update dependencies, add script to create RAG of a github repo, add API endpoint to query it, add chatbot page to interact with it

.gitignore

@@ -5,7 +5,9 @@ node_modules
 /package
 .env
 .env.*
+.envrc
 !.env.example
 vite.config.js.timestamp-*
 vite.config.ts.timestamp-*
 .vercel
+vectorstore

package.json

@@ -12,49 +12,54 @@
     "lint": "prettier --check . && eslint .",
     "format": "prettier --write .",
     "model-pipeline:run": "node scripts/model-pipeline.js",
-    "generate-embeddings": "node scripts/generate-embeddings.js"
+    "generate-embeddings": "node scripts/generate-embeddings.js",
+    "finetune": "node scripts/finetune.js"
   },
   "devDependencies": {
-    "@sveltejs/adapter-auto": "^3.0.0",
-    "@sveltejs/adapter-node": "^5.0.1",
-    "@sveltejs/kit": "^2.0.0",
-    "@sveltejs/vite-plugin-svelte": "^3.0.0",
-    "@tailwindcss/typography": "^0.5.13",
-    "@theatre/core": "^0.7.1",
-    "@theatre/studio": "^0.7.1",
-    "@threlte/theatre": "^2.1.7",
-    "@types/eslint": "^8.56.0",
+    "@sveltejs/adapter-auto": "^3.2.4",
+    "@sveltejs/adapter-node": "^5.2.2",
+    "@sveltejs/kit": "^2.5.22",
+    "@sveltejs/vite-plugin-svelte": "^3.1.1",
+    "@tailwindcss/typography": "^0.5.14",
+    "@theatre/core": "^0.7.2",
+    "@theatre/studio": "^0.7.2",
+    "@threlte/theatre": "^2.1.8",
+    "@types/eslint": "^8.56.11",
     "@types/three": "^0.159.0",
-    "@typescript-eslint/eslint-plugin": "^7.0.0",
-    "@typescript-eslint/parser": "^7.0.0",
-    "autoprefixer": "^10.4.19",
-    "daisyui": "^4.11.1",
-    "eslint": "^8.56.0",
+    "@typescript-eslint/eslint-plugin": "^7.18.0",
+    "@typescript-eslint/parser": "^7.18.0",
+    "autoprefixer": "^10.4.20",
+    "daisyui": "^4.12.10",
+    "eslint": "^8.57.0",
     "eslint-config-prettier": "^9.1.0",
-    "eslint-plugin-svelte": "^2.35.1",
+    "eslint-plugin-svelte": "^2.43.0",
     "fs-extra": "^11.2.0",
-    "postcss": "^8.4.38",
-    "prettier": "^3.1.1",
-    "prettier-plugin-svelte": "^3.1.2",
-    "svelte": "^4.2.7",
-    "svelte-check": "^3.6.0",
-    "tailwindcss": "^3.4.3",
-    "tslib": "^2.4.1",
-    "typescript": "^5.0.0",
-    "vite": "^5.0.3",
-    "vitest": "^1.2.0"
+    "postcss": "^8.4.41",
+    "prettier": "^3.3.3",
+    "prettier-plugin-svelte": "^3.2.6",
+    "svelte": "^4.2.18",
+    "svelte-check": "^3.8.5",
+    "tailwindcss": "^3.4.10",
+    "tslib": "^2.6.3",
+    "typescript": "^5.5.4",
+    "vite": "^5.4.1",
+    "vitest": "^1.6.0"
   },
   "type": "module",
   "dependencies": {
     "@dimforge/rapier3d-compat": "^0.11.2",
+    "@langchain/anthropic": "^0.3.1",
+    "@langchain/community": "^0.3.1",
+    "@langchain/openai": "^0.3.0",
+    "@langchain/textsplitters": "^0.1.0",
     "@tensorflow-models/universal-sentence-encoder": "^1.3.3",
     "@tensorflow/tfjs-node": "^4.20.0",
-    "@threlte/core": "^7.3.0",
-    "@threlte/extras": "^8.11.2",
-    "@threlte/rapier": "^2.0.0",
+    "@threlte/core": "^7.3.1",
+    "@threlte/extras": "^8.11.5",
+    "@threlte/rapier": "^2.0.1",
     "fuse.js": "^7.0.0",
     "marked": "^12.0.2",
-    "mdsvex": "^0.11.0",
+    "mdsvex": "^0.11.2",
     "three": "^0.159.0"
   }
 }

