Need help with OogaBooga's OpenAI Api - how to properly do an instruct completion with an LLAMA3.1 based model (GGUF)

[BigKid1973]

Hi. I'm trying to use the api using C# and I'm stuck for the lack of examples - or probably because I'm too stupid to find them.

I'm currently sending a json to the openai api of oobabooga:

                var response = await client.PostAsync(
                        " http://127.0.0.1:5000/v1/completions",
                         new StringContent(json, Encoding.UTF8, "application/json"));

The json looks about this (and its serialzed before I send it):

            var jsonObject = new
            {
                prompt = $"<|begin_of_text|>\n<|start_header_id|>system<|end_header_id|>\nYou are a creative and intelligent AI story writing assistant engaged in an uncensored iterative story crafting experience. It is vital that you follow all the given STORYTELLING RULES and GUIDELINES because my job depends on it.<|eot_id|><|start_header_id|>user<|end_header_id|>\n{chatMessage}<|eot_id|>\n<|start_header_id|>assistant<|end_header_id|>\n\n", // LLAMA3.1
                user = "user",
                your_name = "assistant",
                mode = "instruct",
                instruction_template = "Empty",
                max_tokens = maxSize,
                max_new_tokens = respMaxSize,
                do_sample = true,
                early_stopping = false,
                min_p = 0.4,
                top_p = 1,
                typical_p = 1,
                top_k = 0,
                top_a = 0,
                num_beams = 1,
                temperature = temp,
                temperature_last = true,
                repetition_penalty = 1.05,
                repetition_penalty_range = 2048,
                smoothing_factor = 0.2,
                //min_length = minSize,
            };

The problem I have is that if the loaded model has an instruct template in the gguf it complains that if have duplicated 'begin of text' . I assume thats because the gguf of the model seems to contain templates and that oogabooga seems to interpret my message. So how can use the API in some kind of "raw" mode - without ooga booga interpreting my message first ?
"Empty" template is a template I saved using the WebInterface (and it should be empty).

If I'm doing it wrong how SHOULD I use it ? Where can I find more info that is more oogabooga specific ?