Merge branch 'staging' into gork-ai

public/index.html

@@ -1975,7 +1975,7 @@
                                             <span data-i18n="enable_functions_desc_3">Can be utilized by various extensions to provide additional functionality.</span>
                                         </div>
                                     </div>
-                                    <div class="range-block" data-source="openai,openrouter,makersuite,claude,custom,01ai,xai">
+                                    <div class="range-block" data-source="openai,openrouter,mistralai,makersuite,claude,custom,01ai,xai">
                                         <label for="openai_image_inlining" class="checkbox_label flexWrap widthFreeExpand">
                                             <input id="openai_image_inlining" type="checkbox" />
                                             <span data-i18n="Send inline images">Send inline images</span>

public/scripts/custom-request.js

@@ -43,10 +43,12 @@ import EventSourceStream from './sse-stream.js';
  * @property {boolean?} [stream=false] - Whether to stream the response
  * @property {ChatCompletionMessage[]} messages - Array of chat messages
  * @property {string} [model] - Optional model name to use for completion
- * @property {string} chat_completion_source - Source provider for chat completion
+ * @property {string} chat_completion_source - Source provider
  * @property {number} max_tokens - Maximum number of tokens to generate
  * @property {number} [temperature] - Optional temperature parameter for response randomness
- * @property {string} [custom_url] - Optional custom URL for chat completion
+ * @property {string} [custom_url] - Optional custom URL
+ * @property {string} [reverse_proxy] - Optional reverse proxy URL
+ * @property {string} [proxy_password] - Optional proxy password
  */
 
 /** @typedef {Record<string, any> & ChatCompletionPayloadBase} ChatCompletionPayload */
@@ -80,7 +82,6 @@ export class TextCompletionService {
      */
     static createRequestData({ stream = false, prompt, max_tokens, model, api_type, api_server, temperature, min_p, ...props }) {
         const payload = {
-            ...props,
             stream,
             prompt,
             max_tokens,
@@ -90,6 +91,7 @@ export class TextCompletionService {
             api_server: api_server ?? getTextGenServer(api_type),
             temperature,
             min_p,
+            ...props,
         };
 
         // Remove undefined values to avoid API errors
@@ -387,9 +389,8 @@ export class ChatCompletionService {
      * @param {ChatCompletionPayload} custom
      * @returns {ChatCompletionPayload}
      */
-    static createRequestData({ stream = false, messages, model, chat_completion_source, max_tokens, temperature, custom_url, ...props }) {
+    static createRequestData({ stream = false, messages, model, chat_completion_source, max_tokens, temperature, custom_url, reverse_proxy, proxy_password, ...props }) {
         const payload = {
-            ...props,
             stream,
             messages,
             model,
@@ -397,6 +398,11 @@ export class ChatCompletionService {
             max_tokens,
             temperature,
             custom_url,
+            reverse_proxy,
+            proxy_password,
+            use_makersuite_sysprompt: true,
+            claude_use_sysprompt: true,
+            ...props,
         };
 
         // Remove undefined values to avoid API errors

public/scripts/extensions/shared.js

@@ -1,7 +1,7 @@
 import { CONNECT_API_MAP, getRequestHeaders } from '../../script.js';
 import { extension_settings, openThirdPartyExtensionMenu } from '../extensions.js';
 import { t } from '../i18n.js';
-import { oai_settings } from '../openai.js';
+import { oai_settings, proxies } from '../openai.js';
 import { SECRET_KEYS, secret_state } from '../secrets.js';
 import { textgen_types, textgenerationwebui_settings } from '../textgen-settings.js';
 import { getTokenCountAsync } from '../tokenizers.js';
@@ -310,9 +310,10 @@ export class ConnectionManagerRequestService {
      * @param {boolean?} [custom.includePreset=true]
      * @param {boolean?} [custom.includeInstruct=true]
      * @param {Partial<InstructSettings>?} [custom.instructSettings] Override instruct settings
+     * @param {Record<string, any>} [overridePayload] - Override payload for the request
      * @returns {Promise<import('../custom-request.js').ExtractedData | (() => AsyncGenerator<import('../custom-request.js').StreamResponse>)>} If not streaming, returns extracted data; if streaming, returns a function that creates an AsyncGenerator
      */
-    static async sendRequest(profileId, prompt, maxTokens, custom = this.defaultSendRequestParams) {
+    static async sendRequest(profileId, prompt, maxTokens, custom = this.defaultSendRequestParams, overridePayload = {}) {
         const { stream, signal, extractData, includePreset, includeInstruct, instructSettings } = { ...this.defaultSendRequestParams, ...custom };
 
         const context = SillyTavern.getContext();
@@ -330,6 +331,8 @@ export class ConnectionManagerRequestService {
                         throw new Error(`API type ${selectedApiMap.selected} does not support chat completions`);
                     }
 
