@aj-archipelago/cortex 1.0.14 → 1.0.15

package/helper_apps/MediaFileChunker/docHelper.js

@@ -2,6 +2,7 @@ import pdfjsLib from 'pdfjs-dist';
 import fs from 'fs/promises';
 import mammoth from 'mammoth';
 import XLSX from 'xlsx';
+import Papa from 'papaparse';
 
 export async function txtToText(filePath) {
     const text = await fs.readFile(filePath, 'utf-8');
@@ -43,6 +44,18 @@ export async function pdfToText(filePath) {
     return finalText;
 }
 
+export async function csvToText(filePath) {
+    const text = await fs.readFile(filePath, 'utf-8');
+    const results = Papa.parse(text);
+    let finalText = '';
+
+    results.data.forEach(row => {
+        finalText += row.join(' ') + '\n';
+    });
+
+    return finalText;
+}
+
 export async function documentToText(filePath) {
     const fileExtension = filePath.split('.').pop();
 
@@ -55,6 +68,8 @@ export async function documentToText(filePath) {
             return docxToText(filePath);
         case 'xlsx':
             return xlsxToText(filePath);
+        case 'csv':
+            return csvToText(filePath);
         default:
             throw new Error(`Unsupported file type: ${fileExtension}`);
     }

package/helper_apps/MediaFileChunker/index.js

@@ -7,6 +7,7 @@ import { documentToText, easyChunker } from './docHelper.js';
 import path from 'path';
 import os from 'os';
 import { v4 as uuidv4 } from 'uuid';
+import fs from 'fs';
 
 const useAzure = process.env.AZURE_STORAGE_CONNECTION_STRING ? true : false;
 console.log(useAzure ? 'Using Azure Storage' : 'Using local file system');
@@ -38,7 +39,7 @@ async function main(context, req) {
         return
     }
 
-    const { uri, requestId } = req.body?.params || req.query;
+    const { uri, requestId, save } = req.body?.params || req.query;
     if (!uri || !requestId) {
         context.res = {
             status: 400,
@@ -63,14 +64,29 @@ async function main(context, req) {
         await publishRequestProgress({ requestId, progress, completedCount, totalCount, numberOfChunks, data });
     }
 
-    const isDocument = ['.pdf', '.txt', '.docx', '.xlsx'].some(ext => uri.toLowerCase().endsWith(ext));
+    const isDocument = ['.pdf', '.txt', '.docx', '.xlsx', '.csv'].some(ext => uri.toLowerCase().endsWith(ext));
 
     try {
         if (isDocument) {
             const extension = path.extname(uri).toLowerCase();
             const file = path.join(os.tmpdir(), `${uuidv4()}${extension}`);
-            await downloadFile(uri,file)
-            result.push(...easyChunker(await documentToText(file)));
+            await downloadFile(uri, file)
+            const text = await documentToText(file);
+            if (save) {
+                const fileName = `${uuidv4()}.txt`; // generate unique file name
+                const filePath = path.join(os.tmpdir(), fileName);
+                const tmpPath = filePath;
+                fs.writeFileSync(filePath, text); // write text to file
+        
+                // save file to the cloud or local file system
+                const saveResult = useAzure ? await saveFileToBlob(filePath, requestId) : await moveFileToPublicFolder(filePath, requestId);
+                result.push(saveResult);
+        
+                // delete temporary file
+                fs.unlinkSync(tmpPath);
+            } else {
+                result.push(...easyChunker(text));
+            }
         }else{
 
