Caption: Add custom model input field (#4956)

293ee0a31009f6df44595ca1916c6c7c2e6f240c

Cohee <18619528+Cohee1207@users.noreply.github.com>

Signed
Parents: e40b31b06d20122c14714351cdf9ac39d65691a4
3 files changed, +24 -1Ignore whitespace
public/scripts/extensions/caption/index.js+4 -0
@@ -699,6 +699,10 @@ jQuery(async function () {
699699 extension_settings.caption.ollama_custom_model = String($('#caption_ollama_custom_model').val()).trim();
700700 saveSettingsDebounced();
701701 });
702+ $('#caption_custom_model').val(extension_settings.caption.custom_model || '').on('input', () => {
703+ extension_settings.caption.custom_model = String($('#caption_custom_model').val()).trim();
704+ saveSettingsDebounced();
705+ });
702706 $('#caption_refresh_models').on('click', async () => {
703707 extension_settings.caption.multimodal_model = '';
704708 await switchMultimodalBlocks();
public/scripts/extensions/caption/settings.html+8 -0
@@ -177,8 +177,16 @@
177177 <option data-type="koboldcpp" value="koboldcpp_current" data-i18n="currently_loaded">[Currently loaded]</option>
178178 <option data-type="vllm" value="vllm_current" data-i18n="currently_selected">[Currently selected]</option>
179179 <option data-type="custom" value="custom_current" data-i18n="currently_selected">[Currently selected]</option>
180+ <option data-type="custom" value="custom_custom" data-i18n="[Custom model]">[Custom model]</option>
180181 </select>
181182 </div>
183+ <div data-type="custom">
184+ <label for="caption_custom_model">
185+ <span data-i18n="Model Id">Model Id</span>
186+ <small data-i18n="(for [Custom model] option)">(for [Custom model] option)</small>
187+ </label>
188+ <input id="caption_custom_model" class="text_pole" type="text" placeholder="e.g. gpt-4o" />
189+ </div>
182190 <div data-type="ollama">
183191 <div>
184192 The model must be downloaded first! Do it with the <code>ollama pull</code> command or <a href="#" id="caption_ollama_pull">click here</a>.
public/scripts/extensions/shared.js+12 -1
@@ -108,8 +108,15 @@ export async function getMultimodalCaption(base64Img, prompt) {
108108 }
109109
110110 if (isCustom) {
111+ if (extension_settings.caption.multimodal_model === 'custom_current') {
112+ requestBody.model = oai_settings.custom_model || '';
113+ }
114+
115+ if (extension_settings.caption.multimodal_model === 'custom_custom') {
116+ requestBody.model = extension_settings.caption.custom_model || '';
117+ }
118+
111119 requestBody.server_url = oai_settings.custom_url;
112- requestBody.model = oai_settings.custom_model || 'gpt-4-turbo';
113120 requestBody.custom_include_headers = oai_settings.custom_include_headers;
114121 requestBody.custom_include_body = oai_settings.custom_include_body;
115122 requestBody.custom_exclude_body = oai_settings.custom_exclude_body;
@@ -245,6 +252,10 @@ function throwIfInvalidModel(useReverseProxy) {
245252 throw new Error('Custom API URL is not set.');
246253 }
247254
255+ if (multimodalApi === 'custom' && multimodalModel === 'custom_custom' && !extension_settings.caption.custom_model) {
256+ throw new Error('Custom OpenAI-compatible Model ID is not set.');
257+ }
258+
248259 if (multimodalApi === 'aimlapi' && !secret_state[SECRET_KEYS.AIMLAPI]) {
249260 throw new Error('AI/ML API key is not set.');
250261 }