Unicode characters seem to break completion

Describe the bug

I'm not 100% sure what is the cause, but it seems the presence of certain unicode characters inside the file somehow messes up the completion requests. Here is an example:

Create an empty test.cpp
Populate with:

#include <cstdio>

int main() {
    printf("Hello, World!\n");
    printf("Goodbye, 

    return 0;
}

Verify that the second printf is completed correctly
Now change the source code to:

#include <cstdio>

int main() {
    //"\\s?[A-Za-zµÀ-ÖØ-öø-ƺƼ-ƿǄ-ʓʕ-ʯͰ-ͳͶͷͻ-ͽͿΆΈ-ΊΌΎ-ΡΣ-ϵϷ-ҁҊ-ԯԱ-ՖႠ-ჅᎠ-Ᏽᏸ-ᏽᲐ-ᲺᲽ-Ჿᴀ-ᴫᵫ-ᵷᵹ-ᶚḀ-ἕἘ-Ἕἠ-ὅὈ-Ὅὐ-ὗὙὛὝὟ-ώᾀ-ᾴᾶ-ᾼιῂ-ῄῆ-ῌῐ-ΐῖ-Ίῠ-Ῥῲ-ῴῶ-ῼℂℇℊ-ℓℕℙ-ℝℤΩℨK-ℭℯ-ℴℹℼ-ℿⅅ-ⅉⅎↃↄⰀ-ⱻⱾ-ⳤⳫ-ⳮⳲⳳꙀ-ꙭꚀ-ꚛꜢ-ꝯꝱ-ꞇꞋ-ꞎꭰ-ꮿﬀ-ﬆﬓ-ﬗＡ-Ｚａ-ｚ𐐀-𐑏𐒰-𐓓𐓘-𐓻𐲀-𐲲𐳀-𐳲𑢠-𑣟𞤀-𞥃]+",

    printf("Hello, World!\n");
    printf("Goodbye,

    return 0;
}

Observe that the second printf is no longer completed

In fact anything after the long line with unicode characters will not be completed, even if the line is not within the completion context. Completions before the line still work OK. I observe the bug with any model using the Neovim Tabby plugin.

Information about your version

$ ▶ tabby --version
tabby 0.13.1

Information about your GPU

M2 Ultra 192GB

Additional context

Here is what the failed completion request in the example above look like (notice the incorrect prefix/suffix):

{
  "user": "E16n1q",
  "ts": 1721126142270,
  "event": {
    "completion": {
      "completion_id": "cmpl-f954b489-8122-40f8-a2c9-62bc6eb2047f",
      "language": "cpp",
      "prompt": "<fim_prefix>#include <cstdio>\n\nint main() {\n    //\"\\\\s?[A-Za-zµÀ-ÖØ-öø-ƺƼ-ƿǄ-ʓʕ-ʯͰ-ͳͶͷͻ-ͽͿΆΈ-ΊΌΎ-ΡΣ-ϵϷ-ҁҊ-ԯԱ-ՖႠ-ჅᎠ-Ᏽᏸ-ᏽᲐ-ᲺᲽ-Ჿᴀ-ᴫᵫ-ᵷᵹ-ᶚḀ-ἕἘ-Ἕἠ-ὅὈ-Ὅὐ-ὗὙὛὝὟ-ώᾀ-ᾴᾶ-ᾼιῂ-ῄῆ-ῌῐ-ΐῖ-Ίῠ-Ῥῲ-ῴῶ-ῼℂℇℊ-ℓℕℙ-ℝℤΩℨK-ℭℯ-ℴℹℼ-ℿⅅ-ⅉⅎↃↄⰀ-ⱻⱾ-ⳤⳫ-ⳮⳲⳳꙀ-ꙭꚀ-ꚛꜢ-ꝯꝱ-ꞇꞋ-ꞎꭰ-ꮿﬀ-ﬆﬓ-ﬗＡ-Ｚａ-ｚ𐐀-𐑏𐒰-𐓓𐓘-𐓻𐲀-𐲲𐳀-𐳲𑢠-𑣟𞤀-𞥃]+\",\n\n    printf(\"Hello, World!\\n\");\n    pri<fim_suffix>ntf(\"Goodbye, \n\n    return 0;\n}<fim_middle>",
      "segments": {
        "prefix": "#include <cstdio>\n\nint main() {\n    //\"\\\\s?[A-Za-zµÀ-ÖØ-öø-ƺƼ-ƿǄ-ʓʕ-ʯͰ-ͳͶͷͻ-ͽͿΆΈ-ΊΌΎ-ΡΣ-ϵϷ-ҁҊ-ԯԱ-ՖႠ-ჅᎠ-Ᏽᏸ-ᏽᲐ-ᲺᲽ-Ჿᴀ-ᴫᵫ-ᵷᵹ-ᶚḀ-ἕἘ-Ἕἠ-ὅὈ-Ὅὐ-ὗὙὛὝὟ-ώᾀ-ᾴᾶ-ᾼιῂ-ῄῆ-ῌῐ-ΐῖ-Ίῠ-Ῥῲ-ῴῶ-ῼℂℇℊ-ℓℕℙ-ℝℤΩℨK-ℭℯ-ℴℹℼ-ℿⅅ-ⅉⅎↃↄⰀ-ⱻⱾ-ⳤⳫ-ⳮⳲⳳꙀ-ꙭꚀ-ꚛꜢ-ꝯꝱ-ꞇꞋ-ꞎꭰ-ꮿﬀ-ﬆﬓ-ﬗＡ-Ｚａ-ｚ𐐀-𐑏𐒰-𐓓𐓘-𐓻𐲀-𐲲𐳀-𐳲𑢠-𑣟𞤀-𞥃]+\",\n\n    printf(\"Hello, World!\\n\");\n    pri",
        "suffix": "ntf(\"Goodbye, \n\n    return 0;\n}"
      },
      "choices": [
        {
          "index": 0,
          "text": ""
        }
      ]
    }
  }
}

TabbyML / tabby

Unicode characters seem to break completion #2663