             if (isYoutubeUrl) {

package/helper_apps/MediaFileChunker/package-lock.json

@@ -18,6 +18,7 @@
         "fluent-ffmpeg": "^2.1.2",
         "ioredis": "^5.3.1",
         "mammoth": "^1.6.0",
+        "papaparse": "^5.4.1",
         "pdfjs-dist": "^3.9.179",
         "public-ip": "^6.0.1",
         "uuid": "^9.0.0",
@@ -1989,6 +1990,11 @@
       "resolved": "https://registry.npmjs.org/pako/-/pako-1.0.11.tgz",
       "integrity": "sha512-4hLB8Py4zZce5s4yd9XzopqwVv/yGNhV1Bl8NTmCq1763HeK2+EwVTv+leGeL13Dnh2wfbqowVPXCIO0z4taYw=="
     },
+    "node_modules/papaparse": {
+      "version": "5.4.1",
+      "resolved": "https://registry.npmjs.org/papaparse/-/papaparse-5.4.1.tgz",
+      "integrity": "sha512-HipMsgJkZu8br23pW15uvo6sib6wne/4woLZPlFf3rpDyMe9ywEXUsuD7+6K9PRkJlVT51j/sCOYDKGGS3ZJrw=="
+    },
     "node_modules/parseurl": {
       "version": "1.3.3",
       "resolved": "https://registry.npmjs.org/parseurl/-/parseurl-1.3.3.tgz",
@@ -4098,6 +4104,11 @@
       "resolved": "https://registry.npmjs.org/pako/-/pako-1.0.11.tgz",
       "integrity": "sha512-4hLB8Py4zZce5s4yd9XzopqwVv/yGNhV1Bl8NTmCq1763HeK2+EwVTv+leGeL13Dnh2wfbqowVPXCIO0z4taYw=="
     },
+    "papaparse": {
+      "version": "5.4.1",
+      "resolved": "https://registry.npmjs.org/papaparse/-/papaparse-5.4.1.tgz",
+      "integrity": "sha512-HipMsgJkZu8br23pW15uvo6sib6wne/4woLZPlFf3rpDyMe9ywEXUsuD7+6K9PRkJlVT51j/sCOYDKGGS3ZJrw=="
+    },
     "parseurl": {
       "version": "1.3.3",
       "resolved": "https://registry.npmjs.org/parseurl/-/parseurl-1.3.3.tgz",

package/helper_apps/MediaFileChunker/package.json

@@ -18,6 +18,7 @@
     "fluent-ffmpeg": "^2.1.2",
     "ioredis": "^5.3.1",
     "mammoth": "^1.6.0",
+    "papaparse": "^5.4.1",
     "pdfjs-dist": "^3.9.179",
     "public-ip": "^6.0.1",
     "uuid": "^9.0.0",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@aj-archipelago/cortex",
-  "version": "1.0.14",
+  "version": "1.0.15",
   "description": "Cortex is a GraphQL API for AI. It provides a simple, extensible interface for using AI services from OpenAI, Azure and others.",
   "repository": {
     "type": "git",

package/pathways/index.js

@@ -6,6 +6,7 @@ import cognitive_insert from './cognitive_insert.js';
 import cognitive_search from './cognitive_search.js';
 import complete from './complete.js';
 import entities from './entities.js';
+import language from './language.js';
 import paraphrase from './paraphrase.js';
 import sentiment from './sentiment.js';
 import summary from './summary.js';
@@ -28,6 +29,7 @@ export {
     complete,
     embeddings,
     entities,
+    language,
     paraphrase,
     sentiment,
     summary,

package/server/plugins/azureCognitivePlugin.js

@@ -2,6 +2,11 @@
 import { callPathway } from '../../lib/pathwayTools.js';
 import ModelPlugin from './modelPlugin.js';
 import { v4 as uuidv4 } from 'uuid';
+import path from 'path';
+import { config } from '../../config.js';
+import { axios } from '../../lib/request.js';
+
+const API_URL = config.get('whisperMediaApiUrl');
 
 const TOP = 1000;
 
@@ -109,6 +114,19 @@ class AzureCognitivePlugin extends ModelPlugin {
         }
     }
 
+    async markCompletedForCleanUp(requestId) {
+        try {
+            if (API_URL) {
+                //call helper api to mark processing as completed
+                const res = await axios.delete(API_URL, { params: { requestId } });
+                console.log(`Marked request ${requestId} as completed:`, res.data);
+                return res.data;
+            }
+        } catch (err) {
+            console.log(`Error marking request ${requestId} as completed:`, err);
+        }
+    }
+
     // Execute the request to the Azure Cognitive API
     async execute(text, parameters, prompt, pathwayResolver) {
         const { requestId, pathway, savedContextId, savedContext } = pathwayResolver;
@@ -118,6 +136,28 @@ class AzureCognitivePlugin extends ModelPlugin {
         url = this.ensureIndex(url, indexName);
         const headers = this.model.headers;
 
+        const { file } = parameters;
+        if(file){ 
+            let url = file;
+            //if not txt file, use helper app to convert to txt
+            const extension = path.extname(file).toLowerCase();
+            if (extension !== '.txt') {
+                try {
+                    const {data} = await axios.get(API_URL, { params: { uri: file, requestId, save: true } });
+                    url = data[0]
+                } catch (error) {
+                    console.log(`Error converting file ${file} to txt:`, error);
+                    throw error;
+                }
+            }
+ 
+            const { data } = await axios.get(url);
+            await this.markCompletedForCleanUp(requestId);
+
+            //return await this.execute(data, {...parameters, file:null}, prompt, pathwayResolver); 
+            return await callPathway(this.config, 'cognitive_insert', {...parameters, file:null, text:data });
+        }
+
         const { data, params } = await this.getRequestParameters(text, parameters, prompt, mode, indexName, savedContextId, {headers, requestId, pathway, url});
 
         // update contextid last used