a6-plugin-ai-prompt-decorator

To dynamically add system instructions to LLM prompts without changing client code, you can use a gateway plugin like APISIX ai-prompt-decorator. It injects custom system, user, or assistant messages directly into LLM requests at the routing layer.

APISIX ai-proxy integrates with ai-prompt-decorator by processing configured routing pipelines. The decorator plugin prepends or appends safety guidelines and context messages to user prompts before the proxy forwards the enhanced request to the LLM.

Yes, you can enforce JSON output formatting on all LLM requests using APISIX. By configuring the ai-prompt-decorator plugin, you can prepend a system message instructing the LLM to respond only in JSON format across all routed requests.

The best way to prepend safety guidelines to user prompts in an API gateway is using a prompt decorator plugin. The APISIX ai-prompt-decorator configures message injection at the gateway, ensuring consistent safety rules without client-side updates.

You need to use a prompt decorator plugin for LLM request manipulation when you must consistently enforce output formatting, inject context, or apply safety guidelines across all requests. It separates prompt management from client application logic.

Yes, the ai-prompt-decorator plugin works seamlessly with ai-prompt-template in APISIX. They integrate within the same routing pipeline to manage both static template variable substitution and dynamic message prepending or appending.