shaw/SillyTavern
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

feat(openrouter): disable reasoning if Request model reasoning is off and effort is minimum (#5079)

Browse Source 
* feat(openrouter): disable reasoning if "Request model reasoning" is disabled

* feat(openrouter): map minimum reasoning to none if request reasoning is off

* Add hint how to disable reasoning

---------

Co-authored-by: Cohee <18619528+Cohee1207@users.noreply.github.com>
This commit is contained in:
Brioch
2026-02-23 20:19:04 +01:00
committed by GitHub
parent 496b3570e6
commit 0cef10f63f
3 changed files with 15 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
public/index.html
+4 -1
View File
@@ -2129,7 +2129,7 @@
<span data-i18n="Allows the model to return its thinking process.">
Allows the model to return its thinking process.
</span>
<strong data-i18n="This setting affects visibility only." data-source-mode="except" data-source="zai,moonshot">
<strong data-i18n="This setting affects visibility only." data-source-mode="except" data-source="zai,moonshot,openrouter">
This setting affects visibility only.
</strong>
</div>
@@ -2151,6 +2151,9 @@
<div class="toggle-description justifyLeft marginBot5" data-source="openai,custom,xai,aimlapi,openrouter,pollinations,perplexity,cometapi,electronhub,azure_openai,chutes" data-i18n="OpenAI-style options: low, medium, high. Minimum and maximum are aliased to low and high. Auto does not send an effort level.">
OpenAI-style options: low, medium, high. Minimum and maximum are aliased to low and high. Auto does not send an effort level.
</div>
<strong class="toggle-description justifyLeft marginBot5" data-source="openrouter">
Request model reasoning = Off with Reasoning Effort = Minimum disables reasoning entirely on models that support that, but can cause errors with some models.
</strong>
<div class="toggle-description justifyLeft marginBot5" data-source="claude" data-i18n="Allocates a portion of the response length for thinking (min: 1024 tokens, low: 10%, medium: 25%, high: 50%, max: 95%), but minimum 1024 tokens. Auto does not request thinking.">
Allocates a portion of the response length for thinking (min: 1024 tokens, low: 10%, medium: 25%, high: 50%, max: 95%), but minimum 1024 tokens. Auto does not request thinking.
</div>
public/scripts/openai.js
+4
View File
@@ -2462,6 +2462,10 @@ function getReasoningEffort(settings = null, model = null) {
case reasoning_effort_types.auto:
return undefined;
case reasoning_effort_types.min:
if (chat_completion_sources.OPENROUTER === settings.chat_completion_source && !settings.show_thoughts) {
return 'none';
}
return [chat_completion_sources.OPENAI, chat_completion_sources.AZURE_OPENAI].includes(settings.chat_completion_source) && /^gpt-5/.test(model)
? reasoning_effort_types.min
: reasoning_effort_types.low;
src/endpoints/backends/chat-completions.js
+7 -4
View File
@@ -2070,10 +2070,13 @@ router.post('/generate', async function (request, response) {
apiKey = readSecret(request.user.directories, SECRET_KEYS.OPENROUTER);
// OpenRouter needs to pass the Referer and X-Title: https://openrouter.ai/docs#requests
headers = { ...OPENROUTER_HEADERS };
const includeReasoning = Boolean(request.body.include_reasoning);
bodyParams = {
'transforms': getOpenRouterTransforms(request),
'plugins': getOpenRouterPlugins(request),
'include_reasoning': Boolean(request.body.include_reasoning),
transforms: getOpenRouterTransforms(request),
plugins: getOpenRouterPlugins(request),
reasoning: {
exclude: !includeReasoning,
},
};
if (request.body.min_p !== undefined) {
@@ -2105,7 +2108,7 @@ router.post('/generate', async function (request, response) {
}
if (request.body.reasoning_effort) {
bodyParams['reasoning'] = { effort: request.body.reasoning_effort };
bodyParams['reasoning']['effort'] = request.body.reasoning_effort;
}
if (request.body.verbosity) {