Skip to content

Prompt Template

The Prompt Template policy applies reusable, parameterized prompt templates to requests. Instead of sending a raw prompt, clients send a template reference (template://<name>?param=value) and the gateway substitutes the parameters into the configured template before forwarding the request to the LLM.

Configuration Parameters

Parameter Required Description
Templates Yes A list of named templates. Each template has a name (identifier) and a prompt string containing placeholders in the format [[parameter-name]].

Template Placeholder Syntax

Use [[parameter-name]] inside the prompt string to mark substitution points:

Translate the following text from [[from]] to [[to]]: [[text]]

When a client calls template://translate?from=English&to=Spanish&text=Hello, the policy replaces each placeholder with the corresponding query parameter value.

Advanced Settings

Parameter Default Description
JSON Path — JSONPath expression to limit template resolution to a specific field in the request payload. If omitted, template references are resolved across the entire request payload.
On Missing Template error Behavior when a referenced template name is not found. error returns an immediate error response; passthrough leaves the original template reference unchanged.
On Unresolved Placeholder keep Behavior when a placeholder has no matching query parameter. keep leaves the placeholder as-is; empty replaces it with an empty string; error returns an immediate error response.

Add This Policy

  1. Navigate to AI Workspace > LLM Providers or LLM Proxies.
  2. Click on the provider or proxy name.
  3. Go to the Guardrails tab.
  4. Click + Add Guardrail and select Prompt Template from the sidebar.
  5. Add one or more templates, each with a name and prompt string using [[param]] placeholders.
  6. Optionally expand Advanced Settings to configure JSON Path, On Missing Template, or On Unresolved Placeholder behavior.
  7. Click Add (for providers) or Submit (for proxies).
  8. Deploy the provider or proxy to apply the changes.

Example: Translation Template

Configured template:

Field Value
Name translate
Prompt Translate the following text from [[from]] to [[to]]: [[text]]

Client request body:

{
  "model": "gpt-4o",
  "messages": [
    {
      "role": "user",
      "content": "template://translate?from=English&to=Spanish&text=Hello, how are you?"
    }
  ]
}

Request forwarded to LLM:

{
  "model": "gpt-4o",
  "messages": [
    {
      "role": "user",
      "content": "Translate the following text from English to Spanish: Hello, how are you?"
    }
  ]
}