scripts/finetune.js

@@ -0,0 +1,35 @@
+// scripts/finetune.js
+import { GithubRepoLoader } from '@langchain/community/document_loaders/web/github';
+import { RecursiveCharacterTextSplitter } from '@langchain/textsplitters';
+import { OpenAIEmbeddings } from '@langchain/openai';
+import { FaissStore } from '@langchain/community/vectorstores/faiss';
+import { fileURLToPath } from 'url';
+import { dirname, join } from 'path';
+
+const __filename = fileURLToPath(import.meta.url);
+const __dirname = dirname(__filename);
+
+async function finetune() {
+	console.log('Fine-tuning model...');
+
+	const loader = new GithubRepoLoader('https://github.com/silentsilas/Authenticator', {
+		branch: 'develop',
+		recursive: true,
+		unknown: 'warn',
+		maxConcurrency: 5
+	});
+	const docs = await loader.load();
+
+	const textSplitter = new RecursiveCharacterTextSplitter({ chunkSize: 1000 });
+	const splitDocs = await textSplitter.splitDocuments(docs);
+
+	const embeddings = new OpenAIEmbeddings();
+	const vectorStore = await FaissStore.fromDocuments(splitDocs, embeddings);
+
+	const directory = join(__dirname, '..', 'vectorstore');
+	await vectorStore.save(directory);
+
+	console.log('Fine-tuning complete. Vector store saved.');
+}
+
+finetune();

src/hooks.server.ts

@@ -1,5 +1,7 @@
 import { getModel } from '$lib/utils/search';
 import { building } from '$app/environment';
+import type { Handle } from '@sveltejs/kit';
+import { v4 as uuidv4 } from 'uuid';
 
 if (!building) {
 	getModel().catch((error) => {
@@ -8,3 +10,18 @@ if (!building) {
 
 	console.log('Model loaded successfully!');
 }
+
+export const handle: Handle = async ({ event, resolve }) => {
+	// Check for existing session
+	let sessionId = event.cookies.get('sessionId');
+
+	// If no session exists, create a new one
+	if (!sessionId) {
+		sessionId = uuidv4();
+		event.cookies.set('sessionId', sessionId, { path: '/', httpOnly: true, sameSite: 'strict', maxAge: 60 * 60 * 24 * 7 }); // 1 week
+	}
+	// Add sessionId to locals for easy access in routes
+	event.locals = { ...event.locals, sessionId };
+
+	return resolve(event);
+};

src/routes/ai/+layout.svelte

@@ -0,0 +1,9 @@
+<script lang="ts">
+	import SearchResults from '$lib/components/SearchResults.svelte';
+	import AppContainer from '$lib/components/scenes/app/AppContainer.svelte';
+</script>
+
+<AppContainer>
+	<slot />
+	<SearchResults />
+</AppContainer>

src/routes/ai/+layout.ts

@@ -0,0 +1,2 @@
+export const prerender = false;
+export const ssr = false;

