open ai image tool for codecompanion

image.lua

local xml2lua = require("codecompanion.utils.xml.xml2lua")
local log = require("codecompanion.utils.log")

local function get_secret()
  local cmd = os.getenv("HOME") .. "/scripts/get_secret.sh"
  local handle = io.popen(cmd, "r")
  if handle then
    local result = handle:read("*a")
    log:trace("Executed cmd: %s", cmd)
    handle:close()
    return result:gsub("%s+$", "")
  else
    return log:error("Error: Could not execute cmd: %s", cmd)
  end
end

---@class CodeCompanion.Tool
return {
  name = "img",
  cmds = {
    {
      "curl",
      "${url}",
      "-H",
      "Accept: application/json",
      "-H",
      "Content-Type: application/json",
      "-H",
      "Authorization: Bearer ${api_key}",
      "-d",
      '{ "model": "gpt-4o-mini", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "${query}" }, { "type": "image_url", "image_url": { "url": "${image_url}" } } ] } ], "max_tokens": 300 }',
    },
  },
  schema = {
    tool = {
      _attr = { name = "img" },
      action = {
        url = "https://media.istockphoto.com/id/1944476521/photo/portrait-of-a-pretty-blue-merle-australian-shepherd-dog-looking-straigth-at-the-camera-with.jpg?s=1024x1024&w=is&k=20&c=cUz-C8K91qUiT5PXys3mu5cYGFLwRYShqg8jH0V89jQ=",
        query = "Describe the image",
      },
    },
  },
  env = function(tool)
    local api_key = get_secret()
    return {
      url = "https://api.openai.com/v1/chat/completions",
      image_url = tool.action.url,
      api_key = api_key,
      query = tool.action.query,
    }
  end,
  system_prompt = function(schema)
    local xmlstring = xml2lua.toXml({ tools = { schema } })
    local temp = string.format(
      [[### You have gained access to a new tool!

Name: IMG (image anaylisis with vision ai)
Purpose: This gives you the ability to anaylise images
Why: Sometimes you may need to analyize images from urls to help you with your responses
Usage: To use this tool, you need to return an XML markdown code block (with backticks). Consider the following schema:

```xml
%s
```

You must:
- Only use the tool when you are provided with an image url in the context
- Be mindful that you may not be required to use the tool in all of your responses
- However if the user states you should use the image tool directly, you should use it without bothering the user
- Ensure the XML markdown code block is valid and follows the schema]],
      xmlstring
    )
    return temp
  end,
  output_error_prompt = function(error)
    if type(error) == "table" then
      error = table.concat(error, "\n")
    end
    return string.format(
      [[After the tool completed, there was an error:

```
%s
```
]],
      error
    )
  end,
  output_prompt = function(output)
    if type(output) == "table" then
      output = table.concat(output, "\n")
      output = vim.json.decode(output)
      output = output.choices[1].message.content
    end

    return string.format(
      [[After viewing the image, this is what the tool returned:

### tool output

```
%s
```
]],
      output
    )
  end,
}