@saltcorn/large-language-model 0.4.3 → 0.4.5

package/generate.js

@@ -23,21 +23,31 @@ const getEmbedding = async (config, opts) => {
         {
           embeddingsEndpoint: config.embed_endpoint,
           bearer: config.api_key,
-          embed_model: config.model,
+          embed_model: config.embed_model || config.model,
         },
         opts
       );
     case "Local Ollama":
-      if (!ollamaMod) throw new Error("Not implemented for this backend");
+      if (config.embed_endpoint) {
+        return await getEmbeddingOpenAICompatible(
+          {
+            embeddingsEndpoint: config.embed_endpoint,
+            embed_model: config.embed_model || config.model,
+          },
+          opts
+        );
+      } else {
+        if (!ollamaMod) throw new Error("Not implemented for this backend");
 
-      const { Ollama } = ollamaMod;
-      const ollama = new Ollama();
-      const olres = await ollama.embeddings({
-        model: opts?.model || config.embed_model || config.model,
-        prompt: opts.prompt,
-      });
-      //console.log("embedding response ", olres);
-      return olres.embedding;
+        const { Ollama } = ollamaMod;
+        const ollama = new Ollama();
+        const olres = await ollama.embeddings({
+          model: opts?.model || config.embed_model || config.model,
+          prompt: opts.prompt,
+        });
+        //console.log("embedding response ", olres);
+        return olres.embedding;
+      }
     default:
       throw new Error("Not implemented for this backend");
   }
@@ -71,7 +81,7 @@ const getCompletion = async (config, opts) => {
       const ollama = new Ollama();
       const olres = await ollama.generate({
         model: config.model,
-        prompt: opts.prompt,
+        ...opts,
       });
       //console.log("the response ", olres);
       return olres.response;
@@ -140,8 +150,11 @@ const getCompletionOpenAICompatible = async (
   );
 };
 
-const getEmbeddingOpenAICompatible = async (config, { prompt, model }) => {
-  const { embeddingsEndpoint, bearer, debugResult, embed_model } = config;
+const getEmbeddingOpenAICompatible = async (
+  config,
+  { prompt, model, debugResult }
+) => {
+  const { embeddingsEndpoint, bearer, embed_model } = config;
   const headers = {
     "Content-Type": "application/json",
     Accept: "application/json",
@@ -162,6 +175,7 @@ const getEmbeddingOpenAICompatible = async (config, { prompt, model }) => {
   if (debugResult)
     console.log("OpenAI response", JSON.stringify(results, null, 2));
   if (results.error) throw new Error(`OpenAI error: ${results.error.message}`);
+  if (Array.isArray(prompt)) return results?.data?.map?.((d) => d?.embedding);
   return results?.data?.[0]?.embedding;
 };
 module.exports = { getCompletion, getEmbedding };

package/index.js

@@ -100,22 +100,30 @@ const configuration_workflow = () =>
                 name: "embed_model",
                 label: "Embedding model",
                 type: "String",
-                showIf: { backend: "Local Ollama" },
+                showIf: { backend: ["OpenAI-compatible API", "Local Ollama"] },
               },
               {
                 name: "endpoint",
                 label: "Chat completions endpoint",
                 type: "String",
-                sublabel: "Example: http://localhost:8080/v1/chat/completions",
+                sublabel: "Example: http://127.0.0.1:8080/v1/chat/completions",
                 showIf: { backend: "OpenAI-compatible API" },
               },
               {
                 name: "embed_endpoint",
                 label: "Embedding endpoint",
                 type: "String",
-                sublabel: "Example: http://localhost:8080/v1/embeddings",
+                sublabel: "Example: http://127.0.0.1:8080/v1/embeddings",
                 showIf: { backend: "OpenAI-compatible API" },
               },
+              {
+                name: "embed_endpoint",
+                label: "Embedding endpoint",
+                type: "String",
+                sublabel:
+                  "Optional. Use an alternative OpenAI-compatible API for embeddings. Example: http://127.0.0.1:8080/v1/embeddings",
+                showIf: { backend: "Local Ollama" },
+              },
             ],
           });
         },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@saltcorn/large-language-model",
-  "version": "0.4.3",
+  "version": "0.4.5",
   "description": "Large language models and functionality for Saltcorn",
   "main": "index.js",
   "dependencies": {