fixing-examples

JS/edgechains/examples/chat-with-llm/dist/index.js

@@ -1,7 +1,8 @@
+//@ts-ignore
 import { ArakooServer } from "@arakoodev/edgechains.js/arakooserver";
 import Jsonnet from "@arakoodev/jsonnet";
 //@ts-ignore
-import createClient from "sync-rpc";
+import createClient from 'sync-rpc';
 import { fileURLToPath } from "url";
 import path from "path";
 const server = new ArakooServer();

JS/edgechains/examples/chat-with-llm/dist/lib/generateResponse.cjs

@@ -38,7 +38,8 @@ function openAICall() {
             return openai.zodSchemaResponse({ prompt, schema: schema }).then((res) => {
                 return JSON.stringify(res);
             });
-        } catch (error) {
+        }
+        catch (error) {
             return error;
         }
     };

JS/edgechains/examples/chat-with-llm/package.json

@@ -1,22 +1,22 @@
 {
-    "name": "ownChatGpt",
-    "version": "1.0.0",
-    "description": "",
-    "main": "index.js",
-    "type": "module",
-    "keywords": [],
-    "author": "",
-    "scripts": {
-        "start": "tsc && node --experimental-wasm-modules ./dist/index.js"
-    },
-    "license": "ISC",
-    "dependencies": {
-        "@arakoodev/edgechains.js": "^0.1.23",
-        "@arakoodev/jsonnet": "^0.2.1",
-        "sync-rpc": "^1.3.6",
-        "zod": "^3.23.8"
-    },
-    "devDependencies": {
-        "@types/node": "^20.12.12"
-    }
+  "name": "ownchatgpt",
+  "version": "1.0.0",
+  "description": "",
+  "main": "index.js",
+  "type": "module",
+  "keywords": [],
+  "author": "",
+  "scripts": {
+    "start": "tsc && node --experimental-wasm-modules ./dist/index.js"
+  },
+  "license": "ISC",
+  "dependencies": {
+    "@arakoodev/edgechains.js": "^0.1.23",
+    "@arakoodev/jsonnet": "^0.3.1",
+    "sync-rpc": "^1.3.6",
+    "zod": "^3.23.8"
+  },
+  "devDependencies": {
+    "@types/node": "^20.12.12"
+  }
 }

JS/edgechains/examples/chat-with-pdf/jsonnet/main.jsonnet

@@ -0,0 +1,26 @@
+local CUSTOM_TEMPLATE = |||
+                            You are a senior software developer seeking accurate answers based on provided content. Your task is to find answers exclusively from the given content and the answer should be verbose and detailedfull. If the answer is not present within the provided content, respond with "Sorry! I don't know the answer.
+                            Content:{content}
+                            Question:{}
+                        |||;
+
+
+local updateQueryPrompt(promptTemplate, query, content) =
+    local updatedPrompt = std.strReplace(promptTemplate, '{}', query + "\n");
+    local updatedContent = std.strReplace(updatedPrompt, '{content}', content + "\n");
+    updatedContent;
+
+
+local query = std.extVar("query");
+local content = std.extVar("content");
+
+local getQueryMatch(query)= 
+    local queryEmbeddings = arakoo.native("getEmbeddings")(query);
+    local content = arakoo.native("getQueryMatch")(queryEmbeddings);
+    content;
+
+local updatedQueryPrompt = updateQueryPrompt(CUSTOM_TEMPLATE, query, getQueryMatch(query));
+
+local getOpenAiResponse = arakoo.native("openAICall")(updatedQueryPrompt);
+
+getOpenAiResponse

JS/edgechains/examples/chat-with-pdf/package.json

@@ -14,10 +14,11 @@
     "license": "ISC",
     "dependencies": {
         "@arakoodev/edgechains.js": "^0.1.23",
-        "@arakoodev/jsonnet": "^0.2.0",
+        "@arakoodev/jsonnet": "^0.3.1",
         "@babel/preset-env": "^7.24.4",
         "cli-spinner": "^0.2.10",
         "regenerator-runtime": "^0.14.1",
+        "sync-rpc": "^1.3.6",
         "zod": "^3.23.8"
     },
     "devDependencies": {

