, use the Image Critic node.
Related
Generate Text - Any LLM
Image Critic
Choose Best
Router
Credits
Last updated on
Extract Field
Pull one field or a nested path, such as title or authorMeta.name, out of JSON data, and send it on as text, as a list to run once per item, or as JSON.
JSON Process
JSON Process is a legacy node that filters and reshapes JSON. It is no longer in the node picker, but saved workflows keep running. Use Extract Field instead.