starred/humanify

Fork 0

mirror of https://github.com/jehna/humanify.git synced 2026-04-27 01:26:00 +03:00

[GH-ISSUE #69] SAFETY RULES for Gemini. #35

New issue

Open

opened 2026-03-03 13:52:24 +03:00 by kerem · 10 comments

kerem commented

2026-03-03 13:52:24 +03:00

Owner

Originally created by @Anooxy17 on GitHub (Aug 29, 2024).
Original GitHub issue: https://github.com/jehna/humanify/issues/69

I've been trying so many times, even tried to add my own, but I am still getting errors about safety, tf

Originally created by @Anooxy17 on GitHub (Aug 29, 2024). Original GitHub issue: https://github.com/jehna/humanify/issues/69 I've been trying so many times, even tried to add my own, but I am still getting errors about safety, tf ![image](https://github.com/user-attachments/assets/c7dda0f1-78f5-4fb2-9ab2-89acfddd1779) ![image](https://github.com/user-attachments/assets/75d1b3e9-1770-4939-aa0c-e52cfef44c5b) ![image](https://github.com/user-attachments/assets/b642099a-0b1c-40d9-bb8c-08ae62e47253)

kerem added the

bug

label

2026-03-03 13:52:24 +03:00

kerem commented

2026-03-03 13:52:24 +03:00

Author

Owner

@jehna commented on GitHub (Aug 29, 2024):

Huh, that's weird. Gemini is a very recent addition, have you tried with OpenAI or local version yet?

@jehna commented on GitHub (Aug 29, 2024): Huh, that's weird. Gemini is a very recent addition, have you tried with OpenAI or local version yet?

kerem commented

2026-03-03 13:52:24 +03:00

Author

Owner

@jehna commented on GitHub (Aug 29, 2024):

I found one more, can you try:

[
      {
        category: HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT,
        threshold: HarmBlockThreshold.BLOCK_NONE
      },
      {
        category: HarmCategory.HARM_CATEGORY_HARASSMENT,
        threshold: HarmBlockThreshold.BLOCK_NONE
      },
      {
        category: HarmCategory.HARM_CATEGORY_HATE_SPEECH,
        threshold: HarmBlockThreshold.BLOCK_NONE
      },
      {
        category: HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT,
        threshold: HarmBlockThreshold.BLOCK_NONE
      },
      {
        category: HarmCategory.HARM_CATEGORY_UNSPECIFIED,
        threshold: HarmBlockThreshold.BLOCK_NONE
      }
]

@jehna commented on GitHub (Aug 29, 2024): I found one more, can you try: ``` [ { category: HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT, threshold: HarmBlockThreshold.BLOCK_NONE }, { category: HarmCategory.HARM_CATEGORY_HARASSMENT, threshold: HarmBlockThreshold.BLOCK_NONE }, { category: HarmCategory.HARM_CATEGORY_HATE_SPEECH, threshold: HarmBlockThreshold.BLOCK_NONE }, { category: HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT, threshold: HarmBlockThreshold.BLOCK_NONE }, { category: HarmCategory.HARM_CATEGORY_UNSPECIFIED, threshold: HarmBlockThreshold.BLOCK_NONE } ] ```

kerem commented

2026-03-03 13:52:25 +03:00

Author

Owner

@jehna commented on GitHub (Aug 29, 2024):

If you have some example code snippets that would result in safety errors, I'd be happy to test against them. It would help to resolve the issue.

@jehna commented on GitHub (Aug 29, 2024): If you have some example code snippets that would result in safety errors, I'd be happy to test against them. It would help to resolve the issue.

kerem commented

2026-03-03 13:52:25 +03:00

Author

Owner

@Manouchehri commented on GitHub (Aug 29, 2024):

@Anooxy17 Could you share the input script that trigger the failure? I'll try on my company's Vertex AI account to see if we get a different outcome.

@Manouchehri commented on GitHub (Aug 29, 2024): @Anooxy17 Could you share the input script that trigger the failure? I'll try on my company's Vertex AI account to see if we get a different outcome.

kerem commented

2026-03-03 13:52:25 +03:00

Author

Owner

@Anooxy17 commented on GitHub (Aug 30, 2024):

@jehna @Manouchehri
asd.txt

I put here whole file, if it's possible put output. I've only 3 requests on ChatGPT, and then syntax and safety on Gemini all the time, so if u have maybe output of this thing I'm trying to deobfuscate since I remember, and only your script is making me 60% max, feel free to put output, keep me updated if there's something wrong ;)

@Anooxy17 commented on GitHub (Aug 30, 2024): @jehna @Manouchehri [asd.txt](https://github.com/user-attachments/files/16817985/asd.txt) I put here whole file, if it's possible put output. I've only 3 requests on ChatGPT, and then syntax and safety on Gemini all the time, so if u have maybe output of this thing I'm trying to deobfuscate since I remember, and only your script is making me 60% max, feel free to put output, keep me updated if there's something wrong ;)

kerem commented

2026-03-03 13:52:25 +03:00

Author

Owner

@jehna commented on GitHub (Sep 3, 2024):

Thanks for the file! I'll check on it later – unfortunately something came up, so I don't have time to work on humanify for a little while. I'll get back to you a bit later on this 👍

@jehna commented on GitHub (Sep 3, 2024): Thanks for the file! I'll check on it later – unfortunately [something came up](https://x.com/Disobey_fi/status/1830590725748392159), so I don't have time to work on humanify for a little while. I'll get back to you a bit later on this 👍

kerem commented

2026-03-03 13:52:25 +03:00

Author

Owner

@Anooxy17 commented on GitHub (Sep 4, 2024):

Thannks for respond now, good luck 😊

@Anooxy17 commented on GitHub (Sep 4, 2024): Thannks for respond now, good luck 😊

kerem commented

2026-03-03 13:52:25 +03:00

Author

Owner

@Anooxy17 commented on GitHub (Sep 16, 2024):

Sorry for another interrupt, but have to ask, any ideas? 😄

@Anooxy17 commented on GitHub (Sep 16, 2024): Sorry for another interrupt, but have to ask, any ideas? 😄

kerem commented

2026-03-03 13:52:25 +03:00

Author

Owner

@tt-thoma commented on GitHub (Dec 30, 2024):

Maybe killing children processes?

@tt-thoma commented on GitHub (Dec 30, 2024): Maybe killing children processes?

kerem commented

2026-03-03 13:52:25 +03:00

Author

Owner

@tt-thoma commented on GitHub (Dec 30, 2024):

Things that makes sense in programming but could be badly interpreted by non-programmers

@tt-thoma commented on GitHub (Dec 30, 2024): Things that makes sense in programming but could be badly interpreted by non-programmers

kerem referenced this issue

2026-03-03 13:53:08 +03:00

[PR #35] [MERGED] Bump openai from 4.55.5 to 4.55.7 #115