Update README.md

Signed-off-by: Andreia Ocănoaia <[email protected]>

README.md

@@ -0,0 +1,123 @@
+# Generative AI - RAG Workshop
+
+This repository contains the source code for the workshop ** ** held at [C3 Festival 2024](https://c3fest.com/).
+
+The application is a RAG application that takes as an input the user description and interests and generates a list of must-see speakers from C3 Festival.
+
+Feel free to fork it and use it as a template for your own projects.
+
+TODO: Add gitpod and codespaces buttons:
+
+
+<div style="display: flex; align-items: center;">
+  <a href="https://gitpod.io/#https://github.com/andreia-oca/rag-workshop" style="margin-right: 10px;">
+    <img src="https://gitpod.io/button/open-in-gitpod.svg" alt="Open in Gitpod" style="height: 30px;"/>
+  </a>
+  <a href="https://open.vscode.dev/andreia-oca/rag-workshop" style="margin-right: 10px;">
+    <img src="https://img.shields.io/static/v1?label=Open%20in%20VSCode&message=Open&color=blue&logo=visualstudiocode" alt="Open in Visual Studio Code" style="height: 30px;"/>
+  </a>
+  <a href="https://codespaces.new/andreia-oca/rag-workshop">
+    <img src="https://github.com/codespaces/badge.svg" alt="Open in GitHub Codespaces" style="height: 30px;"/>
+  </a>
+</div>
+
+## Step-by-step tutorial
+
+LLMs challenges are:
+1. **Adding new or proprietary data** - LLMs are trained on public tokens (public data from the Internet) up to a certain date. If you want to add new data, you need to retrain the model or to provide the information in the prompt.
+2. **Costs** - each prompt cost increases with the number of tokens in the prompt. The longer the prompt, the more expensive it is.
+
+The answer to these challenges is to implement a RAG application over a pre-trained LLM. The RAG application is a two-step process:
+1. **Retrieval** - the application retrieves the most relevant documents from a vector store based on the user query.
+2. **Generation** - the application generates the answer based on the retrieved documents.
+
+### Prerequisites
+
+Create a `server/.env` file to export the OPENAI_API_KEY:
+
+```bash
+# .env
+OPENAI_API_KEY="your_openai_api_key"
+```
+
+### Create a vector store
+
+The vector store is a fancy storage for the proprietary data.
+
+TODO - Add a diagram
+
+Load the data into documents:
+
+```typescript
+    // Document loading
+    const loader = new TextLoader("./data/talks.txt");
+    const raw_documents = await loader.load();
+    console.log(raw_documents.length);
+    console.log(raw_documents[0].length);
+```
+
+Split the data into more manageable chunks:
+
+```typescript
+    // Document splitting
+    const splitter = new RecursiveCharacterTextSplitter({
+        separators: ["\n\n", "\n", ",", " ", ""],
+        chunkSize: 1024,
+        chunkOverlap: 256,
+    });
+
+    const documents = await splitter.splitDocuments(raw_documents);
+    console.log(documents.length);
+```
+
+Create the vector store/vector database:
+
+```typescript
+    // Use the OpenAIEmbeddings model to create embeddings from text
+    const embeddings = new OpenAIEmbeddings({openAIApiKey: OPENAI_API_KEY});
+
+    // Create the database directory if it doesn't exist
+    if (!fs.existsSync(DATABASE_PATH)) {
+      try {
+        fs.mkdirSync(DATABASE_PATH);
+      } catch (e) {
+        console.error(`Error creating directory '${DATABASE_PATH}':`, e);
+      }
+    }
+
+    // Connect to the vector store
+    const db = await lancedb.connect(DATABASE_PATH);
+```
+
+Set the vector schema - highly dependant on how the documents metadata looks like:
+
+```typescript
+    const table = await db.createTable(
+      "vectors",
+      [
+        {
+          vector: await embeddings.embedQuery("string"),
+          text: "contents",
+          source: "filename",
+          loc: { lines: { from: 0, to: 0 } },
+        },
+      ],
+      { writeMode: lancedb.WriteMode.Overwrite }
+    );
+```
+
+Save the embeddings in the vector store:
+
+```typescript
+    // Save the data as OpenAI embeddings in a table
+    await LanceDB.fromDocuments(documents, embeddings, { table });
+```
+
+### Retrieve data based on a question
+
+TODO: Add diagram
+
+
+### Easily deploy with Genezio
+
+