JS/edgechains/examples/chat-with-pdf/src/lib/InsertToSupabase.ts

@@ -0,0 +1,64 @@
+import { fileURLToPath } from "url";
+import { Spinner } from "cli-spinner"
+import { PdfLoader } from "@arakoodev/edgechains.js/document-loader";
+import { TextSplitter } from "@arakoodev/edgechains.js/splitter";
+import { createRequire } from 'module';
+import { Supabase } from "@arakoodev/edgechains.js/vector-db";
+import Jsonnet from "@arakoodev/jsonnet";
+import path from "path"
+import fs from "fs";
+
+const require = createRequire(import.meta.url)
+
+const getEmbeddings = require("./getEmbeddings.cjs")
+
+const __dirname = fileURLToPath(import.meta.url);
+
+const pdfPath = path.join(__dirname, "../../../example.pdf");
+const pdfData = fs.readFileSync(pdfPath);
+const bufferPdf = Buffer.from(pdfData);
+const loader = new PdfLoader(bufferPdf);
+const docs = await loader.loadPdf();
+const splitter = new TextSplitter();
+export const splitedDocs = await splitter.splitTextIntoChunks(docs, 500);
+
+const secretsPath = path.join(__dirname, "../../../jsonnet/secrets.jsonnet");
+
+const jsonnet = new Jsonnet();
+
+const secretsLoader = jsonnet.evaluateFile(secretsPath);
+const supabaseApiKey = await JSON.parse(secretsLoader).supabase_api_key;
+const supabaseUrl = await JSON.parse(secretsLoader).supabase_url;
+const supabase = new Supabase(supabaseUrl, supabaseApiKey);
+const client = supabase.createClient();
+
+export async function InsertToSupabase(content:any) {
+    var spinner = new Spinner("Inserting to Supabase.. %s");
+    try {
+        spinner.setSpinnerString("|/-\\");
+        spinner.start();
+
+        const response = await getEmbeddings()(content);
+        for (let i = 0; i < response?.length; i++) {
+            if (content[i].length <= 1) {
+                continue;
+            }
+
+            const element = response[i].embedding;
+            await supabase.insertVectorData({
+                client,
+                tableName: "documents",
+                content: content[i].toLowerCase(),
+                embedding: element,
+            });
+        }
+        if (!response) {
+            return console.log("Error inserting to Supabase");
+        }
+        console.log("Inserted to Supabase");
+    } catch (error) {
+        console.log("Error inserting to Supabase", error);
+    } finally {
+        spinner.stop();
+    }
+}

JS/edgechains/examples/chat-with-pdf/src/lib/generateResponse.cts

@@ -0,0 +1,29 @@
+const path = require("path");
+const { OpenAI } = require("@arakoodev/edgechains.js/openai");
+const z = require("zod");
+const Jsonnet = require("@arakoodev/jsonnet");
+const jsonnet = new Jsonnet();
+
+
+const secretsPath = path.join(__dirname, "../../jsonnet/secrets.jsonnet");
+const openAIApiKey = JSON.parse(jsonnet.evaluateFile(secretsPath)).openai_api_key;
+
+const openai = new OpenAI({ apiKey: openAIApiKey });
+
+const schema = z.object({
+    answer: z.string().describe("The answer to the question"),
+});
+
+function openAICall() {
+    return function (prompt: string) {
+        try {
+            return openai.zodSchemaResponse({ prompt, schema }).then((res: any) => {
+                return JSON.stringify(res);
+            });
+        } catch (error) {
+            return error;
+        }
+    };
+}
+
+module.exports = openAICall;

JS/edgechains/examples/chat-with-pdf/src/lib/getEmbeddings.cts

@@ -0,0 +1,23 @@
+const { OpenAI } = require("@arakoodev/edgechains.js/openai")
+const path = require("path");
+const Jsonnet = require("@arakoodev/jsonnet");
+
+const jsonnet = new Jsonnet();
+
+const secretsPath = path.join(__dirname, "../../jsonnet/secrets.jsonnet");
+const openAIApiKey = JSON.parse(jsonnet.evaluateFile(secretsPath)).openai_api_key;
+
+const llm = new OpenAI({
+    apiKey: openAIApiKey
+});
+
+function getEmbeddings() {
+    return ((content: any) => {
+        const embeddings = llm.generateEmbeddings(content).then((res: any) => {
+            return JSON.stringify(res)
+        })
+        return embeddings;
+    })
+}
+
+module.exports = getEmbeddings;

JS/edgechains/examples/chat-with-pdf/src/lib/getQueryMatch.cts

@@ -0,0 +1,37 @@
+import path from "path"
+const Jsonnet = require("@arakoodev/jsonnet");
+const { Supabase } = require("@arakoodev/edgechains.js/vector-db");
+
+const secretsPath = path.join(__dirname, "../../jsonnet/secrets.jsonnet");
+
+const jsonnet = new Jsonnet();
+
+const secretsLoader = jsonnet.evaluateFile(secretsPath);
+const supabaseApiKey = JSON.parse(secretsLoader).supabase_api_key;
+const supabaseUrl = JSON.parse(secretsLoader).supabase_url;
+
+const supabase = new Supabase(supabaseUrl, supabaseApiKey);
+const client = supabase.createClient();
+
+function getQueryMatch() {
+    return function (embeddings: any) {
+        const response = supabase.getDataFromQuery({
+            client,
+            functionNameToCall: "match_documents",
+            query_embedding: JSON.parse(embeddings)[0].embedding,
+            similarity_threshold: 0.5,
+            match_count: 1,
+        }).then((response:any) => {
+            const contentArr: string[] = [];
+            for (let i = 0; i < response?.length; i++) {
+                const element = response[i];
+                contentArr.push(element.content);
+            }
+            return contentArr.join(" ");
+        })
+        return response;
+    }
+}
+
+module.exports = getQueryMatch
+