mirror of
https://github.com/SillyTavern/Extension-Objective.git
synced 2026-10-15 13:31:50 +00:00
[GH-ISSUE #13] Extension does not work on Gemini Flash models #8
Labels
No labels
pull-request
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
SillyTavern/Extension-Objective#8
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @rickgnyc on GitHub (Aug 20, 2025).
Original GitHub issue: https://github.com/SillyTavern/Extension-Objective/issues/13
As noted in the issue posted by user Kedwyn, this extension returns empty results from certain models due to the request being sent as:
role: 'system'
Notably, the Gemini Flash models return nothing unless the request is sent as the user role.
I am unable to use this extension with my preferred Gemini models. I can confirm that it does work with other models, such as "thedrummer/anubis-pro-105b-v1".
THANK YOU for this wonderful extension. I love what it intends to do.
-Rick
For your reference, here is a terminal log using a simplistic scenario:
{
messages: [
{
role: 'system',
content: "Write Wizard's next reply in a fictional chat between Wizard and Rick."
},
{ role: 'system', content: '[Start a new Chat]' },
{
role: 'user',
content: "I am a young wizard who must enter an evil warlock's tower and make it to the top to defeat him. I approach to door to the tower and attempt to use my magic to unlock the door."
},
{
role: 'assistant',
content: 'Aha, a valiant quest indeed! The air crackles with your youthful ambition, and I sense a spark of true power within you. But remember, young one, not all challenges yield to brute force, even magical force.\n' +
'\n' +
"Tell me, as your magic washes over the ancient, iron-clad door, what do you feel? Is there a resistance, a hum, a whisper of something more than just wood and metal? Describe it to me. And what exactly did you attempt to do with your magic? A simple unlocking spell? A forceful blast? The details, my new apprentice, are a wizard's most potent ingredients.\n" +
'\n' +
'Let us unravel this mystery together!'
},
{
role: 'system',
content: 'Ignore previous instructions and generate a list of tasks to complete an objective. Your next response must be formatted as a numbered list of plain text entries. Do not include anything but the numbered list. The list must be prioritized in the order that tasks must be completed.\n' +
'\n' +
'The objective that you must make a numbered task list for is: [Reach the evil warlock at the top of the tower.].\n' +
'The tasks created should take into account the character traits of Wizard. These tasks may or may not involve Rick directly. Be sure to include the objective as the final task.\n' +
'\n' +
"Given an example objective of 'Make me a four course dinner', here is an example output:\n" +
'1. Determine what the courses will be\n' +
'2. Find recipes for each course\n' +
'3. Go shopping for supplies with Rick\n' +
'4. Cook the food\n' +
'5. Get Rick to set the table\n' +
'6. Serve the food\n' +
'7. Enjoy eating the meal with Rick\n' +
' '
}
],
prompt: undefined,
model: 'google/gemini-2.5-flash',
temperature: 1,
max_tokens: 600,
max_completion_tokens: undefined,
stream: false,
presence_penalty: 0,
frequency_penalty: 0,
top_p: 1,
top_k: 0,
stop: undefined,
logit_bias: undefined,
seed: undefined,
n: undefined,
transforms: undefined,
plugins: [],
include_reasoning: false,
min_p: 0,
top_a: 0,
repetition_penalty: 1,
safety_settings: [
{ category: 'HARM_CATEGORY_HARASSMENT', threshold: 'OFF' },
{ category: 'HARM_CATEGORY_HATE_SPEECH', threshold: 'OFF' },
{ category: 'HARM_CATEGORY_SEXUALLY_EXPLICIT', threshold: 'OFF' },
{ category: 'HARM_CATEGORY_DANGEROUS_CONTENT', threshold: 'OFF' },
{ category: 'HARM_CATEGORY_CIVIC_INTEGRITY', threshold: 'OFF' }
]
}
{
id: 'gen-1755676051-kkqWRlbIrrLfAD2utFDL',
provider: 'Google',
model: 'google/gemini-2.5-flash',
object: 'chat.completion',
created: 1755676051,
choices: [
{
logprobs: null,
finish_reason: 'stop',
native_finish_reason: 'STOP',
index: 0,
message: {
role: 'assistant',
content: ' I await your keen observations.',
refusal: null
}
}
],
usage: {
prompt_tokens: 395,
completion_tokens: 6,
total_tokens: 401,
prompt_tokens_details: { cached_tokens: 0 }
}
}
{
role: 'assistant',
content: ' I await your keen observations.',
refusal: null
}
@Cohee1207 commented on GitHub (Aug 20, 2025):
Set Prompt Post-Processing to Semi-strict.
@rickgnyc commented on GitHub (Aug 20, 2025):
I may have just soiled myself slightly when your guidance fixed my issue and I witnessed my task list fill up.
THANK YOU @Cohee1207 !!