Support additional request body parameters in OpenAI-compatible nodes

[toddkim95]

Feature Description

Background

When working with OpenAI-compatible backends, some providers require additional non-standard fields in the request body.

For example, when serving Qwen3 models via vLLM, reasoning (thinking) behavior is controlled through request body parameters such as:

{
  "chat_template_kwargs": {
    "enable_thinking": false
  }
}

In the OpenAI Python/JS SDK, these parameters are typically passed via extra_body.

Problem

Currently, Flowise OpenAI-compatible nodes (e.g., ChatOpenAI / OpenAI Custom Model) allow customization of:

• Base URL
• Headers (via Base Options)

However, there is no way to pass additional fields into the request body.

This makes it difficult to use OpenAI-compatible APIs that rely on extended request parameters (such as vLLM + Qwen deployments).

Expected Behavior

It would be helpful to support passing arbitrary JSON fields into the request body, similar to the extra_body option in OpenAI SDKs.

For example:

{
  "extra_body": {
    "chat_template_kwargs": {
      "enable_thinking": false
    }
  }
}

Or more generally, a field like:

• Additional Body Params (JSON)

Suggested Solution

Add an optional JSON input field (e.g., extraBody) in OpenAI-compatible nodes, and merge it into the final request payload before sending the request.

Why this matters

• Some OpenAI-compatible backends (e.g., vLLM) expose important features via additional request body fields

• Qwen3 models require parameters like chat_template_kwargs.enable_thinking to control reasoning behavior

• Without this capability, users must:

  • fork Flowise, or
  • introduce a proxy layer to modify requests

Additional Context

Happy to contribute a PR if this direction is acceptable 🙂

Feature Category

API Enhancement

Problem Statement

No response

Proposed Solution

No response

Mockups or References

No response

Additional Context

No response