+                    const proxyPreset = proxies.find((p) => p.name === profile.proxy);
+
                     const messages = Array.isArray(prompt) ? prompt : [{ role: 'user', content: prompt }];
                     return await context.ChatCompletionService.processRequest({
                         stream,
@@ -338,6 +341,9 @@ export class ConnectionManagerRequestService {
                         model: profile.model,
                         chat_completion_source: selectedApiMap.source,
                         custom_url: profile['api-url'],
+                        reverse_proxy: proxyPreset?.url,
+                        proxy_password: proxyPreset?.password,
+                        ...overridePayload,
                     }, {
                         presetName: includePreset ? profile.preset : undefined,
                     }, extractData, signal);
@@ -354,6 +360,7 @@ export class ConnectionManagerRequestService {
                         model: profile.model,
                         api_type: selectedApiMap.type,
                         api_server: profile['api-url'],
+                        ...overridePayload,
                     }, {
                         instructName: includeInstruct ? profile.instruct : undefined,
                         presetName: includePreset ? profile.preset : undefined,

public/scripts/openai.js

@@ -4192,6 +4192,80 @@ function getMaxContextWindowAI(value) {
     }
 }
 
+/**
+ * Get the maximum context size for the Mistral model
+ * @param {string} model Model identifier
+ * @param {boolean} isUnlocked Whether context limits are unlocked
+ * @returns {number} Maximum context size in tokens
+ */
+function getMistralMaxContext(model, isUnlocked) {
+    if (isUnlocked) {
+        return unlocked_max;
+    }
+
+    if (Array.isArray(model_list) && model_list.length > 0) {
+        const contextLength = model_list.find((record) => record.id === model)?.max_context_length;
+        if (contextLength) {
+            return contextLength;
+        }
+    }
+
+    const contextMap = {
+        'codestral-2411-rc5': 262144,
+        'codestral-2412': 262144,
+        'codestral-2501': 262144,
+        'codestral-latest': 262144,
+        'codestral-mamba-2407': 262144,
+        'codestral-mamba-latest': 262144,
+        'open-codestral-mamba': 262144,
+        'ministral-3b-2410': 131072,
+        'ministral-3b-latest': 131072,
+        'ministral-8b-2410': 131072,
+        'ministral-8b-latest': 131072,
+        'mistral-large-2407': 131072,
+        'mistral-large-2411': 131072,
+        'mistral-large-latest': 131072,
+        'mistral-large-pixtral-2411': 131072,
+        'mistral-tiny-2407': 131072,
+        'mistral-tiny-latest': 131072,
+        'open-mistral-nemo': 131072,
+        'open-mistral-nemo-2407': 131072,
+        'pixtral-12b': 131072,
+        'pixtral-12b-2409': 131072,
+        'pixtral-12b-latest': 131072,
+        'pixtral-large-2411': 131072,
+        'pixtral-large-latest': 131072,
+        'open-mixtral-8x22b': 65536,
+        'open-mixtral-8x22b-2404': 65536,
+        'codestral-2405': 32768,
+        'mistral-embed': 32768,
+        'mistral-large-2402': 32768,
+        'mistral-medium': 32768,
+        'mistral-medium-2312': 32768,
+        'mistral-medium-latest': 32768,
+        'mistral-moderation-2411': 32768,
+        'mistral-moderation-latest': 32768,
+        'mistral-ocr-2503': 32768,
+        'mistral-ocr-latest': 32768,
+        'mistral-saba-2502': 32768,
+        'mistral-saba-latest': 32768,
+        'mistral-small': 32768,
+        'mistral-small-2312': 32768,
+        'mistral-small-2402': 32768,
+        'mistral-small-2409': 32768,
+        'mistral-small-2501': 32768,
+        'mistral-small-2503': 32768,
+        'mistral-small-latest': 32768,
+        'mistral-tiny': 32768,
+        'mistral-tiny-2312': 32768,
+        'open-mistral-7b': 32768,
+        'open-mixtral-8x7b': 32768,
+    };
+
+    // Return context size if model found, otherwise default to 32k
+    return Object.entries(contextMap).find(([key]) => model.includes(key))?.[1] || 32768;
+}
+
 /**
  * Get the maximum context size for the Groq model
  * @param {string} model Model identifier
@@ -4460,27 +4534,10 @@ async function onModelChange() {
     }
 
