Problem with gpt-4o-search-preview and Request for a Complete LangChain Code Example

[Yunus01]

Hello everyone,

I encountered an issue when using the OpenAI Chat Model node with the model gpt-4o-search-preview. I get the following error:

Bad request - please check your parameters
Model incompatible request arguments supplied: frequency_penalty, n, presence_penalty, temperature, top_p

However, I’m not providing any of these parameters in my node. I suspect that this model does not support some of the default arguments.

Therefore, I would like to create my own custom Language Model using the LangChain Code node in order to have better control over the parameters. Could someone please provide me with a complete code example for calling the gpt4o (or gpt-4o-search-preview) model via the LangChain Code node?

I need to know exactly what to include in the code of my LangChain Code node to use it correctly.

Thank you very much for your help!

[Yunus01]

so I found an alternative method by using the HTTP Request node in n8n. Here’s how I set it up:

1. Add an HTTP Request Node to your workflow.
2. Configure the node as follows:

• Method: POST
• URL: https://api.openai.com/v1/chat/completions
• Headers:
  • Content-Type: application/json
  • Authorization: Bearer YOUR_OPENAI_API_KEY

3. Set up the Body:
   In the “Body” field, add the following JSON:

{
  "model": "gpt-4o-search-preview",
  "messages": [
    {
      "role": "user",
      "content": "Your question here"
    }
  ]
}

4. Optional - Adding a system prompt:
   If you want to include a system prompt, modify the JSON as follows:

{
  "model": "gpt-4o-search-preview",
  "messages": [
    {
      "role": "system",
      "content": "Your system prompt here"
    },
    {
      "role": "user",
      "content": "Your question here"
    }
  ]
}

This setup allows you to bypass the issues with default parameters and gives you more control over the request sent to the OpenAI API. I hope this helps anyone facing similar issues!

[Yunus01]

I have tested the solution with the LangChain Code node for using the “gpt-4o-search-preview” model, but it doesn’t work either. I’m sharing the solution that works with other models.

Example: Using “LangChain Code” with gpt-4-mini

1. In your LangChain Code node, click on Add Code and select Execute.
2. Paste the following code:

const { PromptTemplate } = require('@langchain/core/prompts');
const { ChatOpenAI } = require('@langchain/openai');

const query = 'Tell me a joke';
const prompt = PromptTemplate.fromTemplate(query);

// Create an instance of ChatOpenAI with the GPT-4o-mini model
const llm = new ChatOpenAI({
  modelName: "gpt-4o-mini", // replace with your desired model
  temperature: 0.7,
  openAIApiKey: 'YOUR API KEY',
});

let chain = prompt.pipe(llm);
const output = await chain.invoke();
return [ { json: { output } } ];

3. In your node’s Input and Output section, add “Main” and leave the default values.
4. Execute the code; it should work as expected with the gpt-4-mini model.

---

This example should help those looking to integrate custom language models using the LangChain Code node in n8n, even though the “gpt-4o-search-preview” model is currently not supported with this method.