Using a tab completions template (moustache) encodes special characters to html

[Lucki]

Before submitting your bug report

• [X] I believe this is a bug. I'll try to join the Continue Discord for questions
• [X] I'm not able to find an open issue that reports the same bug
• [X] I've seen the troubleshooting guide on the Continue Docs

Relevant environment info

- OS: Arch Linux
- Continue: v0.9.80 (pre-release)
- IDE: VSCode 1.87.0

Description

Defining a template in tabAutocompleteOptions will html encode special characters before they're sent to the model. E.g.: " becomes ", ' becomes ' - The model adapts to this and the response is equally useless.

To reproduce

1. Define a template in tabAutocompleteOptions. E.g. for deepseek-coder:

   "tabAutocompleteOptions": {
     "template": "<｜fim▁begin｜>{{prefix}}<｜fim▁hole｜>{{suffix}}<｜fim▁end｜>"
   },
   

2. Create a new file, make sure autocomplete is turned on.
3. Write something containing quotation marks. E.g.: print("asd")
4. Press Return and enter print:
5. View the log of your LLM server. Example from Koboldcpp:

   {
       "model": "AUTODETECT",
       "max_tokens": 1024,
       "temperature": 0,
       "stop": [
           "\n\n",
           "```",
           "function",
           "class",
           "module",
           "export"
       ],
       "prompt": "<\uff5cfim\u2581begin\uff5c>print(&quot;asd&quot;)\np<\uff5cfim\u2581hole\uff5c><\uff5cfim\u2581end\uff5c>",
       "stream": true
   }
   

Log output

Notice " vs \" in both examples as a replacement for ".

Log with template:

{
    "model": "AUTODETECT",
    "max_tokens": 1024,
    "temperature": 0,
    "stop": [
        "\n\n",
        "",
        "function",
        "class",
        "module",
        "export"
    ],
    "prompt": "<\uff5cfim\u2581begin\uff5c>print(&quot;asd&quot;)\np<\uff5cfim\u2581hole\uff5c><\uff5cfim\u2581end\uff5c>",
    "stream": true
}

Log without template: (this doesn't acutally give me any output because it's the wrong format)

{
    "model": "AUTODETECT",
    "max_tokens": 1024,
    "temperature": 0,
    "stop": [
        "<fim_prefix>",
        "<fim_suffix>",
        "<fim_middle>",
        "<|endoftext|>",
        "\n\n",
        "```",
        "function",
        "class",
        "module",
        "export"
    ],
    "prompt": "<fim_prefix>// Path: config.json\n// ]\n// }\n// \n//   \"slashCommands\": [\n//     {\n//       \"name\": \"edit\",\n//       \"description\": \"Edit selected code\"\n//\n\nprint(\"asd\")\nprint<fim_suffix><fim_middle>",
    "stream": true
}

tabAutocompleteModel config:

"tabAutocompleteModel": {
  "title": "asd",
  "provider": "openai",
  "model": "AUTODETECT",
  "apiBase": "http://localhost:5001/v1",
  "template": "deepseek",
  "apiKey": ""
},