Added embedding + QA examples using simple file system vector db

Author: ezimuel

data/.gitignore

@@ -0,0 +1 @@
+vectordb.json

data/Uranus_Neptune_moons.pdf

@@ -0,0 +1,43 @@
+<?php
+/**
+ * Embedding with Ollama (Llama 3)
+ */
+require dirname(dirname(dirname(__DIR__))) . '/vendor/autoload.php';
+
+use LLPhant\Chat\OllamaChat;
+use LLPhant\Embeddings\DataReader\FileDataReader;
+use LLPhant\Embeddings\DocumentSplitter\DocumentSplitter;
+use LLPhant\Embeddings\EmbeddingGenerator\Ollama\OllamaEmbeddingGenerator;
+use LLPhant\Embeddings\VectorStores\FileSystem\FileSystemVectorStore;
+use LLPhant\OllamaConfig;
+
+# You can run ollama locally and install LLama3 from here: https://ollama.com/library/llama3
+$config = new OllamaConfig();
+$config->model = 'llama3';
+$chat = new OllamaChat($config);
+
+# Read PDF file
+printf ("- Reading the PDF files\n");
+$pdfPath = __DIR__ . '/../../../data/Uranus_Neptune_moons.pdf';
+$reader = new FileDataReader($pdfPath);
+$documents = $reader->getDocuments();
+printf("Number of PDF files: %d\n", count($documents));
+
+# Document split
+printf("- Document split\n");
+$splitDocuments = DocumentSplitter::splitDocuments($documents, 1000);
+printf("Number of splitted documents (chunk): %d\n", count($splitDocuments));
+
+# Embedding
+printf("- Embedding\n");
+$embeddingGenerator = new OllamaEmbeddingGenerator($config);
+$embeddedDocuments = $embeddingGenerator->embedDocuments($splitDocuments);
+
+# File system vector store
+printf("- Index all the embeddings using the file system\n");
+$vectorStorePath = __DIR__ . '/../../../data/vectordb.json';
+$store = new FileSystemVectorStore($vectorStorePath);
+
+$store->addDocuments($embeddedDocuments);
+
+printf("Added %d documents in %s\n", count($embeddedDocuments), realpath($vectorStorePath));

src/rag/ollama/embedding_file.php

@@ -0,0 +1,38 @@
+<?php
+/**
+ * RAG architecture with Ollama and Elasticsearch
+ */
+require dirname(dirname(dirname(__DIR__))) . '/vendor/autoload.php';
+
+use LLPhant\Chat\OllamaChat;
+use LLPhant\Embeddings\EmbeddingGenerator\Ollama\OllamaEmbeddingGenerator;
+use LLPhant\Embeddings\VectorStores\FileSystem\FileSystemVectorStore;
+use LLPhant\OllamaConfig;
+use LLPhant\Query\SemanticSearch\QuestionAnswering;
+
+# Ollama with Llama3
+$config = new OllamaConfig();
+$config->model = 'llama3';
+$chat = new OllamaChat($config);
+
+# Embedding
+$embeddingGenerator = new OllamaEmbeddingGenerator($config);
+
+# File system vector store
+$vectorStorePath = __DIR__ . '/../../../data/vectordb.json';
+$store = new FileSystemVectorStore($vectorStorePath);
+
+# RAG
+$qa = new QuestionAnswering(
+    $store,
+    $embeddingGenerator,
+    $chat
+);
+
+$answer = $qa->answerQuestion('How many moons has Neptune?');
+printf("-- Answer:\n%s\n", $answer);
+
+foreach ($qa->getRetrievedDocuments() as $doc) {
+    printf("-- Document: %s\n", $doc->sourceName);
+    printf("-- Content (%d characters): %s\n", strlen($doc->content), $doc->content);
+}