src/routes/ai/+page.svelte

@@ -0,0 +1,141 @@
+<script lang="ts">
+	import '../../app.css';
+	import type { SearchResult } from '$lib/utils/search';
+	import { searchResults } from '$lib/store';
+	import { onMount } from 'svelte';
+	import { marked } from 'marked';
+	import { HumanMessage, AIMessage } from '@langchain/core/messages';
+	import type { ChatHistory } from '../api/ai/+server';
+	import { PUBLIC_LOAD_DUMMY_HISTORY } from '$env/static/public';
+
+	let searchResultsValue: SearchResult[] = [];
+	let query = '';
+	let loading = false;
+	function generateDummyHistory() {
+		return [
+			JSON.parse(JSON.stringify(new HumanMessage({ content: 'Hello, AI!' }))),
+			JSON.parse(JSON.stringify(new AIMessage({ content: 'Hello! How can I assist you today?' }))),
+			JSON.parse(JSON.stringify(new HumanMessage({ content: "What's the weather like?" }))),
+			JSON.parse(
+				JSON.stringify(
+					new AIMessage({
+						content:
+							"I'm sorry, but I don't have access to real-time weather information. You might want to check a weather app or website for the most up-to-date forecast."
+					})
+				)
+			)
+		];
+	}
+
+	let chatHistory: ChatHistory = [];
+
+	onMount(async () => {
+		try {
+			if (PUBLIC_LOAD_DUMMY_HISTORY === 'true') {
+				chatHistory = generateDummyHistory();
+				return;
+			}
+			const response = await fetch('/api/ai');
+			const data = await response.json();
+			chatHistory = data.chatHistory || [];
+		} catch (error) {
+			console.error('Error fetching chat history:', error);
+		}
+	});
+
+	searchResults.subscribe((value: SearchResult[]) => {
+		searchResultsValue = value ? value : [];
+	});
+
+	function getRoleAndContent(message: any): { role: string; content: string } {
+		if (message.type === 'constructor') {
+			const messageType = message.id[2];
+			return {
+				role: messageType.replace('Message', '').toLowerCase(),
+				content: message.kwargs.content
+			};
+		}
+		return {
+			role: 'unknown',
+			content: JSON.stringify(message)
+		};
+	}
+
+	// safely render markdown
+	function renderMarkdown(content: string) {
+		return marked(content);
+	}
+
+	async function handleSubmit() {
+		loading = true;
+		try {
+			const response = await fetch('/api/ai', {
+				method: 'POST',
+				headers: {
+					'Content-Type': 'application/json'
+				},
+				body: JSON.stringify({ query })
+			});
+			const data = await response.json();
+			chatHistory = data.chatHistory || [];
+
+			query = '';
+		} catch (error) {
+			console.error('Error fetching AI response:', error);
+		} finally {
+			loading = false;
+		}
+	}
+</script>
+
+<svelte:head>
+	<title>silentsilas - AI</title>
+</svelte:head>
+
+{#if searchResultsValue.length === 0}
+	<div class="flex-grow flex-col overflow-auto p-4">
+		<div class="space-y-4">
+			{#each chatHistory as message}
+				{@const { role, content } = getRoleAndContent(message)}
+				{#if role === 'human'}
+					<div class="chat chat-end">
+						<div class="chat-bubble chat-bubble-primary">
+							{@html renderMarkdown(content)}
+						</div>
+					</div>
+				{:else}
+					<div class="p-4 rounded-lg bg-base-300 prose md:container">
+						{@html renderMarkdown(content)}
+					</div>
+				{/if}
+			{/each}
+		</div>
+
+		{#if loading}
+			<div class="mt-4">
+				<span class="loading loading-dots loading-lg"></span>
+			</div>
+		{/if}
+		<form on:submit|preventDefault={handleSubmit} class="mt-4 flex-col">
+			<label class="form-control">
+				<div class="label">
+					<span class="label-text">
+						Querying the Authenticator <a
+							href="https://git.silentsilas.com/silentsilas/Authenticator"
+							target="_blank"
+							class="link-primary">repository</a
+						>
+					</span>
+				</div>
+				<textarea
+					bind:value={query}
+					class="textarea textarea-bordered h-24"
+					placeholder="Type your message here..."
+				></textarea>
+			</label>
+			<button type="submit" class="btn btn-block btn-primary mt-2" disabled={loading}>
+				Send
+			</button>
+		</form>
+	</div>
+{/if}

