@saltcorn/large-language-model 0.9.10 → 0.9.12

package/generate.js

@@ -11,6 +11,7 @@ const { google } = require("googleapis");
 const Plugin = require("@saltcorn/data/models/plugin");
 const File = require("@saltcorn/data/models/file");
 const path = require("path");
+const fs = require("fs");
 const { features, getState } = require("@saltcorn/data/db/state");
 const {
   generateText,
@@ -21,7 +22,10 @@ const {
   embedMany,
   experimental_transcribe,
 } = require("ai");
-const { openai, createOpenAI } = require("@ai-sdk/openai");
+const { createOpenAI } = require("@ai-sdk/openai");
+const OpenAI = require("openai");
+const { ElevenLabsClient } = require("@elevenlabs/elevenlabs-js");
+
 let ollamaMod;
 if (features.esm_plugins) ollamaMod = require("ollama");
 
@@ -119,7 +123,46 @@ const getAudioTranscription = async (
   { backend, apiKey, api_key, provider, ai_sdk_provider },
   opts
 ) => {
-  switch (backend) {
+  switch (opts.backend || backend) {
+    case "ElevenLabs":
+      const transcription = await new ElevenLabsClient({
+        apiKey: opts?.api_key || api_key || apiKey,
+      }).speechToText.convert({
+        file: await (await File.findOne(opts.file)).get_contents(),
+        modelId: opts.model || "scribe_v2", // Model to use
+        tagAudioEvents: true, // Tag audio events like laughter, applause, etc.
+        languageCode: opts.languageCode || "eng", // Language of the audio file. If set to null, the model will detect the language automatically.
+        numSpeakers: opts.numSpeakers || null, // Language of the audio file. If set to null, the model will detect the language automatically.
+        diarize: !!opts.diarize, // Whether to annotate who is speaking
+        diarizationThreshold: opts.diarizationThreshold || null
+      });
+      return transcription;
+    case "OpenAI":
+      const client = new OpenAI({
+        apiKey: opts?.api_key || api_key || apiKey,
+      });
+      const fp = opts.file.location
+        ? opts.file.location
+        : typeof opts.file === "string"
+        ? await (
+            await File.findOne(opts.file)
+          ).location
+        : null;
+      const model = opts?.model || "whisper-1";
+      const diarize = model === "gpt-4o-transcribe-diarize";
+      const transcript1 = await client.audio.transcriptions.create({
+        file: Buffer.isBuffer(opts.file) ? opts.file : fs.createReadStream(fp),
+
+        model,
+        ...(diarize
+          ? {
+              response_format: "diarized_json",
+              chunking_strategy: "auto",
+            }
+          : {}),
+      });
+      return transcript1;
+
     case "AI SDK":
       const api_Key = opts?.api_key || api_key || apiKey;
       const prov_obj = createOpenAI({ apiKey: api_Key });

package/index.js

@@ -687,7 +687,6 @@ module.exports = {
               name: "model",
               label: "The model name, for example <code>whisper-1</code>",
               type: "String",
-              required: true,
             },
             {
               name: "prompt_template",
@@ -729,7 +728,6 @@ module.exports = {
               name: "model",
               label: "The model name, for example <code>whisper-1</code>",
               type: "String",
-              required: true,
             },
             {
               name: "prompt_template",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@saltcorn/large-language-model",
-  "version": "0.9.10",
+  "version": "0.9.12",
   "description": "Large language models and functionality for Saltcorn",
   "main": "index.js",
   "dependencies": {
@@ -12,7 +12,9 @@
     "@google-cloud/aiplatform": "^3.34.0",
     "googleapis": "^144.0.0",
     "ai": "5.0.44",
-    "@ai-sdk/openai": "2.0.30"
+    "@ai-sdk/openai": "2.0.30",
+    "openai": "6.16.0",
+    "@elevenlabs/elevenlabs-js": "2.31.0"
   },
   "author": "Tom Nielsen",
   "license": "MIT",