# LLM Seed Configuration Feature

## Overview

This feature introduces a **configurable LLM Seed** value that allows users to control the deterministic behavior of LLM responses.
By setting a specific integer seed, users (especially developers and prompt engineers) can reproduce consistent outputs for the same prompt and parameters.

---

## 1. UI: Add “LLM Seed” Field

### Description

A new optional numeric input field **“LLM Seed”** will be added to the **LLM Configuration** section in the UI.
This allows users to specify an integer seed value that controls the reproducibility of model generations.

### Behavior

* Field label: **Seed**
* Input type: numeric (integer)
* Placeholder: `e.g. 42`
* Optional: If left empty, the backend will not include a seed in the configuration.
* Validation:

* Must be an integer.
* Acceptable range: `0` – `2,147,483,647` (32-bit signed integer).
* Non-numeric values should show a validation warning.

### UI Layout


---

## 2. Kalimera Backend: Expose Seed in Configuration

### Description

The Kalimera Backend must include the new **`seed`** field as part of the configuration object returned from the LLM settings API.

### Technical Details

* **New field:** `seed`
* **Type:** `integer | null`
* **Source:** Value provided from the UI configuration.
* **Default:** `null` (if no seed was entered by the user).


### Example Payload

 "dialogNodes": [
            {
                "topP": 1,
                "nodeType": "Welcome",
                "maxTokens": 15000,
                "userPrompt": "",
                "temperature": 0.1,
                "seed": 42,...

If no value is provided:

"dialogNodes": [
            {
                "topP": 1,
                "nodeType": "Welcome",
                "maxTokens": 15000,
                "userPrompt": "",
                "temperature": 0.1,
                "seed": null,...

---

## 3. VoiceAI Integration: Use Seed When Creating LLM Connection

### Description

The **VoiceAI team** should update the LLM connection logic to consume the new `seed` field from the configuration and include it when initializing or sending requests to the LLM API.

### Implementation Notes

* Retrieve `seed` from the Kalimera configuration.
* When creating the LLM connection or request payload, include:

  "seed": <value>

* If `seed` is `null`, omit the field entirely (to allow random generation as usual).