[GH-ISSUE #404] Bug Report: When the filename contains Chinese characters, all AI-generated code is incorrectly classified as human-written. #152

New issue

Closed

opened 2026-03-02 04:12:20 +03:00 by kerem · 4 comments

kerem commented

2026-03-02 04:12:20 +03:00

Owner

Originally created by @ShuboLiang on GitHub (Jan 26, 2026).
Original GitHub issue: https://github.com/git-ai-project/git-ai/issues/404

Issue Summary

When I use an AI agent to generate code in a file whose filename contains Chinese characters, all the generated lines are incorrectly classified as human-written. Running the command git-ai stats --json produces the following output:

{
  "human_additions": 9,
  "mixed_additions": 0,
  "ai_additions": 0,
  "ai_accepted": 0,
  "total_ai_additions": 9,
  "total_ai_deletions": 0,
  "time_waiting_for_ai": 0,
  "git_diff_deleted_lines": 0,
  "git_diff_added_lines": 9,
  "tool_model_breakdown": {
    "turing-coder::deepseek-v3": {
      "ai_additions": 0,
      "mixed_additions": 0,
      "ai_accepted": 0,
      "total_ai_additions": 9,
      "total_ai_deletions": 0,
      "time_waiting_for_ai": 0
    }
  }
}

The puzzling part is that total_ai_additions correctly reports 9, yet ai_additions is 0, with all additions attributed to human_additions. This perhaps suggests the git-ai tracks AI-generated content internally but fails to classify it properly when the filename includes non-ASCII (e.g., Chinese) characters.

Notably, if the filename uses only English characters, the classification works correctly—AI-generated lines are properly counted under ai_additions.

enviroment

git-ai-version: latest
os: windows11

Originally created by @ShuboLiang on GitHub (Jan 26, 2026). Original GitHub issue: https://github.com/git-ai-project/git-ai/issues/404 # Issue Summary When I use an AI agent to generate code in a file whose **filename contains Chinese characters**, all the generated lines are incorrectly classified as human-written. Running the command `git-ai stats --json` produces the following output: ```json { "human_additions": 9, "mixed_additions": 0, "ai_additions": 0, "ai_accepted": 0, "total_ai_additions": 9, "total_ai_deletions": 0, "time_waiting_for_ai": 0, "git_diff_deleted_lines": 0, "git_diff_added_lines": 9, "tool_model_breakdown": { "turing-coder::deepseek-v3": { "ai_additions": 0, "mixed_additions": 0, "ai_accepted": 0, "total_ai_additions": 9, "total_ai_deletions": 0, "time_waiting_for_ai": 0 } } } ``` The puzzling part is that `total_ai_additions` correctly reports 9, yet `ai_additions` is 0, with all additions attributed to `human_additions`. This perhaps suggests the git-ai tracks AI-generated content internally but fails to classify it properly when the **filename includes non-ASCII (e.g., Chinese) characters**. Notably, if the filename uses only English characters, the classification works correctly—AI-generated lines are properly counted under `ai_additions`. ## enviroment - git-ai-version: latest - os: windows11

kerem

2026-03-02 04:12:20 +03:00

closed this issue
added the
bug

good first issue
labels

kerem commented

2026-03-02 04:12:21 +03:00

Author

Owner

@svarlamov commented on GitHub (Jan 28, 2026):

@ShuboLiang Thank you -- we'll get this patched right away.

@Krishnachaitanyakc Wondering if you'd be interested in taking this one on? Think we should probably also do some TDD with complex utf-8 char file contents, names, etc.

@svarlamov commented on GitHub (Jan 28, 2026): @ShuboLiang Thank you -- we'll get this patched right away. @Krishnachaitanyakc Wondering if you'd be interested in taking this one on? Think we should probably also do some TDD with complex utf-8 char file contents, names, etc.

kerem commented

2026-03-02 04:12:21 +03:00

Author

Owner

@ShuboLiang commented on GitHub (Feb 4, 2026):

hello! is this in progress? @svarlamov

@ShuboLiang commented on GitHub (Feb 4, 2026): hello! is this in progress? @svarlamov

kerem commented

2026-03-02 04:12:21 +03:00

Author

Owner

@svarlamov commented on GitHub (Feb 4, 2026):

Hey! I see no one has picked up yet -- should be in the release after next though. Probably in production by early next week

@svarlamov commented on GitHub (Feb 4, 2026): Hey! I see no one has picked up yet -- should be in the release after next though. Probably in production by early next week

kerem commented

2026-03-02 04:12:21 +03:00

Author

Owner

@Krishnachaitanyakc commented on GitHub (Feb 4, 2026):

@svarlamov sorry, couldn't pick this up earlier. Created a PR, started with chinese but expanded it to others as well to make it comprehensive. I am not sure if it's overkill tbh, let me know.

@Krishnachaitanyakc commented on GitHub (Feb 4, 2026): @svarlamov sorry, couldn't pick this up earlier. Created a PR, started with chinese but expanded it to others as well to make it comprehensive. I am not sure if it's overkill tbh, let me know.

kerem referenced this issue

2026-03-02 04:13:12 +03:00

[PR #152] mock_ai pathspecs #291