src/routes/api/ai/+server.ts

@@ -0,0 +1,82 @@
+import { json } from '@sveltejs/kit';
+import { FaissStore } from '@langchain/community/vectorstores/faiss';
+import { OpenAIEmbeddings } from '@langchain/openai';
+import { ChatAnthropic } from '@langchain/anthropic';
+import { RunnableSequence, RunnablePassthrough } from '@langchain/core/runnables';
+import { StringOutputParser } from '@langchain/core/output_parsers';
+import { ChatPromptTemplate } from '@langchain/core/prompts';
+import { join } from 'path';
+import { HumanMessage, AIMessage, SystemMessage } from '@langchain/core/messages';
+import type { RequestEvent } from '@sveltejs/kit';
+
+const chatHistories: Record<string, ChatHistory> = {};
+
+type VectorDocument = {
+	pageContent: string;
+};
+
+const formatDocumentsAsString = (documents: VectorDocument[]) => {
+	return documents.map((doc) => doc.pageContent).join('\n\n');
+};
+
+export type ChatHistory = HumanMessage[] | AIMessage[] | SystemMessage[];
+
+export async function POST({ request, locals }: RequestEvent): Promise<Response> {
+	const { query } = await request.json();
+	const sessionId = locals.sessionId;
+
+	if (!chatHistories[sessionId]) {
+		chatHistories[sessionId] = [];
+	}
+	const chatHistory = chatHistories[sessionId];
+
+	const directory = join(process.cwd(), 'vectorstore');
+	const embeddings = new OpenAIEmbeddings();
+	const vectorStore = await FaissStore.load(directory, embeddings);
+	const vectorStoreRetriever = vectorStore.asRetriever();
+
+	const model = new ChatAnthropic({
+		modelName: 'claude-3-5-sonnet-20240620',
+		anthropicApiKey: process.env.ANTHROPIC_API_KEY
+	});
+
+	const SYSTEM_TEMPLATE = `Use the following pieces of context to answer the question at the end.
+If you don't know the answer, just say that you don't know, don't try to make up an answer.
+
+----------------
+
+Context:
+{context}
+
+Question:
+{question}`;
+
+	const prompt = ChatPromptTemplate.fromMessages([
+		['system', SYSTEM_TEMPLATE],
+		...chatHistory,
+		['user', '{question}']
+	]);
+
+	const chain = RunnableSequence.from([
+		{
+			context: vectorStoreRetriever.pipe(formatDocumentsAsString),
+			question: new RunnablePassthrough(),
+		},
+		prompt,
+		model,
+		new StringOutputParser()
+	]);
+
+	const answer = await chain.invoke(query);
+
+	chatHistory.push(new HumanMessage({ content: query }));
+	chatHistory.push(new AIMessage({ content: answer }));
+
+	return json({ response: answer, chatHistory });
+}
+
+export async function GET({ locals }): Promise<Response> {
+	const sessionId = locals.sessionId;
+	const chatHistory = chatHistories[sessionId] || [];
+	return json({ chatHistory });
+}

src/routes/projects/+layout.svelte

@@ -1,6 +1,4 @@
 <script lang="ts">
-	import Footer from '$lib/components/Footer.svelte';
-	import NavBar from '$lib/components/NavBar.svelte';
 	import SearchResults from '$lib/components/SearchResults.svelte';
 	import AppContainer from '$lib/components/scenes/app/AppContainer.svelte';
 </script>

svelte.config.js

@@ -6,10 +6,12 @@ import { mdsvex } from 'mdsvex';
 const config = {
 	// Consult https://kit.svelte.dev/docs/integrations#preprocessors
 	// for more information about preprocessors
-	preprocess: [vitePreprocess(), mdsvex({ extensions: ['.md'], layout: {
+	preprocess: [vitePreprocess(), mdsvex({
+		extensions: ['.md'], layout: {
 			poetry: './src/lib/utils/PoetryLayout.svelte',
 			thoughts: './src/lib/utils/ThoughtsLayout.svelte',
-}})],
+		}
+	})],
 
 
 	kit: {