     if (oai_settings.chat_completion_source === chat_completion_sources.MISTRALAI) {
-        if (oai_settings.max_context_unlocked) {
+        const maxContext = getMistralMaxContext(oai_settings.mistralai_model, oai_settings.max_context_unlocked);
-            $('#openai_max_context').attr('max', unlocked_max);
+        $('#openai_max_context').attr('max', maxContext);
-        } else if (['codestral-latest', 'codestral-mamba-2407', 'codestral-2411-rc5', 'codestral-2412', 'codestral-2501'].includes(oai_settings.mistralai_model)) {
-            $('#openai_max_context').attr('max', max_256k);
-        } else if (['mistral-large-2407', 'mistral-large-2411', 'mistral-large-pixtral-2411', 'mistral-large-latest'].includes(oai_settings.mistralai_model)) {
-            $('#openai_max_context').attr('max', max_128k);
-        } else if (oai_settings.mistralai_model.includes('mistral-nemo')) {
-            $('#openai_max_context').attr('max', max_128k);
-        } else if (oai_settings.mistralai_model.includes('mixtral-8x22b')) {
-            $('#openai_max_context').attr('max', max_64k);
-        } else if (oai_settings.mistralai_model.includes('pixtral')) {
-            $('#openai_max_context').attr('max', max_128k);
-        } else if (oai_settings.mistralai_model.includes('ministral')) {
-            $('#openai_max_context').attr('max', max_32k);
-        } else {
-            $('#openai_max_context').attr('max', max_32k);
-        }
         oai_settings.openai_max_context = Math.min(oai_settings.openai_max_context, Number($('#openai_max_context').attr('max')));
         $('#openai_max_context').val(oai_settings.openai_max_context).trigger('input');
-
-        //mistral also caps temp at 1.0
         oai_settings.temp_openai = Math.min(claude_max_temp, oai_settings.temp_openai);
         $('#temp_openai').attr('max', claude_max_temp).val(oai_settings.temp_openai).trigger('input');
     }
@@ -5070,7 +5127,9 @@ export function isImageInliningSupported() {
         'o1-2024-12-17',
         'chatgpt-4o-latest',
         'yi-vision',
-        'pixtral-latest',
+        'mistral-large-pixtral-2411',
+        'mistral-small-2503',
+        'mistral-small-latest',
         'pixtral-12b-latest',
         'pixtral-12b',
         'pixtral-12b-2409',

public/scripts/textgen-models.js

@@ -58,6 +58,8 @@ const OPENROUTER_PROVIDERS = [
     'Minimax',
     'Nineteen',
     'Liquid',
+    'Stealth',
+    'NCompass',
     'InferenceNet',
     'Friendli',
     'AionLabs',
@@ -69,6 +71,9 @@ const OPENROUTER_PROVIDERS = [
     'Targon',
     'Ubicloud',
     'Parasail',
+    'Phala',
+    'Cent-ML',
+    'Venice',
     '01.AI',
     'HuggingFace',
     'Mancer',

src/endpoints/stable-diffusion.js

@@ -1166,7 +1166,8 @@ falai.post('/models', async (_request, response) => {
         const models = data
             .filter(x => !x.title.toLowerCase().includes('inpainting') &&
                 !x.title.toLowerCase().includes('control') &&
-                !x.title.toLowerCase().includes('upscale'))
+                !x.title.toLowerCase().includes('upscale') &&
+                !x.title.toLowerCase().includes('lora'))
             .sort((a, b) => a.title.localeCompare(b.title))
             .map(x => ({ value: x.modelUrl.split('fal-ai/')[1], text: x.title }));
         return response.send(models);