NanoGPT: Pull max context and vision capability from endpoint

5eea41d315eb1e53729c3db2383f7824ed6ce105

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

public/index.html+1 -1 Viewed

                                             <strong data-i18n="enable_functions_desc_4">Not supported when Prompt Post-Processing with "no tools" is used!</strong>
                                         </div>
                                     </div>
-                                    <div class="range-block" data-source="openai,aimlapi,openrouter,mistralai,makersuite,vertexai,claude,custom,xai,pollinations,moonshot,cohere,cometapi">
+                                    <div class="range-block" data-source="openai,aimlapi,openrouter,mistralai,makersuite,vertexai,claude,custom,xai,pollinations,moonshot,cohere,cometapi,nanogpt">
                                         <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/openai.js+26 -6 Viewed

     return max_32k;
+}
+/**
+ * Get the maximum context size for the NanoGPT model
+ * @param {string} model Model identifier
+ * @param {boolean} isUnlocked Whether context limits are unlocked
+ * @returns {number} Maximum context size in tokens
+ */
+function getNanoGptMaxContext(model, isUnlocked) {
+    if (isUnlocked) {
+        return unlocked_max;
+    }
+    if (Array.isArray(model_list)) {
+        const modelInfo = model_list.find(m => m.id === model);
+        if (modelInfo?.context_length) {
+            return modelInfo.context_length;
+        }
+    }
+    return max_128k;
+}
 async function onModelChange() {
     biasCache = undefined;
     let value = String($(this).val() || '');

+    }
     if (oai_settings.chat_completion_source === chat_completion_sources.NANOGPT) {
-        if (oai_settings.max_context_unlocked) {
+        const maxContext = getNanoGptMaxContext(oai_settings.nanogpt_model, oai_settings.max_context_unlocked);
-            $('#openai_max_context').attr('max', unlocked_max);
+        $('#openai_max_context').attr('max', maxContext);
-        } else {
-            $('#openai_max_context').attr('max', max_128k);
-        }
         oai_settings.openai_max_context = Math.min(Number($('#openai_max_context').attr('max')), oai_settings.openai_max_context);
         $('#openai_max_context').val(oai_settings.openai_max_context).trigger('input');
+        oai_settings.temp_openai = Math.min(oai_max_temp, oai_settings.temp_openai);
         $('#temp_openai').attr('max', oai_max_temp).val(oai_settings.temp_openai).trigger('input');
+    }

             return true;
         case chat_completion_sources.MOONSHOT:
             return visionSupportedModels.some(model => oai_settings.moonshot_model.includes(model));
+        case chat_completion_sources.NANOGPT:
+            return (Array.isArray(model_list) && model_list.find(m => m.id === oai_settings.nanogpt_model)?.capabilities?.vision);
         default:
             return false;
+    }

src/endpoints/backends/chat-completions.js+13 -6 Viewed

 import util from 'node:util';
 import express from 'express';
 import fetch from 'node-fetch';
+import urlJoin from 'url-join';
 import {
     AIMLAPI_HEADERS,

 router.post('/status', async function (request, statusResponse) {
     if (!request.body) return statusResponse.sendStatus(400);
-    let apiUrl;
+    let apiUrl = '';
-    let apiKey;
+    let apiKey = '';
-    let headers;
+    let headers = {};
+    let queryParams = {};
     if (request.body.chat_completion_source === CHAT_COMPLETION_SOURCES.OPENAI) {
         apiUrl = new URL(request.body.reverse_proxy || API_OPENAI).toString();

         apiUrl = API_NANOGPT;
         apiKey = readSecret(request.user.directories, SECRET_KEYS.NANOGPT);
         headers = {};
+        queryParams = { detailed: true };
     } else if (request.body.chat_completion_source === CHAT_COMPLETION_SOURCES.DEEPSEEK) {
-        apiUrl = new URL(request.body.reverse_proxy || API_DEEPSEEK.replace('/beta', ''));
+        apiUrl = new URL(request.body.reverse_proxy || API_DEEPSEEK.replace('/beta', '')).toString();
         apiKey = request.body.reverse_proxy ? request.body.proxy_password : readSecret(request.user.directories, SECRET_KEYS.DEEPSEEK);
         headers = {};
     } else if (request.body.chat_completion_source === CHAT_COMPLETION_SOURCES.XAI) {
-        apiUrl = new URL(request.body.reverse_proxy || API_XAI);
+        apiUrl = new URL(request.body.reverse_proxy || API_XAI).toString();
         apiKey = request.body.reverse_proxy ? request.body.proxy_password : readSecret(request.user.directories, SECRET_KEYS.XAI);
         headers = {};
     } else if (request.body.chat_completion_source === CHAT_COMPLETION_SOURCES.AIMLAPI) {

+    }
     try {
-        const response = await fetch(apiUrl + '/models', {
+        const modelsUrl = new URL(urlJoin(apiUrl, '/models'));
+        Object.keys(queryParams).forEach(key => {
+            modelsUrl.searchParams.append(key, queryParams[key]);
+        });
+        const response = await fetch(modelsUrl, {
             method: 'GET',
             headers: {
                 'Authorization': 'Bearer ' + apiKey,