@saltcorn/large-language-model 0.7.4 → 0.7.6

package/constants.js

@@ -1,18 +1,19 @@
 const OPENAI_MODELS = [
   "gpt-3.5-turbo",
   "gpt-3.5-turbo-16k",
-  "gpt-3.5-turbo-1106",
-  "gpt-3.5-turbo-0125",
-  "gpt-3.5-turbo-0613",
-  "gpt-3.5-turbo-16k-0613",
   "gpt-4o-mini",
   "gpt-4",
   "gpt-4-32k",
   "gpt-4-turbo-preview",
-  "gpt-4-1106-preview",
-  "gpt-4-0125-preview",
   "gpt-4-turbo",
   "gpt-4o",
+  "gpt-4.1",
+  "gpt-4.1-mini",
+  "gpt-4.1-nano",
+  "o1",
+  "o3",
+  "o3-mini",
+  "o4-mini",
 ];
 
 // https://github.com/ollama/ollama/blob/main/docs/faq.md#where-are-models-stored

package/generate.js

@@ -56,7 +56,9 @@ const getEmbedding = async (config, opts) => {
         if (!ollamaMod) throw new Error("Not implemented for this backend");
 
         const { Ollama } = ollamaMod;
-        const ollama = new Ollama();
+        const ollama = new Ollama(
+          config.ollama_host ? { host: config.ollama_host } : undefined
+        );
         const olres = await ollama.embeddings({
           model: opts?.model || config.embed_model || config.model,
           prompt: opts.prompt,
@@ -319,10 +321,16 @@ const getCompletionGoogleVertex = async (config, opts, oauth2Client) => {
   });
   const generativeModel = vertexAI.getGenerativeModel({
     model: config.model,
+    systemInstruction: {
+      role: "system",
+      parts: [{ text: opts.systemPrompt || "You are a helpful assistant." }],
+    },
+    generationCon0fig: {
+      temperature: config.temperature || 0.7,
+    },
   });
   const chatParams = {
     history: convertChatToVertex(opts.chat),
-    systemPrompt: opts.systemPrompt || "You are a helpful assistant.",
   };
   if (opts?.tools?.length > 0) {
     chatParams.tools = [
@@ -344,6 +352,7 @@ const getCompletionGoogleVertex = async (config, opts, oauth2Client) => {
       if (part.functionCall) {
         const toolCall = {
           function: prepFuncArgsForChat(part.functionCall),
+          id: Math.floor(Math.random() * 1000000),
         };
         if (!result.tool_calls) result.tool_calls = [toolCall];
         else result.tool_calls.push(toolCall);

package/index.js

@@ -65,6 +65,13 @@ ${domReady(`
                   onChange: "backendChange(this)",
                 },
               },
+              {
+                name: "ollama_host",
+                label: "Host",
+                sublabel: "Optional, for remote ollama server",
+                type: "String",                
+                showIf: { backend: "Local Ollama" },
+              },              
               {
                 name: "client_id",
                 label: "Client ID",
@@ -95,10 +102,28 @@ ${domReady(`
                 type: "String",
                 showIf: { backend: "Google Vertex AI" },
                 attributes: {
-                  options: ["gemini-1.5-pro", "gemini-1.5-flash"],
+                  options: [
+                    "gemini-1.5-pro",
+                    "gemini-1.5-flash",
+                    "gemini-2.0-flash",
+                  ],
                 },
                 required: true,
               },
+              {
+                name: "temperature",
+                label: "Temperature",
+                type: "Float",
+                sublabel:
+                  "Controls the randomness of predictions. Higher values make the output more random.",
+                showIf: { backend: "Google Vertex AI" },
+                default: 0.7,
+                attributes: {
+                  min: 0,
+                  max: 1,
+                  decimal_places: 1,
+                },
+              },
               {
                 name: "embed_model",
                 label: "Embedding model",
@@ -303,6 +328,11 @@ const routes = (config) => {
       url: "/large-language-model/vertex/authorize",
       method: "get",
       callback: async (req, res) => {
+        const role = req?.user?.role_id || 100;
+        if (role > 1) {
+          req.flash("error", req.__("Not authorized"));
+          return res.redirect("/");
+        }
         const { client_id, client_secret } = config || {};
         const baseUrl = (
           getState().getConfig("base_url") || "http://localhost:3000"
@@ -325,6 +355,11 @@ const routes = (config) => {
       url: "/large-language-model/vertex/callback",
       method: "get",
       callback: async (req, res) => {
+        const role = req?.user?.role_id || 100;
+        if (role > 1) {
+          req.flash("error", req.__("Not authorized"));
+          return res.redirect("/");
+        }
         const { client_id, client_secret } = config || {};
         const baseUrl = (
           getState().getConfig("base_url") || "http://localhost:3000"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@saltcorn/large-language-model",
-  "version": "0.7.4",
+  "version": "0.7.6",
   "description": "Large language models and functionality for Saltcorn",
   "main": "index.js",
   "dependencies": {