feat: improve support for text-generation-webui

[randoentity]

Via OpenAI API plugin: https://github.com/oobabooga/text-generation-webui/blob/main/docs/12%20-%20OpenAI%20API.md

Without these changes:

• 0 tokens were generated because the character was missing.
• Streaming was broken and popped up a new hover for each token.
• Temperature and p settings seem to improve quality, but I did not benchmark it in this context.
• API key is optional (but still recommended) when on a local network.

Untested: if the missing API key warning still fires.

I'm just getting started with nvim and Lua.

I don't have any cloud based LLM services set up so I didn't test against those.

There are some necessary configuration changes to not get errors. Here's an example for anyone looking to get started (with text-generation-webui):

chatgpt.lua

return {
  "jackMort/ChatGPT.nvim",
    event = "VeryLazy",
  config = function()
    require("chatgpt").setup({
      api_host_cmd = "echo -n http://127.0.0.1:5000",
      actions_paths = { "~/.config/nvim/lua/plugins/chatgpt-actions.json" },
      openai_params = {
        character = "Example",
        model = "",
        max_tokens = 300,
        frequency_penalty = 0,
        temperature = 0.99,
        stream = true,
        mode = "chat",
      },
      openai_edit_params = {
        model = "",
        max_tokens = 300,
        frequency_penalty = 0,
        temperature = 0.99,
        stream = false,
      },
    })
  end,
    dependencies = {
      "MunifTanjim/nui.nvim",
      "nvim-lua/plenary.nvim",
      "nvim-telescope/telescope.nvim"
    }
}

chatgpt-actions.json

{
  "explain_code": {
    "type": "chat",
    "opts": {
      "title": "Explain Code",
      "template": "Explain the following code:\n\nCode:\n```{{filetype}}\n{{input}}\n```\n\nUse markdown format.\nHere's what the above code is doing:\n```",
      "strategy": "display",
      "params": {
        "mode": "instruct",
        "model": "",
        "stop": [
          "```"
        ],
        "stream": false
      }
    }
  },
  "summarize": {
    "type": "chat",
    "opts": {
      "template": "Summarize the following text.\n\nText:\n\"\"\"\n{{input}}\n\"\"\"\n\nSummary:",
      "strategy": "edit",
      "params": {
        "mode": "instruct",
        "model": "",
        "stream": false
      }
    }
  }
}

[jackMort]

I don't see the reason to add to main repo this specific for text-generation-webui config, you can still overwrite the settings as you mentioned.