Use local language model with Dev Proxy

By connecting Dev Proxy to a local language model, you can improve Dev Proxy's functionality. Select Dev Proxy plugins use the language model to improve their output, where natural language is involved. By using a local language model, you can benefit from the improved functionality without incurring extra costs.

Prerequisites

Dev Proxy supports Ollama as the host for local language models. To set it up on your machine, follow the instructions in the Ollama documentation.

Configure Dev Proxy to use a local language model

To configure Dev Proxy to use a local language model, use the languageModel setting in the devproxyrc.json file.

{
  "languageModel": {
    "enabled": true
  }
}

You can use the following options as part of the languageModel setting:

Option	Description	Default value
cacheResponses	Specifies whether to cache responses from the language model.	true
enabled	Specifies whether to use a local language model.	false
model	The name of the language model to use.	phi3
url	The URL of the Ollama instance.	http://localhost:11434

By default, Dev Proxy uses the default Ollama configuration with the Phi-3 model. It also caches responses from the language model, which means, that for the same prompt, you get an immediate response without waiting for the language model to process it.

Important

When using a local language model, be sure to start Ollama before starting Dev Proxy.

Dev Proxy plugins that can use a local language model

The following Dev Proxy plugins use a local language model if available:

• OpenApiSpecGeneratorPlugin