deepseek-coder-v2 doesn't seem to get the context

[cprn]

Hi. I tried code generation with llama3 model and it worked fine but with a custom Ollama agent running deepseek-coder-v2 it seems that the system_prompt isn't send, therefore the received response contains more than just a code snippet. Is this a bug or am I doing something wrong?

The relevant part of the inspect method with a rather empty payload (full output in attachment):

  _queries = {
    ["21840817_687b_40e1_9cda_faacb9efd554"] = {
      buf = 5,
      ex_id = 1,
      first_line = 0,
      handler = <function 29>,
      last_line = 0,
      ns_id = 27,
      on_exit = <function 30>,
      payload = {
        messages = { {
            content = "",
            role = "system"
          }, {
            content = "What's your system_prompt",
            role = "user"
          } },
        model = "deepseek-coder-v2",
        stream = true,
        temperature = 1,
        top_p = 1
      },
      provider = "ollama",
      [...cut...]
      response = " As an intelligent assistant DeepSeek Coder developed by the Chinese company DeepSeek, my system prompt is designed to provide information assistance and answer questions. My responses are based on artificial intelligence algorithms trained with large amounts of data. I am ready to assist users in acquiring knowledge and solving problems through dialogues.",
      timestamp = 1721658433
    }
  }

And this is my config (using Lazy):

    {'robitx/gp.nvim',
        config = function()
            local conf = {
                providers = {
                    openai = {disable = true},
                    googleai = {disable = true},
                    ollama = {disable = false},
                },
                agents = {
                    {name = "ChatOllamaLlama3", disable = true},
                    {name = "CodeOllamaLlama3", disable = true},
                    {
                        provider = "ollama",
                        name = "ChatOllamaDeepseekCoderV2",
                        chat = true,
                        command = false,
                        model = {
                            model = "deepseek-coder-v2",
                            num_ctx = 8192,
                        },
                        system_prompt = "You are a general AI assistant.",

                    },
                    {
                        provider = "ollama",
                        name = "CodeOllamaDeepseekCoderV2",
                        chat = false,
                        command = true,
                        model = {
                            model = "deepseek-coder-v2",
                            temperature = 1.9,
                            top_p = 1,
                            num_ctx = 8192,
                        },
                        system_prompt = "You are an AI working as a code editor providing answers.\n\n"
                            .. "Use 4 SPACES FOR INDENTATION.\n"
                            .. "Please AVOID COMMENTARY OUTSIDE OF THE SNIPPET RESPONSE.\n"
                            .. "START AND END YOUR ANSWER WITH:\n\n```",
                    },
                },
            }
            require("gp").setup(conf)
        end,
    },

full GpInspectPlugin output

[cprn]

Works great, thanks! Also, it was a blazing fast fix! 🚀